Modulation of the Visual Retrieval System in Writing: Functional MRI Study on the Japanese

Kimihiro Nakamura1, Manabu Honda2, Shigeru Hirano1,

Tatsuhide Oga1, Nobukatsu Sawamoto1, Takashi Hanakawa1, Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 Hiroshi Inoue3, Jin Ito3, Tetsu Matsuda1, Hidenao Fukuyama1, and Hiroshi Shibasaki1

Abstract & We used functional magnetic resonance imaging (fMRI) to left sensorimotor areas and right cerebellum. The versus examine whether the act of writing involves different neuro- comparison showed increased responses in the left psychological mechanisms between the two systems of prefrontal and anterior cingulate areas. Especially, the lPITC the : kanji () and kana (phono- showed a significant task-by-script interaction. Two additional gram). The main experiments employed a 2 Â 2 factorial control tasks, repetition (REP) and semantic judgment (SJ), design that comprised writing-to-dictation and visual mental activated the bilateral perisylvian areas, but enhanced the lPITC recall for kanji and kana. For both scripts, the actual writing response only weakly. These results suggest that writing of the produced a widespread fronto-parietal activation in the left ideographic and phonographic scripts, although using the hemisphere. Especially, writing of kanji activated the left largely same cortical regions, each modulates the visual word- posteroinferior temporal cortex (lPITC), whereas that of retrieval system according to their graphic features. Further- kana also yielded a trend of activation in the same area. more, comparisons with two additional tasks indicate that the Mental recall for both scripts activated similarly the left parieto- activity of the lPITC increases especially in expressive language temporal regions including the lPITC. The writing versus operations regardless of sensory modalities of the input mental recall comparison revealed greater activations in the stimulus. &

INTRODUCTION Among a body of relating kanji–kana disso- Neurolinguistic studies of Japanese have centered on ciation in brain-damaged patients, kanji-selective agra- the question that neural processing of written language phia in left posterior temporal lesions constitutes a may differ among different script systems, for brain distinct neuropsychological symptom that has been lesions may affect or writing abilities differently described by many investigators (Hamasaki et al., 1995; between the ideographic (kanji) and phonographic (ka- Sakurai, , Sakuta, & Iwata, 1994; Yokota, Ishiai, na) scripts, which is a phenomenon seemingly unique to Furukawa, & Tsukagoshi, 1990; Soma et al., 1989; Kawa- this particular language (.g., Mayeux & Kandel, 1991). hata, Nagata, & Shishido, 1988; Mochizuki & Ohtomo, This ideogram–phonogram dissociation in Japanese, 1988; Kawamura, Hirayama, Hasegawa, Takahashi, & however, may simply represent a ‘‘phenotypic’’ variant Yamamura, 1987). As suggested by the observation that derived from dysfunction of common neuropsycholog- the agraphic patients complain of ‘‘forgetting’’ kanji, the ical mechanisms that should work universally in the symptom has been interpreted as a disorder of mental human brain for processing of written . This is recall of visual graphic images required for normal especially the case for certain forms of the dissociation writing of kanji (Iwata, 1984). The critical lesions for that are thought to have equivalent language disorders producing the symptom converge on Brodmann’s area in Western languages (e.g., Soma, Sugishita, Kitamura, 37 in the left posteroinferior temporal cortex (lPITC) Maruyama, & Imanaga, 1989). (Soma et al., 1989). Conversely, kana-selective agraphia, the other form of the dissociative agraphia, may appear in parietal or frontal lesions (Sakurai, Matsumura, Iwat- 1Kyoto University, 2National Institute for Physiological subo, & Momose, 1997; Tanaka, Yamadori, & Murata, Sciences, Okazaki, , 3Rakuwa-kai Otowa Hospital, , 1987), but none of the reported cases was associated Japan with lPITC damage. This may support the idea that

