- Email: [email protected]
Changes in brain activation patterns associated with learning of Korean words by Japanese: an fMRI study
NeuroImage 20 (2003) 1–11
www.elsevier.com/locate/ynimg
Changes in brain activation patterns associated with learning of Korean words by Japanese: an fMRI study Hyung-Suk Lee,a,* Toshikatsu Fujii,a Jiro Okuda,a Takashi Tsukiura,b,c Atsushi Umetsu,d Maki Suzuki,a Tatsuo Nagasaka,e Shoki Takahashi,d and Atsushi Yamadoria a
Division of Neuropsychology, Department of Disability Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan b The Japan Society for the Promotion of Science, Tokyo, Japan c Neuroscience Research Institute, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan d Department of Diagnostic Radiology, Tohoku University Graduate School of Medicine, Sendai, Japan e Division of Radiology, Tohoku University Hospital, Sendai, Japan Received 15 July 2002; revised 3 April 2003; accepted 24 April 2003
Abstract We used functional magnetic resonance imaging (fMRI) to explore the change in brain activation associated with the learning of Korean words written in Han-gul characters (K-words) by young Japanese at two stages. Subjects were 12 right-handed native Japanese without previous knowledge of Korean words and characters. On the first day they were taught the pronunciation and meaning of 20 K-words. Then, after the first fMRI session (on day 2), they were given a set of 20 cards with the words and corresponding photographs. They also received a tape and were instructed to memorize the 20 K-words by studying them every day until the day of the second fMRI session (day 16). During the fMRI sessions, 20 Japanese words written in kana syllabograms (J-words) and the 20 previously presented K-words, as well as 20 new K-words (Kn-words) were presented visually for silent reading. The first J-word reading, relative to the first K-word reading, showed activation in the left angular gyrus. K-word reading relative to J-word reading during both sessions showed activation in occipital regions. Within these activated areas, session by condition interaction was found only in the left angular gyrus. The interaction between session and condition resulted from the fact that the differences in blood oxygenation-level-dependent signals between K-words and J-words and between Kn-words and J-words were significantly greater in the first session than in the second session. From the results, we concluded that patterns of brain activation changed as the memory of the 20 K-words became fixed through daily practice and that reading of both Korean words and Japanese syllabograms engaged the left angular gyrus. © 2003 Elsevier Inc. All rights reserved. Keywords: Second language; Learning; Angular gyrus; Functional magnetic resonance imaging (fMRI)
Introduction Although there have been several studies of the differences and similarities between the first language as a mother tongue and a second language acquired later as a foreign language (Klein et al., 1994, 1995; Perani et al., 1998; Illes et al., 1999; Chee et al., 1999a, 1999b, 2000), no study has been conducted with regard to the changes in brain activa* Corresponding author. Division of Neuropsychology, Department of Disability Medicine, Tohoku University Graduate School of Medicine, 2-1 Seiryo-machi, Aoba-ku, Sendai 980-8575, Japan. Fax: ⫹81-22-717-7360. E-mail address: [email protected] (H.-S. Lee). 1053-8119/03/$ – see front matter © 2003 Elsevier Inc. All rights reserved. doi:10.1016/S1053-8119(03)00254-4
tion patterns associated with the learning of foreign words. We wondered how brain activation patterns would be affected when subjects intensively learned a limited number of foreign words that had previously been totally unknown to them in a short period of time. We anticipated that if the number of words the subjects were required to master was sufficiently small, they would be able to learn them well enough to enable us to analyze the relationship between the activated brain patterns during the processing of unfamiliar foreign words, familiar foreign words, and native words. The Korean and Japanese languages both employ a phonological script system in addition to Chinese ideographic characters to transcribe native pronunciation. Korean in-
2
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Fig. 1. Overall experimental design.
volves phonological symbols called Han-gul, a symbol representing a phoneme—similar to languages that use an alphabet. However, unlike the alphabetic system, these phonemic symbols are not arranged in a serial order, but are combined into a single form to represent a syllable. These syllabic units are spatially separated from each other. Thus, in a sense, Han-gul characters can also be regarded as syllabograms. Conversely, Japanese employs purely syllabic characters called kana. Unlike a Han-gul syllabogram, a kana syllabogram is an independent sign in itself and does not consist of phonemic units. In this study we investigated the change in brain activation pattern in 12 right-handed native Japanese without previous knowledge of Han-gul words who studied 20 words every day for 16 days with the goal of memorizing them.
card aloud; on the comprehension test, they were asked to verbally report the meaning of the words. The number of correct responses and the reaction times were measured for both tests (Fig. 1). Pre-fMRI study phase (day 1) Stimuli consisted of 20 three-syllable Korean words written in Han-gul characters (K-words), all of which were the names of familiar animals or common objects such as dolphin and bicycle. Each word was printed on a card (105 ⫻ 148 mm in size). On the reverse side of the card there was a photograph representing the meaning of the word (Fig. 2). One of the authors (HSL), a native Korean speaker, taught the subjects the pronunciation and meaning of these 20 K-words repeatedly, until each of them could remember the pronunciation and meaning of at least 18 of the 20 words (90% or more).
Methods Daily learning phase (from days 2 to 15) Subjects Twelve Japanese volunteers (ages 19 –23 years, 6 males and 6 females), who had no previous knowledge of the Korean language, participated in this study. All subjects were right-handed and scored above ⫹85 on the Edinburgh Handedness Inventory (Oldfield, 1971). They gave their written informed consent in accordance with the Declaration of Helsinki. Overall experimental design The experiment consisted of a pre-fMRI study phase (day 1), the first fMRI scanning (day 2), a consecutive learning phase (from day 2 to day 15), and the second fMRI scanning (day 16), which employed the same procedure as the first. In order to monitor how well subjects had learned the 20 K-words, they were given two performance tests on days 1, 2, 15, and 16 using 20 learning cards. On the reading aloud test, the subjects were asked to read each word on a
The second performance test for the 20 words was carried out after the first fMRI measurement on day 2. In addition, each of the subjects was given a set of the cards used in the learning session on the first day, and a magnetic tape on which the pronunciation of these 20 words had been recorded by the same Korean speaker. Subjects were required to study the pronunciation and meaning of the words at home, with the help of the cards and tape, for at least 10 min a day until day 15. They were also asked to record the time they spent on this exercise every day. Tasks during fMRI scanning The subjects performed the same task in two fMRI sessions on days 2 and 16. For this task, we prepared eight lists (list A, B, C, D, E, F, G, and H), each consisting of 10 stimuli. Lists A and E consisted of 10 meaningless figures (Vanderplas and Garvin, 1959). Lists B and F consisted of 10 familiar (studied) Korean words (K-words). Lists C and
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
3
Table 1 Results of reading aloud tests and of reading comprehension tests Day
(1) Reading aloud test Reaction time (s) The number of correct responses (2) Reading comprehension test Reaction time (s) The number of correct responses
1st
2nd
6.1 ⫾ 2.9 19.7 ⫾ 0.5
6.1 ⫾ 1.1 19.8 ⫾ 0.5
2.5 ⫾ 0.5 20 ⫾ 0
2.2 ⫾ 0.5 20 ⫾ 0
5.2 ⫾ 2.23 19.3 ⫾ 0.9
4.5 ⫾ 0.1 18.9 ⫾ 1.2
2.2 ⫾ 0.6 19.8 ⫾ 0.4
1.9 ⫾ 0.4 19.8 ⫾ 0.5
G consisted of 10 new (nonstudied) Korean words (Knwords), which had the same number of Han-gul syllabograms as the studied words. Lists D and H consisted of 10 Japanese words written in kana characters (J-words), which had the same meaning and same number of syllables as the K-words. During fMRI scanning, the lists were presented in order from A to H. Stimuli were projected through a projector and back-projected onto a screen placed just beyond the subject’s feet on the scanner table. Subjects saw the screen through a mirror fixed on a head cage. The presentation time of a stimulus was set at 4 s and the interstimulus interval was 1 s. The subjects were asked to read a projected stimulus silently if they were able to read it and simply to look at it if it was a word they could not read or a meaningless figure. Scanning methods All measurements were carried out using a 1.5-T General Electric Signa scanner. Subjects were positioned in the scanner with their head immobilized in support cushions. A gradient echo planar imaging (EPI) modified by an EPIBOLD pulse sequence was used for functional imaging with the following parameters: TR ⫽ 5000 ms, TE ⫽ 60 ms, field of view ⫽ 24 ⫻ 24 cm2, flip angle ⫽ 90°, matrix size ⫽ 64 ⫻ 64 pixels, slice thickness ⫽ 4 mm, interslice thickness ⫽ 1 mm. For each run, 84 scans (including the first 4 dummy scans) were carried out and 27 axial slices per scan were obtained.
15th
16th
Individual analysis (Level 1) A general linear model was applied to the functional data (Friston et al., 1995). Activation patterns for each task relative to the baseline were determined for each subject. Random effects analysis (Level 2) The weighted sum of the parameter estimates in the individual analysis constituted “contrast” images, which were used for the group analysis (Friston et al., 1999). The contrast images obtained by individual analysis represent each task relative to the baseline. Contrasts of the condition effects of each voxel were assessed using two-sample t test, resulting in a standard image. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, K ⬎ 10 voxels). In a region-of-interests (ROIs) analysis, we performed two-way repeated-measures ANOVAs, with session (first fMRI and second fMRI) and condition (K-words, Knwords, and J-words) as factors on each region.
Results Learning time spent by the subjects to memorize the K-words All the subjects reported that they spent at least 10 min every day mastering the 20 K-words. The mean of the time spent for this task as reported by them was 14.2 min a day.
Data analysis
Results of behavioral tests
Analysis of the fMRI data obtained involved three steps using Statistical Parametric Mapping 99 (SPM99, Wellcome Department of Cognitive Neurology, London, UK) implemented in Matlab 5 (Mathworks, Sherborn, MA, USA). First, 80 volumes in each session acquired from a subject were realigned to their first volume. Second, the realigned images from each subject were transformed and normalized into standard space (Talairach and Tournoux, 1988) using an EPI template and smoothed using a Gaussian kernel of FWHM ⫽ 10 mm. Third, statistical analysis was conducted at the following two levels.
We analyzed the results from two kinds of performance test obtained at four points of time for eight subjects using one-way ANOVA; we were unable to test the other four subjects on day 16. Table 1 shows the results. For the reading aloud test, there was a significant difference in the time required to read a word between day 1 and day 15 [F(1, 7) ⫽ 12.15, P ⬍ 0.025], day 1 and day 16 [F(1, 7) ⫽ 12.5, P ⬍ 0.01], day 2 and day 15 [F(1, 7) ⫽ 107.05, P ⬍ 0.005], and day 2 and day 16 [F(1, 7) ⫽ 100.4, P ⬍ 0.005]. There was no difference in the number of correct responses. For the reading comprehension test, there was
4
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Fig. 2. Example of a card used for memorizing a Korean word in the present study. On the surface of the card, three Han-gul syllables, ol-Pfe-Mi, are printed. The syllabogram ol, the character on the left, is composed of two phonemic signs, i.e., o and l. In fact, in the character ol, three phonemic signs (no sound sign, o, and l) are combined vertically. On the reverse side there is a picture of the meaning of ol-Pfe-Mi.
again a significant difference in the time required to produce a Japanese word in response to a stimulus between day 1 and day 15 [F(1, 7) ⫽ 18.06, P ⬍ 0.005], day 1 and day 16 [F(1, 7) ⫽ 17.69, P ⬍ 0.005], day 2 and day 15 [F(1, 7) ⫽ 90.83, P ⬍ 0.005], and day 2 and day 16 [F(1, 7) ⫽ 55.87, P ⬍ 0.005]. There was no difference in the number of correct responses.
right superior occipital gyrus (Table 2). Next, we compared reading of the J-words and K-words with reading of previously unpresented Korean words (Kn-words). These two comparisons (J1-Kn1 and K1-Kn1) failed to produce any significantly activated areas (Table 2).
Brain activation
To identify areas of brain activation during the late stage of learning of the K-words, we analyzed the data using the same method as that employed for the first analysis. Comparison of J-words with K-words (J2-K2) showed no activated areas (Table 2). A significantly activated area for the second reading of the K-words, compared to that of the J-words (K2-J2), was found in the right superior occipital gyrus (Table 2 and Fig. 5). Silent reading of the J-words and the K-words compared with the Kn-words (J2-Kn2 and K2-Kn2) failed to produce significant activation (Table 2).
Brain activation during the first fMRI (on day 2) To identify areas of brain activation during the early stage of K-word learning, we compared silent reading of J-words with silent reading of K-words at early exposure (J1-K1) and vice versa (K1-J1). Significantly activated areas during silent J-word reading, compared with silent K-word reading, were found in the left angular gyrus (Table 2 and Fig. 3). The converse contrast (K1-J1) showed activation in the left inferior occipital gyrus (Table 2 and Fig. 4) and the
Brain activation during the second fMRI (on day 16)
Fig. 3. The activation in the left angular gyrus revealed by contrasts between J-word reading and K-word reading in the early stage (J1-K1). The top left image shows the axial image (z ⫽ 22) and the top right one shows the rendering image on the lateral surface of the left hemisphere. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, k ⬎ 10 voxels). Bar graphs show percentage signal changes in each condition during the first and second fMRI sessions. The upper graph shows the signal changes in the three conditions (K-words, Kn-words, and J-words) during the two fMRI sessions. The lower left graph shows the signal changes in the two conditions (K-words and J-words) during the two fMRI sessions. The lower right graph shows the signal changes in the two conditions (Kn-words and J-words) during the two fMRI sessions. K1 ⫽ the reading of Korean words during the first fMRI, K2 ⫽ the reading of Korean words during the second fMRI, J1 ⫽ the reading of Japanese words during the first fMRI, J2 ⫽ the reading of Japanese words during the second fMRI, Kn1 ⫽ the presenting of new (nonstudied) Korean words during the first fMRI, Kn2 ⫽ the presenting of new (nonstudied) Korean words during the second fMRI.