D 2002 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 14:1, pp. 104–115

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 writing of kana involves a different neuropsychological the standard brain template. Lateralized activation of the mechanism from that of kanji; people may spell out left posterior inferior temporal gyrus was commonly words in the phonographic script by converting their found in the iWD, iMR, and pMR tasks. However, the acoustic images directly to written forms, rather than by pWD task also showed a trend of activation in this area visualizing their graphic forms, because each kana char- (Z = 3.95, uncorrected). acter, having a simple visual structure, has a one-to-one The iWD task yielded extensive activations in the left correspondence with a syllable. A different, fronto-pari- fronto-parietal areas including the middle and inferior etal neural network has been thought to work for this frontal gyri and pre- and postcentral gyri. Bilateral process (Iwata, 1984). activations were found in the medial frontal gyrus, Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 We previously demonstrated, using functional mag- cingulate cortex, superior temporal gyrus, superior pa- netic resonance imaging (fMRI), that activity of the rietal lobule, inferior parietal sulcus, basal ganglia, and lPITC increases in both actual writing and visual cerebellum. The pWD task activated the inferior frontal mental recall of in normal subjects (Naka- gyrus, sensorimotor cortices, superior parietal lobule, mura et al., 2000). This result supported Iwata’s and inferior parietal sulcus in the left hemisphere. (1984) proposal that the region participates in writing Bilateral activations were observed in the superior tem- of kanji as a storehouse of visual graphic memory. poral gyrus, medial frontal gyrus, and cingulate cortex. However, despite the ample lesion studies, there has There were also activations in the left basal ganglia and been neuroimaging evidence that the area acts thalamus and right cerebellum. In the iMR task, there differently between the ideographic and phonographic were bilateral activations in the middle and inferior scripts during the process of normal writing. To frontal gyrus, superior temporal gyrus, inferior parietal address the issue, the present fMRI study employed sulcus, and superior parietal lobule. In the pMR task, a2Â 2 factorial design in which the effects of interest bilateral activations were observed in the cingulate were task (actual writing vs. mental recall) and script cortices, superior temporal gyrus, and inferior parietal type (kanji vs. kana). Since previous neuroimaging sulci. There was also an activation site in the left inferior studies have shown that the lPITC is active when frontal gyrus. Other activation sites included the right people are engaged in visual imagery of letter forms basal ganglia and cerebellar hemisphere and upper regardless of the writing systems used (Nakamura brainstem. et al., 2000; Goldenberg et al., 1989), it is expected In the individual-based analysis, the iWD and pWD that the mental recall of letter images should activate task commonly activated the left fronto-parietal area, the area similarly for kanji and kana. In actual writing, bilateral fronto-temporal junction, left inferior parietal however, the activity of the lPITC may differ between sulcus, left basal nuclei, and right cerebellum in all the the script types, resulting in a task-by-script interaction subjects. The iWD and pWD tasks yielded activations in in this area. In addition, we tested two different the left dorsolateral prefrontal cortex in eight and five behavioral tasks using the same word stimuli as con- subjects, respectively. In both tasks, activation of the left trols for a baseline activation that may reflect early cingulate cortex and medial frontal gyrus and bilateral auditory–verbal processing of the verbal materials. superior parietal lobules was observed in seven subjects. Both of the mental recall tasks activated the bilateral RESULTS superior temporal areas and left inferior parietal sulcus in all the subjects. The iMR task activated the bilateral Behavioral Data superior parietal lobules in nine subjects, while the pMR Response accuracy for the writing-to-dictation of kanji task yielded activation in the same area in seven sub- (iWD), writing-to-dictation of kana phono- jects. The iMR and pMR tasks activated the left dorso- grams (pWD), mental recall of kanji ideograms (iMR), lateral prefrontal cortex in seven and six subjects, mental recall of kana phonograms (pMR), and semantic respectively. Activation of the left anterior cingulate judgment (SJ) tasks during scanning was 96.5 ± 3.8%, cortex and medial frontal area was found in eight 98.5 ± 2.7%, 95.5 ± 4.2%, 97.3 ± 4.2%, and 95.8 ± 4.3%, subjects in iMR, while activation of these areas was respectively (mean ± SD). Accuracy of all the subjects observed in six subjects in pMR. exceeded 87% in each of the tasks. They also reported For the lPITC, the number of activated voxels and that they could repeat most of the auditory word stimuli peak Z scores of the individual-based analysis are re- successfully in the repetition (REP) task. ported in Table 2. Activation of this area was observed in all the subjects in the iWD, iMR, and pMR tasks. The fMRI Results pWD task also yielded smaller clusters of activated voxels in the same region in seven subjects. In Figure 2, Writing and Mental Recall Tasks statistical parametric maps for each task are superim- Activated brain areas in group analysis are presented in posed across the nine subjects on the coronal of Table 1 for each task (Z > 3.09, corrected at p < .05). the brain template. The activation foci converged in the Figure 1 illustrates the activation sites projected onto left posterior inferior temporal gyrus in the iWD, iMR,

Nakamura et al. 105

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02October 2021 0 ora fCgiieNuocec oue1,Nme 1 Number 14, Volume Neuroscience Cognitive of Journal 106

Table 1. Brain Regions (Brodmann’s Area in parenthesis) Activated by the Writing-to-Dictation and Mental Recall Tasks (Group Analysis)

Kanji Writing (iWD) Kana Writing (pWD) Kanji Mental Recall (iMR) Kana Mental Recall ( pMR)

Brain Area Coordinate Z Score Coordinate Z Score Coordinate Z Score Coordinate Z Score

L dorsolateral prefrontal cortex (46) À47 +35 +14 3.90 À40 +30 +28 4.12 À41 +28 +19 3.90 L inferior frontal gyrus (44, 45) À50 +4 +26 5.45 À48 +1 +26 5.54 À40 +3 +32 5.07 À50 +4 +26 4.13 R inferior frontal gyrus +48 +12 +21 4.96 +48 +10 +20 4.46 L cingulate gyrus (32) À12 +18 +41 4.92 À1 +3 +44 5.17 À3 +3 +42 4.09 R cingulate gyrus +2 +6 +44 6.37 +3 +3 +46 6.00 +3 +1 +38 +4.68 L medial frontal gyrus (6) À8 +22 +43 +4.30 À6 À14 +54 3.57 À12 +6 +48 3.86 À1 +1 +49 3.22 R medial frontal gyrus +4 À7 +50 5.09 +8 À5 +47 4.07 +4 À5 +51 5.06 +1 À5 +51 4.01 L fronto-parietal junction À22 À25 +53 7.21 À22 À29 +49 4.84 L superior temporal gyrus (21, 22) À56 À23 +4 6.10 À47 À36 +7 5.26 À52 À23 +4 6.10 À52 À23 +2 5.37 R superior temporal gyrus +52 À7 À2 6.46 +50 À11 À3 5.87 +50 À11 À3 6.15 +55 À19 +1 4.72 L inferior temporal gyrus (37) À43 À58 À9 5.09 À45 À58 À9 5.32 À47 À56 À15 4.65 L superior parietal lobule (40) À29 À50 +50 6.35 À34 À42 +50 5.82 R superior parietal lobule +32 À52 +50 6.67 +31 À46 +45 4.69 L inferior parietal sulcus À27 À40 +24 4.49 À40 À36 +40 5.58 À38 À34 +40 5.75 À33 À46 +41 4.73 R inferior parietal sulcus +29 À50 +36 4.20 +27 À54 +41 4.38 L basal ganglia/thalamus À24 À7 +6 6.50 À13 À21 +7 4.91 R cerebellum +8 À69 À20 6.85 +11 À48 À19 4.97 +8 À67 À20 5.31