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
5
6
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Table 2 Data for anatomical structures, the Talairach coordinates, and the z values of peak activation for K1-J1, J1-K1, K1-Kn1, J1-Kn1, K2-J2, J2-K2, K2-Kn2, J2-Kn2 (corrected, P ⬍ .05 K ⬎ 10 voxels) Regions (Brodmann’s area) J1-K1 L angular gyrus (39) K1-J1 L inferior occipital gyrus (18) R superior occipital gyrus (19) J1-Kn1 No activation K1-Kn1 No activation J2-K2 No activation K2-J2 R superior occipital gyrus (19) J2-Kn2 No activation K2-Kn2 No activation
x
y
z
z value
Voxels
⫺56
⫺66
22 5.15
31
⫺32 28
⫺94 ⫺76
⫺6 5.58 40 4.93
217 15
30
⫺78
40 5.44
46
121.052, P ⬍ 0.0001], and an interaction between two factors [F(1, 11) ⫽ 67.144, P ⬍ 0.0001]. The ANOVA with factors of session (first fMRI and second fMRI) and condition (Kn-words and J-words) showed a significant main effect of condition [F(1, 11) ⫽ 55.239, P ⬍ 0.0001], and an interaction between two factors [F(1, 11) ⫽ 14.095, P ⬍ 0.003], whereas no main effect of session was found (Fig. 3). These interactions between session and condition resulted from the fact that the differences in BOLD signals between K-words and J-words and between Kn-words and J-words were significantly greater in the first session than in the second session. Left inferior occipital gyrus (BA18; x, y, z ⫽ ⫺32, ⫺94, ⫺6) The ANOVA showed a significant main effect of condition [F(2, 22) ⫽ 123.599, P ⬍ 0.0001], but no main effect of session or any interaction between two factors (Fig. 4).
Note. K1, the reading of Korean words during the first fMRI; K2, the reading of Korean words during the second fMRI; J1, the reading of Japanese words during the first fMRI; J2, the reading of Japanese words during the second fMRI; Kn1, the presenting of new (nonstudied) Korean words during the first fMRI; Kn2, the presenting of new (nonstudied) Korean words during the second fMRI.
Right superior occipital gyrus (BA19; x, y, z ⫽ 30, ⫺78, 40) The ANOVA showed a significant main effect of condition [F(2, 22) ⫽ 57.163, P ⬍ 0.0001], but no main effect of session or any interaction between two factors (Fig. 5).
Changes in BOLD signals before and after daily study of K-words
Discussion
To avoid the possible confounds of nonspecific session and time effects, it is necessary to characterize the learningrelated effects in terms of interaction effects (i.e., session ⫻ condition interaction). To identify learning-related changes in blood oxygenation-level-dependent (BOLD) signals, we extracted mean BOLD signals within three activated regions from each task (task vs baseline) for each subject using region-of-interests (ROIs) with MarsBar (http://www.mrccbu.cam.ac.uk/Imaging/marsbar.html). We chose three regions, i.e., the left angular gyrus and the left inferior occipital gyrus found in the first fMRI session and the right superior occipital gyrus found in the second fMRI session, because the location of the area of maximum activation of the superior occipital gyrus found in the first session was very close to that found in the second session. Left angular gyrus (BA39; x, y, z ⫽ ⫺56, ⫺66, 22) The ANOVA showed a significant main effect of session [F(1, 11) ⫽ 25.042, P ⬍ 0.0001] and condition [F(2, 22) ⫽ 56.552, P ⬍ 0.0001] and an interaction between two factors [F(2, 22) ⫽ 37.890, P ⬍ 0.0001]. Consequently, we performed the following two two-way ANOVAs to find the cause of the interaction. The ANOVA with factors of session (first fMRI and second fMRI) and condition (K-words and J-words) showed a significant main effect of session [F(1, 11) ⫽ 17.988, P ⬍ 0.001] and condition [F(1, 11) ⫽
In this study we investigated the change in brain activation patterns in 12 native Japanese who had no previous knowledge of the Korean language or of written Korean words (Han-gul words or K-words) at two stages of word learning, i.e., the early stage of learning and the later stage when they seemed to have completely mastered the 20 K-words. Analysis of behavioral data showed significant shortening of reaction times for both the reading aloud and reading comprehension tests on day 15 and day 16 relative to day 1 and day 2, suggesting that the subjects had become skilled in oral reading and in accessing the meaning of the K-words as a result of daily practice. The most interesting finding of this study is the change in activation patterns as the subjects progressively mastered the pronunciation and meaning of a set of 20 K-words. When the subjects were exposed to the set on the second day, i.e., on the day after the first learning, K-word reading activated the left inferior occipital gyrus and the right superior occipital gyrus, suggesting that the words were processed visually (K1-J1). In contrast, the reading of a set of 20 kana words activated the left angular gyrus (J1-K1). The left angular gyrus is classically known as the visual word center (Dejerine, 1891, 1892; Geschwind, 1965; Damasio and Damasio, 1983). Thus, activation of the angular gyrus by kana-word reading is consistent with the classical view that this area is involved in written-word processing. That this area showed a significant interaction between session
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
(first and second fMRI) and condition (J-words and Kwords) suggests that, at the first scan, the K-words must have been treated as nonlinguistic visual patterns, whereas the same K-words came to be treated as linguistic symbols with definite pronunciation and meaning at the second scan, in the same way as the kana words did at both scans. Although the material was limited to only 20 foreign words, this change in activation patterns is consistent with the hypothesis that the difference between the activation patterns seen in the first (or native) and second language depends on the subject’s degree of proficiency in the latter (Perani et al., 1998). As already mentioned, the first J-word reading relative to the first K-word reading (J1-K1) activated the left angular gyrus. Also, we observed interaction in BOLD signals in this region between session and condition. Although on day 1, the day prior to the first fMRI, the pronunciation and meaning of the K-words had been taught to and memorized by all the subjects, they must have been less proficient at reading these words than the J-words. This speculation is supported by a slow reaction time (average of 6.1 s in the reading aloud test) to the K-words in oral reading and comprehension (as was shown in the behavioral data), which is far beyond the limit of the 4-s stimulus presentation time for reading during fMRI measurement. Thus, it is probable that the subjects treated most of the K-words more or less like nonstudied K-words, at least within the scope of presentation time. Since graphemic processing is cancelled out by the subtraction, the angular gyrus activation seen in J1-K1 can be assumed to be associated with either phonological or semantic processing of the J-words. By the second fMRI measurement, all the subjects had mastered the 20 K-words completely, which would explain why activation of the left angular gyrus was not found in the comparison between J-words and K-words during the second fMRI session. We speculate that, by this time, the 20 K-words were processed as linguistic symbols like kana words, resulting in the observed interaction between session and condition. The interaction found in the angular gyrus between session (first and second fMRI) and condition (J-words and Kn-words) is interesting in terms of speculation about the role of this gyrus. The Kn-words were unfamiliar to the subjects and were only seen by them during the two fMRI sessions. As the Kn-words selected were common Korean words, they inevitably contained some syllabograms that were also present in the K-words. In fact, of the 60 syllabograms used in the Kn-words, 23 (38.3%) were also used in the K-words. As the subjects had mastered the K-words by the time of the second fMRI session, the syllabograms that were also used in the Kn-words must have been processed as familiar and pronounceable Han-gul characters, making the second Kn-word reading qualitatively different from the first. However, as the subjects did not know the meaning of the Kn-words, or the pronunciation of the words as a whole,
7
we speculate that the interaction between session and condition (J-words and Kn-words) found in the left angular gyrus is related to syllabogram–sound association, but not to written word–sound association or written word–meaning association. Clinical data from Japanese cases support this speculation. As is well known to linguists and neuropsychologists, phonographic characters (kana) are used in combination with logographic ones (kanji) in the Japanese script system (Yamadori, 1975; Yamadori and Ikumura, 1975; Iwata, 1984). This hybrid use of different symbols for different levels of word features results in a very interesting dissociation between kana and kanji reading in brain-damaged patients. A left angular lesion produces severe difficulty in kana character reading, with much less marked difficulty in kanji reading (Yamadori, 1975, 1998; Yamadori and Ikumura, 1975; Iwata, 1984, 1985; Hirayama and Kawamura, 1993). On the other hand, a lesion that is situated in the inferior–posterior region of the left temporal lobe, further down from the angular gyrus, results in alexia that is selective for kanji characters (Yamadori, 1986; Iwata, 1984, 1985). Furthermore, in typical cases of angular alexia, single kana reading is commonly affected. Since a single kanji character functions as a word and carries a certain-meaning, sparing of kanji reading in angular damage indicates that the angular gyrus is mostly involved in syllabogram–sound association and not in written word–sound or written word– semantic association. These data are consistent with the present imaging study. Recently, in an extensive review of the brain mechanisms of language processing based on functional imaging studies, Price hypothesized that the function of the angular gyrus in reading is mainly semantic, because reading aloud relative to rest does not activate the angular gyrus, whereas when the meaning of words is accessed, activation is observed in the left posterior temporoparietal cortex, including the angular gyrus (Price, 2000). It may be necessary to employ not a simple rest, but a stimulus that is meaningless but not unlike a target word as a control state in order to activate the angular gyrus. In this regard, it is of interest to note that activation of the angular gyrus has not been found consistently in functional imaging studies. Some reports have found activation of the angular gyrus in reading (Jessen et al., 1999). In a positron emisson tomography (PET) study of Japanese kana reading, Law et al. (1991) found a significant increase of blood flow in the left angular gyrus. However, there are also quite a few reports of studies in which this gyrus was not activated in reading tasks (Howard et al., 1992; Price et al., 1994, 1996a, 1996b; Herbster et al., 1997; Hagoort et al., 1999). In Japanese kana reading, Sakurai reported that this gyrus was not activated in PET measurement (Sakurai et al., 2000). Using the fMRI method, Fujimaki et al. (1999) reported that kana reading activated Broca’s area, the insula, supramarginal gyrus, and posterior superior temporal area, but not the angular gyrus. In a magnetoencephalography study on sin-
8
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Fig. 4. The activation in the left inferior occipital gyrus revealed by contrasts between K-word reading and J-word reading in the early stage (K1 ⫺ J1). The top left image shows the axial image (z ⫽ ⫺6) and the top right one shows the rendering image on the lateral surface of the left hemisphere. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, k ⬎ 10 voxels). The bar graph shows percent signal changes in each condition during the first and second fMRI sessions. Abbreviations are the same as in Fig. 3.
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
9
Fig. 5. The activated right superior occipital gyrus revealed by contrasts between K-word reading and J-word reading in the late stage (K2-J2). The top left image shows the axial image (z ⫽ 40) and the top right one shows the rendering image on the lateral surface of the right hemisphere. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, k ⬎ 10 voxels). The bar graph shows percentage signal changes in each condition during the first and second fMRI sessions. Abbreviations are the same as in Fig. 3.
10
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
gle kana character reading, Koyama et al. (1998) reported that only one of six subjects showed activity in the left angular gyrus. At present it is difficult to reconcile these different results. Since different languages employ different script systems, cross-cultural studies such as ours, which utilize a different script system as a control, may be helpful in separating graphemic, phonologic, and semantic aspects of character/word processing. Can the activation in the angular gyrus be related to the activity of the neural representation of the visual word form? Because our study did not employ the paradigm that can isolate the stage of visual word-form detection from that of oral reading/comprehension, we cannot address this issue. However, it can be said that the areas usually activated when processing word form have mainly been located in the lateral occipital lobe away from the angular area (Petersen et al., 1990). Another possible interpretation of the angular activity would be in terms of working memory, since the supramarginal gyrus has been found to be activated in tasks that demand the temporary store of linguistic information (Paulesu et al., 1993). Our present paradigm did not require working memory for its performance. In addition, our review clearly showed that the regions reported to have been activated are much more anterior and do not extend caudally to the angular gyrus (Fujii, 1998).
Acknowledgments This work was supported by a grant-in-aid for scientific research from the Ministry of Education, Science and Culture of Japan and a grant from the Japan Society for the Promotion of Science.
References Chee, M.W.L., Tan, E.W.L., Thiel, T., 1999a. Mandarin and English single words processing studied with functional Magnetic Resonance Imaging. J. Neurosci. 19, 3050 –3056. Chee, M.W.L., Caplan, D., Soon, C.S., Sriram, N., Tan, E.W.L., Thiel, T., Weekes, B., 1999b. Processing of visually presented sentences in Mandarin and English studied with fMRI. Neuron 23, 127–137. Chee, M.W.L., Weekes, B., Lee, K.M., Soon, C.S., Schreiber, A., Hoon, J.J., Chee, M., 2000. Overlap and dissociation of semantic processing of Chinese characters, English words, and pictures: evidence from fMRI. Neuroimage 12, 392– 403. Damasio, A., Damasio, H., 1983. The anatomic basis of pure alexia. Neurology 33, 1573–1583. Dejerine, J., 1891. Sur un cas de ce´ cite verbale avec agraphie, suivi d’autopsie. C. R. Soc. Biol. 3, 197–201. Dejerine, J., 1892. Contribution a l’e´ tude anatomo-pathologique et clinique des diffe´ rentes varie´ te´ s de cecite verbale. C. R. Soc. Biol. 4, 61–90. Friston, K.J., Holmes, A.P., Worsley, K.J., Frith, C.D., Frackowiak, R.S.J., 1995. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Mapp. 2, 189 –210.