Each of the task conditions is contrasted with their respective baselines. The Z values are thresholded at 3.09 (corrected at p < .05 for multiple comparison). The location coordinates are presented in mm

according to the stereotaxic atlas of Talairach and Tournoux (1988). Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 May 18 on guest by http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf from Downloaded Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021

Figure 1. Activated brain areas in the writing and mental recall tasks (group analysis). The figures are thresholded at Z > 3.09, corrected at p < .05 for multiple comparison. The iWD task produced a distinct activation focus in the lPITC (arrow) in addition to the extensive activation in the left sensorimotor cortices. In the pWD task, a smaller cluster of activated voxels in the lPITC did not survive the statistical criterion (see Results). The mental recall task activated the lPITC similarly for both scripts (arrow). The last two tasks also activated other cortical areas including the left prefrontal cortices and bilateral inferior parietal areas.

Nakamura et al. 107

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 Table 2. Maximum Z Scores and Number of Activated Voxels in the lPITC for the Writing-to-Dictation and Mental Recall Tasks (Individual Analysis)

Kanji Writing (ID) Kana Writing (PD) Kanji Mental Recall (IR) Kana Mental Recall (PR)

Subject Z Score No. of Voxels Z Score No. of Voxels Z Score No. of Voxels Z Score No. of Voxels

1 5.49 140 3.83 16 5.77 349 3.89 28 2 5.21 83 – – 5.77 244 4.87 113 3 4.51 118 3.73 22 6.22 575 5.18 112 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 4 3.79 23 – – 4.18 24 3.97 26 5 6.29 249 4.89 88 5.07 128 5.18 112 6 4.90 50 3.70 15 – – 4.93 151 7 4.46 25 – – 4.89 70 5.44 158 8 5.24 68 – – 5.79 407 4.92 108 9 6.21 285 5.56 92 6.23 174 5.79 34

and pMR tasks, whereas activated voxels in the lPITC did central gyrus, and superior and middle temporal gyri. not overlap in the pWD task. There were also bilateral activations in the basal ganglia For the brain areas identified in the group analysis, and cerebellar hemispheres. However, only a weak trend Figure 3 compares the average percent signal change of activation was found in the lPITC (Z = 2.89, uncor- among the four tasks. Two-way repeated measures rected). The SJ task activated the bilateral fronto-tem- ANOVA on the percent signal change revealed a signifi- poral areas including the superior and middle temporal cantly greater activation in writing relative to mental gyri and inferior frontal gyrus. Bilateral activations recall in the left fronto-parietal areas and right cerebel- were found in the cingulate cortices. There was a lum [F(1,8) = 9.78, p < .05; F = 8.24, p < .05, weak tendency of activation in the lPITC (Z = 3.05, respectively]. The main effect of task was marginally uncorrected). significant in the left medial frontal area (F =4.56, In individual analysis, the REP task yielded a cluster of p = .07). However, no brain region showed enhanced activated voxels in the left inferior temporal gyrus in responses in mental recall relative to writing. In contrast, three subjects. In the SJ task, activated voxels were there were regions that showed increased responses in detected in the lPITC in five subjects. Figure 4B illus- kanji relative to kana: the left dorsolateral prefrontal trates the lPITC response averaged among the nine cortex (F = 6.60, p = .03), the left inferior frontal gyrus (F = 6.09, p = .04), and the left anterior cingulate gyrus (F = 26.48, p < .001). However, no brain region showed a greater response in kana relative to kanji. In addition, none of these regions showed a significant interaction between the task and script type. ThesignalresponseinthelPITCincreasedmost greatly in the iWD task, whereas the response nearly equaled in the iMR and pMR tasks (see Figure 3). The pWD task exhibited a weaker response in this area. Two- way repeated measures ANOVA on the percent signal changes revealed that neither task nor script type reached statistical significance (F = 0.27, p = .62; F = 3.00, p = .12, respectively). However, there was a significant interaction between the two factors (F = 8.22, p = .02).

Comparison With the Repetition and Semantic Tasks For the other two tasks, the group results are displayed Figure 2. Statistical parametric maps of the nine subjects super- imposed on the coronal plane ( y = À58 mm). A cluster of activated in Figure 4A (corrected at p < .05). The REP task yielded voxels converged in the lPITC across more than three subjects in the significant bilateral activations in the fronto-temporal iWD, iMR, and pMR tasks (arrows), whereas the individual-level junctions including the ventrolateral frontal cortex, pre- activations did not overlap in the area in the pWD task.