Friston, K.J., Holmes, A.P., Price, C.J., Buchel, C., Worsley, K.J., 1999. Multisubject fMRI studies and conjunction analyses. Neuroimage 10, 385–396. Fujii, T., 1998. Neural correlates of working memory. J. Psychol. Rev. 41, 157–171 [in Japanese]. Fujimaki, N., Miyauchi, S., Pu¨ tz, B., Sasaki, Y., Takino, R., Sakai, K., Tamada, T., 1999. Functional Magnetic Resonance Imaging of neural activity related to orthographic, phonological, and lexico-semantic judgments of visually presented characters and words. Hum. Brain Mapp. 8, 44 –59. Geschwind, N., 1965. Disconnexion syndromes in animals and man. Brain 88, 585– 644. Hagoort, P., Indefrey, P., Brown, C., Herzog, H., Steinmetz, H., Seitz, R.J., 1999. The neural circuitry involved in the reading of German words and pseudowords: a PET study. J. Cogn. Neurosci. 11, 383–398. Herbster, A.N., Mintun, M.A., Nebes, R.D., Becker, J.T., 1997. Regional cerebral blood flow during word and nonword reading. Hum. Brain Mapp. 5, 84 –92. Hirayama, K., Kawamura, M., 1993. MRI-Cerebral Topodiagnosis. Igakushoin Press, Tokyo, [in Japanese]. Howard, D., Patterson, K., Wise, R., Brown, W.D., Friston, K., Weiller, C., Frackowiak, R., 1992. The cortical localization of the lexicons: Positron Emission Tomography evidence. Brain 115, 1769 –1782. Illes, J., Francis, W.S., Desmond, J.E., Gabrieli, J.D.E., Glover, G.H., Poldrack, R., Lee, C.J., 1999. Convergent cortical representation of semantic processing in bilinguals. Brain Lang. 70, 347–363. Iwata, M., 1984. Kanji versus Kana: neuropsychological correlates of the Japanese writing system. Trends Neurosci. 8, 290. Iwata, M., 1985. Neural mechanism of reading and writing. Tokyo University Press, Tokyo. Jessen, F., Erb, M., Klose, U., Lotze, M., Grodd, W., Heun, R., 1999. Activation of human language processing brain regions after the presentation of random letter strings demonstrated with event-related functional magnetic resonance imaging. Neurosci. Lett. 270, 13–16. Klein, D., Zatorre, R.J., Milner, B., Meyer, E., Evans, A.C., 1994. Left putaminal activation when speaking a second language: evidence from PET. NeuroReport 5, 2295–2297. Klein, D., Milner, B., Zatorre, R.J., Meyer, E., Evans, A.C., 1995. The neural substrates underlying words generation: a bilingual functionalimaging study. Proc. Natl. Acad. Sci. USA 92, 2899 –2903. Koyama, S., Kakigi, R., Hoshiyama, M., Kitamura, Y., 1998. Reading of Japanese Kanji (morphograms) and Kana (syllabograms): a magnetoencephalographic study. Neuropsychologia 36, 83–98. Law, I., Iwao, K., Fujita, H., Lassen, N.A., Miura, S., Uemura, K., 1991. Left supramarginal/angular gyri activation during reading of syllabograms in the Japanese language. J. Neurolingusitics 6, 243–251. Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9, 97–113. Paulesu, E., Frith, C.D., Frackowiak, R.S.J., 1993. The neural correlates of the verbal component of working memory. Nature 362, 342–345. Perani, D., Paulesu, E., Sebastian, N., Dupoux, E., Dehaene, S., Bettinardi, V., Cappa, S.F., Ferruccio, F., Mehler, J., 1998. The bilingual brain: proficiency and age of acquisition of the second language. Brain 121, 1841–1852. Petersen, S.E., Fox, P.T., Snyder, A.Z., Raichle, M.E., 1990. Activation of extrastriate and frontal cortical areas by visual words and word-like stimuli. Nature 249, 1041–1044. Price, C.J., Wise, J.S., Watson, J.D.G., Patterson, K., Howard, D., Frackowiak, R.S.J., 1994. Brain activity during reading: The effects of exposure duration and task. Brain 177, 1255–1269. Price, C.J., Moore, C.J., Frackowiak, S.J., 1996a. The effect of varying stimulus rate and duration on brain activity during reading. Neuroimage 3, 40 –52. Price, C.J., Wise, R.J.S., Frackowiak, S.J., 1996b. Demonstrating the implicit processing of visually presented words and pseudowords. Cereb. Cortex 6, 62–70.
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11 Price, C.J., 2000. The anatomy of language: contributions from functional neuroimaging. J. Anat. 197, 335–359. Sakurai, Y., Momose, T., Iwata, M., Sudo, Y., Ohtomo, K., Kanazawa, I., 2000. Different cortical activity in reading of Kanji words, Kana words and Kana nonwords. Brain Res. Cogn. Brain Res. 9, 111–115. Talairach, J., Tournoux, P., 1988. Co-planar Stereotaxic Atlas of the Human Brain. Thieme, New York. Vanderplas, J.M., Garvin, E.A., 1959. The association value of random shapes. J. Exp. Psychol. 57, 147–154. Yamadori, A., 1975. Ideogram reading in alexia. Brain 98, 231–238.
11
Yamadori, A., Ikumura, G., 1975. Gentral (or conduction) aphasia in a Japanese patient. Cortex 11, 73– 82. Yamadori, A., 1986. Category specific alexia and a neuropsychological model of alexia, in: Henry, S.R.K., Hoosain, R. (Eds.), Linguistics, Psychology, and the Chinese language, Hong Kong University Press, Hong Kong, pp. 255–264. Yamadori, A., 1998. Aphasia in ideograph readers: the case of Japanese, in: Coppens, P., Lebrun, Y., Basso, A. (Eds.), Aphasia in Atypical Populations, Laurence Erlbaum Associates Press, Mahwah, NJ, pp. 143– 174.
www.elsevier.com/locate/ynimg
Changes in brain activation patterns associated with learning of Korean words by Japanese: an fMRI study Hyung-Suk Lee,a,* Toshikatsu Fujii,a Jiro Okuda,a Takashi Tsukiura,b,c Atsushi Umetsu,d Maki Suzuki,a Tatsuo Nagasaka,e Shoki Takahashi,d and Atsushi Yamadoria a
Division of Neuropsychology, Department of Disability Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan b The Japan Society for the Promotion of Science, Tokyo, Japan c Neuroscience Research Institute, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan d Department of Diagnostic Radiology, Tohoku University Graduate School of Medicine, Sendai, Japan e Division of Radiology, Tohoku University Hospital, Sendai, Japan Received 15 July 2002; revised 3 April 2003; accepted 24 April 2003
Abstract We used functional magnetic resonance imaging (fMRI) to explore the change in brain activation associated with the learning of Korean words written in Han-gul characters (K-words) by young Japanese at two stages. Subjects were 12 right-handed native Japanese without previous knowledge of Korean words and characters. On the first day they were taught the pronunciation and meaning of 20 K-words. Then, after the first fMRI session (on day 2), they were given a set of 20 cards with the words and corresponding photographs. They also received a tape and were instructed to memorize the 20 K-words by studying them every day until the day of the second fMRI session (day 16). During the fMRI sessions, 20 Japanese words written in kana syllabograms (J-words) and the 20 previously presented K-words, as well as 20 new K-words (Kn-words) were presented visually for silent reading. The first J-word reading, relative to the first K-word reading, showed activation in the left angular gyrus. K-word reading relative to J-word reading during both sessions showed activation in occipital regions. Within these activated areas, session by condition interaction was found only in the left angular gyrus. The interaction between session and condition resulted from the fact that the differences in blood oxygenation-level-dependent signals between K-words and J-words and between Kn-words and J-words were significantly greater in the first session than in the second session. From the results, we concluded that patterns of brain activation changed as the memory of the 20 K-words became fixed through daily practice and that reading of both Korean words and Japanese syllabograms engaged the left angular gyrus. © 2003 Elsevier Inc. All rights reserved. Keywords: Second language; Learning; Angular gyrus; Functional magnetic resonance imaging (fMRI)
Introduction Although there have been several studies of the differences and similarities between the first language as a mother tongue and a second language acquired later as a foreign language (Klein et al., 1994, 1995; Perani et al., 1998; Illes et al., 1999; Chee et al., 1999a, 1999b, 2000), no study has been conducted with regard to the changes in brain activa* Corresponding author. Division of Neuropsychology, Department of Disability Medicine, Tohoku University Graduate School of Medicine, 2-1 Seiryo-machi, Aoba-ku, Sendai 980-8575, Japan. Fax: ⫹81-22-717-7360. E-mail address: [email protected] (H.-S. Lee). 1053-8119/03/$ – see front matter © 2003 Elsevier Inc. All rights reserved. doi:10.1016/S1053-8119(03)00254-4
tion patterns associated with the learning of foreign words. We wondered how brain activation patterns would be affected when subjects intensively learned a limited number of foreign words that had previously been totally unknown to them in a short period of time. We anticipated that if the number of words the subjects were required to master was sufficiently small, they would be able to learn them well enough to enable us to analyze the relationship between the activated brain patterns during the processing of unfamiliar foreign words, familiar foreign words, and native words. The Korean and Japanese languages both employ a phonological script system in addition to Chinese ideographic characters to transcribe native pronunciation. Korean in-
2
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Fig. 1. Overall experimental design.
volves phonological symbols called Han-gul, a symbol representing a phoneme—similar to languages that use an alphabet. However, unlike the alphabetic system, these phonemic symbols are not arranged in a serial order, but are combined into a single form to represent a syllable. These syllabic units are spatially separated from each other. Thus, in a sense, Han-gul characters can also be regarded as syllabograms. Conversely, Japanese employs purely syllabic characters called kana. Unlike a Han-gul syllabogram, a kana syllabogram is an independent sign in itself and does not consist of phonemic units. In this study we investigated the change in brain activation pattern in 12 right-handed native Japanese without previous knowledge of Han-gul words who studied 20 words every day for 16 days with the goal of memorizing them.
card aloud; on the comprehension test, they were asked to verbally report the meaning of the words. The number of correct responses and the reaction times were measured for both tests (Fig. 1). Pre-fMRI study phase (day 1) Stimuli consisted of 20 three-syllable Korean words written in Han-gul characters (K-words), all of which were the names of familiar animals or common objects such as dolphin and bicycle. Each word was printed on a card (105 ⫻ 148 mm in size). On the reverse side of the card there was a photograph representing the meaning of the word (Fig. 2). One of the authors (HSL), a native Korean speaker, taught the subjects the pronunciation and meaning of these 20 K-words repeatedly, until each of them could remember the pronunciation and meaning of at least 18 of the 20 words (90% or more).
Methods Daily learning phase (from days 2 to 15) Subjects Twelve Japanese volunteers (ages 19 –23 years, 6 males and 6 females), who had no previous knowledge of the Korean language, participated in this study. All subjects were right-handed and scored above ⫹85 on the Edinburgh Handedness Inventory (Oldfield, 1971). They gave their written informed consent in accordance with the Declaration of Helsinki. Overall experimental design The experiment consisted of a pre-fMRI study phase (day 1), the first fMRI scanning (day 2), a consecutive learning phase (from day 2 to day 15), and the second fMRI scanning (day 16), which employed the same procedure as the first. In order to monitor how well subjects had learned the 20 K-words, they were given two performance tests on days 1, 2, 15, and 16 using 20 learning cards. On the reading aloud test, the subjects were asked to read each word on a
The second performance test for the 20 words was carried out after the first fMRI measurement on day 2. In addition, each of the subjects was given a set of the cards used in the learning session on the first day, and a magnetic tape on which the pronunciation of these 20 words had been recorded by the same Korean speaker. Subjects were required to study the pronunciation and meaning of the words at home, with the help of the cards and tape, for at least 10 min a day until day 15. They were also asked to record the time they spent on this exercise every day. Tasks during fMRI scanning The subjects performed the same task in two fMRI sessions on days 2 and 16. For this task, we prepared eight lists (list A, B, C, D, E, F, G, and H), each consisting of 10 stimuli. Lists A and E consisted of 10 meaningless figures (Vanderplas and Garvin, 1959). Lists B and F consisted of 10 familiar (studied) Korean words (K-words). Lists C and
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
3
Table 1 Results of reading aloud tests and of reading comprehension tests Day
(1) Reading aloud test Reaction time (s) The number of correct responses (2) Reading comprehension test Reaction time (s) The number of correct responses
1st
2nd
6.1 ⫾ 2.9 19.7 ⫾ 0.5
6.1 ⫾ 1.1 19.8 ⫾ 0.5
2.5 ⫾ 0.5 20 ⫾ 0
2.2 ⫾ 0.5 20 ⫾ 0
5.2 ⫾ 2.23 19.3 ⫾ 0.9
4.5 ⫾ 0.1 18.9 ⫾ 1.2
2.2 ⫾ 0.6 19.8 ⫾ 0.4
1.9 ⫾ 0.4 19.8 ⫾ 0.5
G consisted of 10 new (nonstudied) Korean words (Knwords), which had the same number of Han-gul syllabograms as the studied words. Lists D and H consisted of 10 Japanese words written in kana characters (J-words), which had the same meaning and same number of syllables as the K-words. During fMRI scanning, the lists were presented in order from A to H. Stimuli were projected through a projector and back-projected onto a screen placed just beyond the subject’s feet on the scanner table. Subjects saw the screen through a mirror fixed on a head cage. The presentation time of a stimulus was set at 4 s and the interstimulus interval was 1 s. The subjects were asked to read a projected stimulus silently if they were able to read it and simply to look at it if it was a word they could not read or a meaningless figure. Scanning methods All measurements were carried out using a 1.5-T General Electric Signa scanner. Subjects were positioned in the scanner with their head immobilized in support cushions. A gradient echo planar imaging (EPI) modified by an EPIBOLD pulse sequence was used for functional imaging with the following parameters: TR ⫽ 5000 ms, TE ⫽ 60 ms, field of view ⫽ 24 ⫻ 24 cm2, flip angle ⫽ 90°, matrix size ⫽ 64 ⫻ 64 pixels, slice thickness ⫽ 4 mm, interslice thickness ⫽ 1 mm. For each run, 84 scans (including the first 4 dummy scans) were carried out and 27 axial slices per scan were obtained.