108 Journal of Cognitive Neuroscience Volume 14, Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021

Nakamura et al. 109

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 subjects for the two tasks. These tasks weakly enhanced angular gyrus with the frontal perisylvian cortices in the the response of this area in contrast to the large left hemisphere (Iwata, 1984). However, our finding that response in the other four tasks (see Figure 3 for the lPITC showed greater responses in pWD than in two comparison). One-way ANOVA on the percent signal control tasks at group level suggests that the area also increase revealed a significant main effect of task participates in normal writing of kana to a certain [F(5,48) = 4.16, p = .003]. Furthermore, post hoc degree. Suppose the area works for storage of visual ANOVA disclosed that the lPITC responded more graphic memory, the inconsistency with lesion data may strongly to iWD than REP and SJ ( p < .001 and < be reconciled by that subjects simultaneously used the .001, respectively). The area also showed greater re- two (or more) different strategies for writing as men- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 sponses in the mental recall tasks relative to the last tioned before. This explanation implies that the phono- two tasks: iMR versus REP ( p < .03) and pMR versus REP logical conversion system is not the only pathway ( p < .04). The difference in percent signal increase was functioning in actual writing of kana. marginally significant in iMR versus SJ ( p = .05) and The present result that the iMR and pMR tasks equally pMR versus SJ ( p = .06). However, the pWD task activated the lPITC indicates that the local neuronal compared with the last two tasks did not reach signifi- activity necessary for activating the visual memory of cance: pWD versus REP ( p = .49) and pWD versus SJ letter forms does not differ between the two script types. ( p = .67). This may appear rather counterintuitive, because there are reasons to assume that people should have greater difficulty in handling of kanji. First, the kana script system DISCUSSION is comprised of only 46 basic characters and two dia- Retrieval of visual word forms is an effective cognitive critics, whereas more than 3000 kanji characters are strategy for the act of writing that is probably adopted by commonly used in everyday activities. Second, each kana most skilled writers. Selective impairment of this process always represents a single syllable and thus has a strict has been described as lexical agraphia in the western one-to-one correspondence with sound. In contrast, literature, for normal adults may use at least two differ- each kanji usually has several different phonetic values ent strategies for spelling words: lexical and phonolog- and thus the relationship between letter and pronunci- ical (e.g., Roeltgen, 1993). The lexical or orthographic ation is very often one-to-many. Third, each kana char- system utilizes a whole-word retrieval process that may acter has a relatively simple visual form consisting of no incorporate visual word images, whereas the phonolog- more than four strokes, while most kanji characters have ical system converts acoustic word images serially into more complex configurations consisting of several sub- letters (Figure 5). The kanji agraphia in lPITC damage is components. Apparently, access to kanji graphic images thought to represent a Japanese equivalent of lexical should involve greater selection demands because of the agraphia, since normal writing of kanji involves a mental larger graphic memory store and higher phonological operation comparable to the whole-word retrieval. In opacity. Therefore it may seem rather plausible that the contrast, the phonological system is thought to be lPITC should be more active in the iMR task. The current especially important for writing out kana characters result, however, does not support this prediction. (Soma et al., 1989). Rather, it suggests that the area works for storage of The present study demonstrated that the lPITC was visual graphic images for both scripts, which may conflict commonly activated by writing and mental recall of with previous studies suggesting that the visual graphic kanji. However, as expected, we found that this area memory of kana is represented in different cortical areas was also active in mental recall of kana graphic forms. from that of kanji, namely, the left angular gyrus (Sakurai Furthermore, the same area showed enhanced re- et al., 2000; Tokunaga et al., 1999). Several other studies, sponses in actual writing of kana. The location of these however, have suggested that the lPITC subserves visual activations overlaps the lesion site in BA 37 that Soma processing of letter/word forms in both alphabetical and et al. (1989) described as responsible for producing the nonalphabetical languages (Cohen et al., 2000; Naka- selective kanji agraphia. In fact, studies on brain-dam- mura et al., 2000; Nobre, Allison, & McCarthy, 1994; aged patients have suggested that the lPITC is not Goldenberg et al., 1989). In general, studies in humans involved in writing of kana, which remains largely intact and primates have established that the inferior temporal in focal damage to the area. Rather, the skill has been cortex works as a storehouse of visual long-term memory associated with a different neural network linking the (Miyashita, 1993). Given the phylogenetic role of the

Figure 3. Average percent signal changes in the brain areas active in the writing and mental recall tasks. Error bars indicate standard errors. The left dorsolateral prefrontal cortex (DLPFC), inferior frontal gyrus (IFG), and anterior cingulate cortex (ACC) showed greater responses in kanji for both writing and recall. By contrast, the left fronto-parietal junction (FPJ) and right cerebellum (CBL) responded more greatly in writing relative to mental recall. In the lPITC, there was a large signal increase in iWD in contrast to the weaker response in pWD, while the same area responded similarly in iMR and pMR. MdFG = medial frontal gyrus, STG = superior temporal gyrus, SPL = superior parietal lobule, IPS = inferior parietal sulcus, and BG = basal ganglia.

110 Journal of Cognitive Neuroscience Volume 14, Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 ventral visual pathway that may cover a wide range of visual objects, both verbal and nonverbal (Puce, Allison, Asgari, Gore, & McCarthy, 1996; Ungerleider & Haxby, 1994), it seems unlikely that the visual images of kana are stored exceptionally in a separate brain region outside the inferior temporal cortex. In contrast, the kanji versus kana comparison revealed greater responses in the left dorsolateral prefrontal and anterior cingulate cortices, suggesting that the greater Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 selection difficulty in retrieval of kanji resulted in activa- tion of these regions. Recent work in monkeys suggests that ‘‘top-down’’ processes originating from the prefron- tal cortex regulate the retrieval of visual long-term memory stored in the inferotemporal cortex (Hasegawa, Fukushima, Ihara, & Miyashita, 1998). In humans, the right dorsolateral prefrontal cortex has been associated with episodic memory retrieval (Fletcher et al., 1995), while the left homologous area is thought to play a critical role in word retrieval as shown by lesion (Ben- ton, 1968) and neuroimaging studies (Warburton et al., 1996). The anterior cingulate cortex has also been shown to be sensitive to increase in task difficulty (Paus, Koski, Caramanos, & Westbury, 1998). Therefore, en- hanced responses in these structures may be interpreted as compensating for the greater selection demands in retrieving kanji graphic memory. The lPITC showed a task-by-script interaction at both group and individual level, whereas neither of the two factors exhibited a significant effect in this area. These findings indicate that the act of writing modulates the visual word-form retrieval system differently between kanji and kana, and provide the first neuroimaging

Figure 4. (A) Statistical parametric maps for the two additional tasks (group analysis). The figures are thresholded at Z > 3.09 (corrected at p < .05). The REP task activated the bilateral perisylvian and sensorimotor areas. The SJ task activated the bilateral middle and inferior frontal gyri, insula, and superior temporal gyrus. However, neither of the two tasks yielded a significant activation in the lPITC. (B) Average percent signal changes in the lPITC during the REP and SJ tasks (individual analysis). These two tasks also showed a trend of activation in this area. However, the levels of signal increase were significantly lower than those in the other four tasks (see Results). Figure 5. Model of basic language processes operating in writing.

Nakamura et al. 111

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 evidence supporting that neuropsychological processes areas in a different way according to the task demands involved in writing are different between the two scripts. that are inherent in their graphic features. Therefore, A possible account for this interaction is that it may contrary to the traditional view, the present data provide reflect the difference in duration of the motor execution new functional imaging evidence indicating that in nor- occurring in the postretrieval phase, given that the mal people the neural systems for processing the two mental recall tasks, which essentially represent the different scripts are separable only functionally, rather retrieval process itself, recruited the lPITC similarly for than anatomically. both scripts. In fact, actual writing of kanji should be a more time-consuming process in proportion to their METHODS Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 higher visual complexity, in which people must maintain the letter images to be written longer after the successful Subjects retrieval. In contrast, only a transient activation of visual Subjects consisted of 9 healthy volunteers (age range graphic memory may be sufficient to generate kana 21–30 years) recruited from students at Kyoto Univer- characters. Thus, although such phase-related signal sity. All were native Japanese speakers and strongly changes cannot be systematically evaluated with the right-handed as confirmed by the Edinburgh Inventory. blocked-design fMRI, the interaction may be associated Informed consent was obtained from each subject prior with the longer postretrieval phase for converting the to the experiment. The protocol of this study was retrieved graphic images into written forms. Also, it is approved by the Committee of Medical Ethics, Graduate possible that writing strategies other than the whole- School of Medicine, Kyoto University. word retrieval, e.g., the direct sound-to-letter conversion system, predominantly worked for writing of kana. Word Stimuli and Behavioral Tasks Although the current data do not allow us to estimate the relative contribution of the phonological pathway, it Forty auditory word stimuli were used for all the may be addressed by testing, e.g., an interaction be- activation paradigms in the present study. Each of tween the main effects of phonological processing and these words is conventionally written in a compound actual writing for the two scripts. of two kanji characters that are listed among the We previously demonstrated a lateralized activation fundamental vocabulary for Japanese language teach- of the lPITC while subjects manually or mentally trans- ing (National Language Research Institute, 1984). They lated visually presented kana words into kanji script can be spelled with several kana characters, as is the (Nakamura et al., 2000). In the present study, the iWD case with any given word of the Japanese language. and iMR tasks—equivalent behavioral tasks using the None of the word stimuli had words. Half auditory stimuli—replicated the previous finding. Tak- of the words represented concrete and the en together, it is suggested that this area works as a other half abstract nouns. In all the tasks, the single cross-modal association cortex for integrating the per- word stimuli were randomly presented every 3 sec ceptual information for higher-order verbal processing, during each task epoch (see below). Each of the regardless of input modalities of the stimulus materials. active tasks was contrasted with a common baseline This is in good accordance with a PET study that this condition in which subjects were instructed to remain area was active when congenitally blind subjects were relaxed and empty the mind. engaged in a reading task (Bu¨chel, Price, & The set of four activation tasks consisted of iWD, Friston, 1998). In addition, the finding that the writing pWD, iMR, and pMR. A 2 Â 2 factorial design was and mental recall tasks consistently showed a greater adopted to evaluate the effects of task (writing vs. activation in the lPITC throughout the comparisons mental recall) and script type (kanji vs. kana) and their with the two control tasks suggests that the area is interaction (see below for detail). In the iWD task, especially important for the expressive, rather than subjects held a pen on a smooth plastic board and receptive, language operations. This is consistent with roughly wrote the first character of a compound kanji the description that damage to this area most com- word that corresponds to each stimulus word (Figure 6). monly produces a difficulty in word finding in the The interstimulus interval of 3 sec allowed the expected purest type (Goodglass, 1993). writing performance, since normal adults require ap- To summarize, the present experiment, focusing on proximately 1.8 sec to write out a single kanji to dicta- the role of the lPITC in visual word retrieval, differ- tion (Kaiho & Nomura, 1983). The subjects could see entiated the neural response of several brain regions neither the movements of their hand nor the written involved in normal writing between kanji and kana. At output. Responses from each subject were monitored by the same time, however, we found that the two script video camera and scored off-line. In the pWD task, systems used largely overlapping cortical regions for subjects wrote out the first character of kana strings both actual writing and mental recall of graphic images. corresponding to each word stimulus (Figure 6). Other- Taken together, these findings can be explained by that wise the task conditions were designed the same as in normal writing of kanji and kana recruits similar brain the pWD task.