15th
16th
Individual analysis (Level 1) A general linear model was applied to the functional data (Friston et al., 1995). Activation patterns for each task relative to the baseline were determined for each subject. Random effects analysis (Level 2) The weighted sum of the parameter estimates in the individual analysis constituted “contrast” images, which were used for the group analysis (Friston et al., 1999). The contrast images obtained by individual analysis represent each task relative to the baseline. Contrasts of the condition effects of each voxel were assessed using two-sample t test, resulting in a standard image. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, K ⬎ 10 voxels). In a region-of-interests (ROIs) analysis, we performed two-way repeated-measures ANOVAs, with session (first fMRI and second fMRI) and condition (K-words, Knwords, and J-words) as factors on each region.
Results Learning time spent by the subjects to memorize the K-words All the subjects reported that they spent at least 10 min every day mastering the 20 K-words. The mean of the time spent for this task as reported by them was 14.2 min a day.
Data analysis
Results of behavioral tests
Analysis of the fMRI data obtained involved three steps using Statistical Parametric Mapping 99 (SPM99, Wellcome Department of Cognitive Neurology, London, UK) implemented in Matlab 5 (Mathworks, Sherborn, MA, USA). First, 80 volumes in each session acquired from a subject were realigned to their first volume. Second, the realigned images from each subject were transformed and normalized into standard space (Talairach and Tournoux, 1988) using an EPI template and smoothed using a Gaussian kernel of FWHM ⫽ 10 mm. Third, statistical analysis was conducted at the following two levels.
We analyzed the results from two kinds of performance test obtained at four points of time for eight subjects using one-way ANOVA; we were unable to test the other four subjects on day 16. Table 1 shows the results. For the reading aloud test, there was a significant difference in the time required to read a word between day 1 and day 15 [F(1, 7) ⫽ 12.15, P ⬍ 0.025], day 1 and day 16 [F(1, 7) ⫽ 12.5, P ⬍ 0.01], day 2 and day 15 [F(1, 7) ⫽ 107.05, P ⬍ 0.005], and day 2 and day 16 [F(1, 7) ⫽ 100.4, P ⬍ 0.005]. There was no difference in the number of correct responses. For the reading comprehension test, there was
4
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Fig. 2. Example of a card used for memorizing a Korean word in the present study. On the surface of the card, three Han-gul syllables, ol-Pfe-Mi, are printed. The syllabogram ol, the character on the left, is composed of two phonemic signs, i.e., o and l. In fact, in the character ol, three phonemic signs (no sound sign, o, and l) are combined vertically. On the reverse side there is a picture of the meaning of ol-Pfe-Mi.
again a significant difference in the time required to produce a Japanese word in response to a stimulus between day 1 and day 15 [F(1, 7) ⫽ 18.06, P ⬍ 0.005], day 1 and day 16 [F(1, 7) ⫽ 17.69, P ⬍ 0.005], day 2 and day 15 [F(1, 7) ⫽ 90.83, P ⬍ 0.005], and day 2 and day 16 [F(1, 7) ⫽ 55.87, P ⬍ 0.005]. There was no difference in the number of correct responses.
right superior occipital gyrus (Table 2). Next, we compared reading of the J-words and K-words with reading of previously unpresented Korean words (Kn-words). These two comparisons (J1-Kn1 and K1-Kn1) failed to produce any significantly activated areas (Table 2).
Brain activation
To identify areas of brain activation during the late stage of learning of the K-words, we analyzed the data using the same method as that employed for the first analysis. Comparison of J-words with K-words (J2-K2) showed no activated areas (Table 2). A significantly activated area for the second reading of the K-words, compared to that of the J-words (K2-J2), was found in the right superior occipital gyrus (Table 2 and Fig. 5). Silent reading of the J-words and the K-words compared with the Kn-words (J2-Kn2 and K2-Kn2) failed to produce significant activation (Table 2).
Brain activation during the first fMRI (on day 2) To identify areas of brain activation during the early stage of K-word learning, we compared silent reading of J-words with silent reading of K-words at early exposure (J1-K1) and vice versa (K1-J1). Significantly activated areas during silent J-word reading, compared with silent K-word reading, were found in the left angular gyrus (Table 2 and Fig. 3). The converse contrast (K1-J1) showed activation in the left inferior occipital gyrus (Table 2 and Fig. 4) and the
Brain activation during the second fMRI (on day 16)
Fig. 3. The activation in the left angular gyrus revealed by contrasts between J-word reading and K-word reading in the early stage (J1-K1). The top left image shows the axial image (z ⫽ 22) and the top right one shows the rendering image on the lateral surface of the left hemisphere. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, k ⬎ 10 voxels). Bar graphs show percentage signal changes in each condition during the first and second fMRI sessions. The upper graph shows the signal changes in the three conditions (K-words, Kn-words, and J-words) during the two fMRI sessions. The lower left graph shows the signal changes in the two conditions (K-words and J-words) during the two fMRI sessions. The lower right graph shows the signal changes in the two conditions (Kn-words and J-words) during the two fMRI sessions. K1 ⫽ the reading of Korean words during the first fMRI, K2 ⫽ the reading of Korean words during the second fMRI, J1 ⫽ the reading of Japanese words during the first fMRI, J2 ⫽ the reading of Japanese words during the second fMRI, Kn1 ⫽ the presenting of new (nonstudied) Korean words during the first fMRI, Kn2 ⫽ the presenting of new (nonstudied) Korean words during the second fMRI.
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
5
6
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Table 2 Data for anatomical structures, the Talairach coordinates, and the z values of peak activation for K1-J1, J1-K1, K1-Kn1, J1-Kn1, K2-J2, J2-K2, K2-Kn2, J2-Kn2 (corrected, P ⬍ .05 K ⬎ 10 voxels) Regions (Brodmann’s area) J1-K1 L angular gyrus (39) K1-J1 L inferior occipital gyrus (18) R superior occipital gyrus (19) J1-Kn1 No activation K1-Kn1 No activation J2-K2 No activation K2-J2 R superior occipital gyrus (19) J2-Kn2 No activation K2-Kn2 No activation
x
y
z
z value
Voxels
⫺56
⫺66
22 5.15
31
⫺32 28
⫺94 ⫺76
⫺6 5.58 40 4.93
217 15
30
⫺78
40 5.44
46
121.052, P ⬍ 0.0001], and an interaction between two factors [F(1, 11) ⫽ 67.144, P ⬍ 0.0001]. The ANOVA with factors of session (first fMRI and second fMRI) and condition (Kn-words and J-words) showed a significant main effect of condition [F(1, 11) ⫽ 55.239, P ⬍ 0.0001], and an interaction between two factors [F(1, 11) ⫽ 14.095, P ⬍ 0.003], whereas no main effect of session was found (Fig. 3). These interactions between session and condition resulted from the fact that the differences in BOLD signals between K-words and J-words and between Kn-words and J-words were significantly greater in the first session than in the second session. Left inferior occipital gyrus (BA18; x, y, z ⫽ ⫺32, ⫺94, ⫺6) The ANOVA showed a significant main effect of condition [F(2, 22) ⫽ 123.599, P ⬍ 0.0001], but no main effect of session or any interaction between two factors (Fig. 4).