112 Journal of Cognitive Neuroscience Volume 14, Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 brain response in verbal operations that do not in- volve the retrieval of visual graphic forms. In the REP task, subjects orally repeated the stimulus words with- out moving the jaw. In addition to the tight constraint on the jaw movement to minimize head motion (see below), they were reminded not to open the mouth wide. These restrictions still allowed minimally audible vocalization required for executing the task. In the SJ task, subjects determined whether the stimulus nouns Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021

Figure 6. Stimulus materials and behavioral paradigms. Forty Japanese nouns were used commonly for all the six activation tasks. The stimulus words could be written in both kanji and kana scripts. Upon auditory presentation of the same set of word stimuli, subjects performed different behavioral tasks according to prespecified instructions. In iWD, subjects wrote down the first character of each two-character kanji word. In iMR, subjects mentally visualized a graphic form of the kanji and judged whether it belonged to the L–R type or not (see Figure 7). Similarly, in pWD and pMR, subjects wrote down and imagined the first character of each kana string, respectively (see Figure 7 for pMR). In REP, subjects orally repeated the auditory stimuli, whereas in SJ they determined whether the stimulus words repre- sented concrete objects or abstract concepts.

The iMR task involved a visual feature decision on kanji graphic images. Among several elemental fea- tures that may determine visual structures of kanji (Saito, 1997; Kaiho & Nomura, 1983), we adopted a ‘‘left–right (L–R) combination of letter-constituents’’ for the current experiment. Namely, some characters are composed of two radicals that are juxtaposed horizontally, while in others letter-constituents are not arranged according to this rule (Figure 7A). In the iMR task, subjects were instructed to mentally transcribe each auditory word stimulus to kanji script, that is, to imagine a visual form of a two-character kanji word corresponding to the stimulus (Figure 6). They responded by squeezing a response bulb when the first character of each kanji word was of the L–R type and withheld the response otherwise. The L–R types appeared in half of the 40 trials. Likewise, each kana character consists of either a single component (SC) or several discrete clusters (Figure 7B). In the Figure 7. Examples of kanji (A) and kana (B) for the mental recall pMR task, subjects responded only when the first tasks. (A) iMR. The graphic forms of kanji can be classified in terms of a character of a kana string corresponding to each basic visual feature; some characters consist of two constituents that stimulus was of the SC type (Figure 6). The SC types are juxtaposed horizontally (left–right combination or L–R type), while others have one or more visual clusters that are arranged differently were presented in half of the 40 trials. (non-L–R type). (B) pMR. Each kana character is comprised of either a Two additional tasks, REP and SJ, were also per- single visual component (SC type) or several discrete clusters (non-SC formed using the same word stimuli to examine the type).

Nakamura et al. 113

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 denote concrete objects or abstract concepts, and series with a boxcar reference waveform, images of the responded only to the abstract nouns by squeezing individual level activation were computed as an ad- the bulb. The abstract nouns appeared in half of the justed mean image per condition per session for each 40 trials. subject. For each task, these condition-specific mean images were subject to a group-based statistics using a paired t test. A significant level threshold of Z > 3.09 fMRI Procedure (corresponding to p < .001 at each voxel level) was Subjects wore headphones connected with paired used to determine the presence of significant activa- plastic tubes and lay supine with eyes closed in the tion foci in each task. The spatial extent of clusters Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 MR scanner. They held with their right hand a pen for was corrected at p < .05 for multiple comparison. the writing-to-dictation tasks and a response bulb for Secondly, the fMRI data were analyzed on a single the iMR, pMR, and SJ tasks. The stimulus sounds were subject basis to examine the variability and consistency generated outside the magnetic shield and delivered of the activation patterns across the subject group. A binaurally to the subjects via air conduction at approx- boxcar reference function was convolved with the imately 100 dB sound pressure level. Background Gaussian-shaped hemodynamic response function, noise was reduced to 70 dB under the tightly and then correlations with the fMRI time-series data occlusive headphones. Head motion was minimized were computed for all voxels and transformed into a Z using foam padding. Scanning was conducted with a score map (Friston, Jezzard, & Turner, 1994). The 1.5 T whole-body MRI system (Vision plus; Siemens, significant Z value was thresholded at Z >3.09 Erlangen, Germany) using a standard head coil opti- (uncorrected for multiple comparison). Activated brain mized for whole-brain echo-planar imaging (EPI). structures were identified by transforming the MNI Functional imaging used a gradient echo EPI sequence coordinate into the standard brain atlas of Talairach with the following parameters: TR 5 sec, TE 35 msec, and Tournoux (1988). flip angle 908, field-of-view 20 Â 20 cm, and 64 Â 64 For between-task statistical comparisons, regions of pixel matrix. Thirty contiguous 4 mm thick slices interest (ROIs) were set up for cortical areas identified without gap were obtained in the axial plane for each in the group analysis described above. Average signal subject. For each task, one scanning session lasting increase relative to the baseline was calculated by 270 sec and yielding 54 functional images was per- sampling a peak voxel within the ROIs for each task formed for each subject. In each session, 9 task for each subject. For the set of writing and mental epochs, namely, 5 for the baseline and 4 for a single recall tasks, the resulting percent signal changes were activation task, alternated every 30 sec. (Thus no two subject to a two-factor repeated measures ANOVA to of the activation tasks were performed within a single examine the main effects of task (writing vs. mental session.) The order of task presentation was counter- recall) and script type (kanji vs. kana) and their balanced across the subjects. interaction. Additionally, planned comparisons with the two control tasks were performed using a post hoc ANOVA (Fisher’s PLSD) on the percent signal Data Analysis changes described above. After image reconstruction, off-line processing of the functional images was performed on a Sparc Ultra-2 workstation (Sun Microsystems, Mountain View, CA) Acknowledgments using the SPM96 software (Wellcome Department of This work was supported by Grants-in-Aid for Scientific Cognitive Neurology, London, UK). Two initial im- Research (A) 09308031, for Scientific Research on Priority ages were discarded to eliminate nonequilibrium Areas 08279106, and for International Scientific Research 10044269 from the Japan Ministry of Education, Science, effects of magnetization. Images were corrected for Sports, and Culture, Research for the Future Program JSPS- head motion, resampled every 2 mm using bilinear RFTF97L00201 from the Japan Society for the Promotion of interpolation, normalized to the standard brain space Science, and General Research Grants for Aging and Health (Friston et al., 1995), and spatially smoothed with an ‘‘Physiological Parameters for evaluation of aging of brain’’ and isotropic Gaussian filter (6 mm full width at half ‘‘Analysis of aged brain function with neuroimaging techni- ques’’ from the Japan Ministry of Health of Health and Welfare. maximum). A high pass filter (0.5 cycles/min) and The authors are grateful to Prof. Shintaro Funahashi for helpful temporal smoothing were applied to remove low comments on the manuscript. frequency noise and enhance the signal-to-noise Reprint requests should be sent to Hiroshi Shibasaki, M.D., ratio, respectively. Department of Brain Pathophysiology, Kyoto University The fMRI data were analyzed at both the group and Graduate School of Medicine, Kyoto 606-8507, Japan, or via individual levels. Firstly, overall activation patterns for e-mail: [email protected].ac.jp. each task relative to the baseline were determined on The data reported in this experiment have been deposited in the basis of a group-based statistical inference (Friston, the fMRI Data Center (http://www.fmridc.org). The accession Holmes, & Worsley, 1999). By fitting the fMRI time- number is 2-2001-1125Q.

114 Journal of Cognitive Neuroscience Volume 14, Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021 REFERENCES Fukuyama, H., Konishi, J., & Shibasaki, H. (2000). Partici- pation of the left posterior inferior temporal cortex in Benton, A. L. (1968). Differential behavioral effects in frontal writing and mental recall of kanji . Brain, 123, lobe disease. Neuropsychologia, 6, 53–60. 954–967. Bu¨chel, C., Price, C., & Friston, K. (1998). A multimodal lan- National Language Research Institute. (1984). A study of the guage region in the ventral visual pathway. Nature, 394, fundamental vocabulary: General description and voca- 274–277. bulary tables [in Japanese]. : Shuei-Shuppan. Cohen, L., Dehaene, S., Naccache, L., Lehe´ricy, S., Dehaene- Nobre, A. C., Allison, T., & McCarthy, G. (1994). Word recog- Lambertz, G., ´naff, M. A., & Michel, F. (2000). The visual nition in the human inferior temporal lobe. Nature, 372, word form area: Spatial and temporal characterization of an 260–263. initial stage of reading in normal subjects and posterior split-

Paus, T., Koski, L., Caramanos, Z., & Westbury, C. (1998). Re- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/14/1/104/1757408/089892902317205366.pdf by guest on 18 May 2021 brain subjects. Brain, 123, 291–307. gional differences in the effects of task difficulty and motor Fletcher, P. C., Frith, C. D., Grasby, P. M., Shallice, T., Fracko- output on blood flow response in the human anterior wiak, R. S. J., & Dolan, R. J. (1995). Brain systems for en- cingulate cortex: A review of 107 PET activation studies. coding and retrieval of auditory–verbal memory: An in vivo NeuroReport, 22, R37–R47. study in humans. Brain, 118, 401–416. Puce, A., Allison, T., Asgari, M., Gore, J. C., & McCarthy, Friston, K. J., Ashburner, J., Poline, J. B., Frith, C. D., Heather, G. (1996). Differential sensitivity of human visual cortex J. D., & Frackowiak, R. S. J. (1995). Spatial registration and to faces, letterstrings, and textures: A functional magnetic normalization of images. Human Brain Mapping, 2, 165–189. resonance imaging study. Journal of Neuroscience, 16, Friston, K. J., Holmes, A. P., & Worsley, K. J. (1999). How many 5205–5215. subjects constitute a study? Neuroimage, 10, 1–5. Roeltgen, D. P. (1993) Agraphia. In K. M. Heilman & E. Va- Friston, K. J., Jezzard, P., & Turner, R. (1994). Analysis of lenstine (Eds.), Clinical neuropsychology (pp. 63–89). New functional MRI time-series. Human Brain Mapping, 1, York: Oxford University Press. 153–171. Saito, H. (1997). Mental lexicon [in Japanese]. In Y. Ohtsu & T. Goldenberg, G., Podreka, ., Steiner, M., Willmes, K., Suess, E., Gunji (Eds.), Gengo-no-kagaku [Linguistic sciences] (pp. & Deecke, L. (1989). Regional cerebral blood flow patterns 93–153). Tokyo: Iwanami Shoten. in visual imagery. Neuropsychologia, 27, 641–664. Sakurai, Y., Matsumura, K., Iwatsubo, T., & Momose, T. Goodglass, H. (1993). Understanding aphasia. San Diego, (1997). Frontal pure agraphia for kanji or kana: Dissocia- CA: Academic Press. tion between morphology and phonology. Neurology, 49, Hamasaki, T., Yasojima, K., Kakita, K., Masaki, H., Ishino, 946–952. S., Murakami, M., Yamaki, T., & Ueda, S. (1995). Alexie- Sakurai, Y., Momose, T., Iwata, M., Sudo, Y., Ohtomo, K., & agraphie pour l’e´criture kanji apre`s le´sion temporale Kanazawa, I. (2000). Different cortical activity in reading of poste´ro-infe´rieur gauche. Revue Neurologique, 151, Kanji words, Kana words and Kana nonwords. Cognitive 16–23. Brain Research, 9, 111–115. Hasegawa, I., Fukushima, T., Ihara, T., & Miyashita, Y. (1998). Sakurai, Y., Sakai, K., Sakuta, M., & Iwata, M. (1994). Naming Callosal window between prefrontal cortices: Cognitive difficulties in alexia with agraphia for kanji after a left pos- interaction to retrieve long-term memory. Science, 281, terior inferior temporal lesion. Journal of Neurology, Neu- 814–818. rosurgery and Psychiatry, 57, 609–613. Iwata, M. (1984). Kanji versus kana: Neuropsychological cor- Soma, Y., Sugishita, M., Kitamura, K., Maruyama, S., & Imanaga, relates of the Japanese . Trends in Neu- H. (1989). Lexical agraphia in the Japanese language: Pure roscience, 7, 290–293. agraphia for kanji due to left posteroinferior temporal le- Kaiho, H., & Nomura, Y. (1983). Kanji jouhou shori no shin- sions. Brain, 112, 1549–1561. rigaku [Psychology of information processing of kanji]. Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic Tokyo: Kyoiku-Shuppan. atlas of the human brain: 3D proportional system: Kawahata, N., Nagata, K., & Shishido, F. (1988). Alexia An approach to cerebral imaging. New York: Thieme. with agraphia due to the left posterior inferior tem- Tanaka, Y., Yamadori, A., & Murata, S. (1987). Selective kana poral lobe lesion—neuropsychological analysis and its agraphia: A case report. Cortex, 23, 679–684. pathogenetic mechanisms. Brain and Language, 33, Tokunaga, H., Nishikawa, T., Ikejiri, Y., Nakagawa, Y., Ya- 296–310. suno, F., Hashikawa, K., Nishimura, T., Sugita, Y., & Kawamura, M., Hirayama, K., Hasegawa, K., Takahashi, N., & Takeda, M. (1999). Different neural substrates for Kanji Yamamura, A. (1987). Alexia with agraphia of kanji ( Japa- and Kana writing: A PET study. NeuroReport, 10, nese morphograms). Journal of Neurology, Neurosurgery 3315–3319. and Psychiatry, 50, 1125–1129. Ungerleider, L. G., & Haxby, J. V. (1994). ‘‘What’’ and ‘‘where’’ Mayeux, R., & Kandel, E. R. (1991). Disorders of language: The in the human brain. Current Opinion in Neurobiology, 4, aphasias. In E. R. Kandel, J. H. Schwartz, & T. M. Jessell 157–165. (Eds.), Principles of neural science (pp. 849–850). New Warburton, E., Wise, R. J. S., Price, C. J., Weiller, C., Hadar, U., York: McGraw-Hill. Ramsay, S., & Frackowiak, R. S. J. (1996). and Miyashita, Y. (1993). Inferior temporal cortex: Where visual retrieval by normal subjects: Studies with PET. Brain, 119, perception meets memory. Annual Review of Neuroscience, 159–179. 16, 245–263. Yokota, T., Ishiai, S., Furukawa, T., & Tsukagoshi, H. (1990). Mochizuki, H., & Ohtomo, R. (1988). Pure alexia in Japanese Pure agraphia of kanji due to thrombosis of the Labbe´ vein. and agraphia without alexia in kanji: The ability dissociation Journal of Neurology, Neurosurgery and Psychiatry, 53, between reading and writing in kanji vs kana. Archives of 335–338. Neurology, 45, 1157–1159. Nakamura, K., Honda, M., Okada, T., Hanakawa, T., Toma, K.,

Nakamura et al. 115

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892902317205366 by guest on 02 October 2021