Note. K1, the reading of Korean words during the first fMRI; K2, the reading of Korean words during the second fMRI; J1, the reading of Japanese words during the first fMRI; J2, the reading of Japanese words during the second fMRI; Kn1, the presenting of new (nonstudied) Korean words during the first fMRI; Kn2, the presenting of new (nonstudied) Korean words during the second fMRI.
Right superior occipital gyrus (BA19; x, y, z ⫽ 30, ⫺78, 40) The ANOVA showed a significant main effect of condition [F(2, 22) ⫽ 57.163, P ⬍ 0.0001], but no main effect of session or any interaction between two factors (Fig. 5).
Changes in BOLD signals before and after daily study of K-words
Discussion
To avoid the possible confounds of nonspecific session and time effects, it is necessary to characterize the learningrelated effects in terms of interaction effects (i.e., session ⫻ condition interaction). To identify learning-related changes in blood oxygenation-level-dependent (BOLD) signals, we extracted mean BOLD signals within three activated regions from each task (task vs baseline) for each subject using region-of-interests (ROIs) with MarsBar (http://www.mrccbu.cam.ac.uk/Imaging/marsbar.html). We chose three regions, i.e., the left angular gyrus and the left inferior occipital gyrus found in the first fMRI session and the right superior occipital gyrus found in the second fMRI session, because the location of the area of maximum activation of the superior occipital gyrus found in the first session was very close to that found in the second session. Left angular gyrus (BA39; x, y, z ⫽ ⫺56, ⫺66, 22) The ANOVA showed a significant main effect of session [F(1, 11) ⫽ 25.042, P ⬍ 0.0001] and condition [F(2, 22) ⫽ 56.552, P ⬍ 0.0001] and an interaction between two factors [F(2, 22) ⫽ 37.890, P ⬍ 0.0001]. Consequently, we performed the following two two-way ANOVAs to find the cause of the interaction. The ANOVA with factors of session (first fMRI and second fMRI) and condition (K-words and J-words) showed a significant main effect of session [F(1, 11) ⫽ 17.988, P ⬍ 0.001] and condition [F(1, 11) ⫽
In this study we investigated the change in brain activation patterns in 12 native Japanese who had no previous knowledge of the Korean language or of written Korean words (Han-gul words or K-words) at two stages of word learning, i.e., the early stage of learning and the later stage when they seemed to have completely mastered the 20 K-words. Analysis of behavioral data showed significant shortening of reaction times for both the reading aloud and reading comprehension tests on day 15 and day 16 relative to day 1 and day 2, suggesting that the subjects had become skilled in oral reading and in accessing the meaning of the K-words as a result of daily practice. The most interesting finding of this study is the change in activation patterns as the subjects progressively mastered the pronunciation and meaning of a set of 20 K-words. When the subjects were exposed to the set on the second day, i.e., on the day after the first learning, K-word reading activated the left inferior occipital gyrus and the right superior occipital gyrus, suggesting that the words were processed visually (K1-J1). In contrast, the reading of a set of 20 kana words activated the left angular gyrus (J1-K1). The left angular gyrus is classically known as the visual word center (Dejerine, 1891, 1892; Geschwind, 1965; Damasio and Damasio, 1983). Thus, activation of the angular gyrus by kana-word reading is consistent with the classical view that this area is involved in written-word processing. That this area showed a significant interaction between session
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
(first and second fMRI) and condition (J-words and Kwords) suggests that, at the first scan, the K-words must have been treated as nonlinguistic visual patterns, whereas the same K-words came to be treated as linguistic symbols with definite pronunciation and meaning at the second scan, in the same way as the kana words did at both scans. Although the material was limited to only 20 foreign words, this change in activation patterns is consistent with the hypothesis that the difference between the activation patterns seen in the first (or native) and second language depends on the subject’s degree of proficiency in the latter (Perani et al., 1998). As already mentioned, the first J-word reading relative to the first K-word reading (J1-K1) activated the left angular gyrus. Also, we observed interaction in BOLD signals in this region between session and condition. Although on day 1, the day prior to the first fMRI, the pronunciation and meaning of the K-words had been taught to and memorized by all the subjects, they must have been less proficient at reading these words than the J-words. This speculation is supported by a slow reaction time (average of 6.1 s in the reading aloud test) to the K-words in oral reading and comprehension (as was shown in the behavioral data), which is far beyond the limit of the 4-s stimulus presentation time for reading during fMRI measurement. Thus, it is probable that the subjects treated most of the K-words more or less like nonstudied K-words, at least within the scope of presentation time. Since graphemic processing is cancelled out by the subtraction, the angular gyrus activation seen in J1-K1 can be assumed to be associated with either phonological or semantic processing of the J-words. By the second fMRI measurement, all the subjects had mastered the 20 K-words completely, which would explain why activation of the left angular gyrus was not found in the comparison between J-words and K-words during the second fMRI session. We speculate that, by this time, the 20 K-words were processed as linguistic symbols like kana words, resulting in the observed interaction between session and condition. The interaction found in the angular gyrus between session (first and second fMRI) and condition (J-words and Kn-words) is interesting in terms of speculation about the role of this gyrus. The Kn-words were unfamiliar to the subjects and were only seen by them during the two fMRI sessions. As the Kn-words selected were common Korean words, they inevitably contained some syllabograms that were also present in the K-words. In fact, of the 60 syllabograms used in the Kn-words, 23 (38.3%) were also used in the K-words. As the subjects had mastered the K-words by the time of the second fMRI session, the syllabograms that were also used in the Kn-words must have been processed as familiar and pronounceable Han-gul characters, making the second Kn-word reading qualitatively different from the first. However, as the subjects did not know the meaning of the Kn-words, or the pronunciation of the words as a whole,
7
we speculate that the interaction between session and condition (J-words and Kn-words) found in the left angular gyrus is related to syllabogram–sound association, but not to written word–sound association or written word–meaning association. Clinical data from Japanese cases support this speculation. As is well known to linguists and neuropsychologists, phonographic characters (kana) are used in combination with logographic ones (kanji) in the Japanese script system (Yamadori, 1975; Yamadori and Ikumura, 1975; Iwata, 1984). This hybrid use of different symbols for different levels of word features results in a very interesting dissociation between kana and kanji reading in brain-damaged patients. A left angular lesion produces severe difficulty in kana character reading, with much less marked difficulty in kanji reading (Yamadori, 1975, 1998; Yamadori and Ikumura, 1975; Iwata, 1984, 1985; Hirayama and Kawamura, 1993). On the other hand, a lesion that is situated in the inferior–posterior region of the left temporal lobe, further down from the angular gyrus, results in alexia that is selective for kanji characters (Yamadori, 1986; Iwata, 1984, 1985). Furthermore, in typical cases of angular alexia, single kana reading is commonly affected. Since a single kanji character functions as a word and carries a certain-meaning, sparing of kanji reading in angular damage indicates that the angular gyrus is mostly involved in syllabogram–sound association and not in written word–sound or written word– semantic association. These data are consistent with the present imaging study. Recently, in an extensive review of the brain mechanisms of language processing based on functional imaging studies, Price hypothesized that the function of the angular gyrus in reading is mainly semantic, because reading aloud relative to rest does not activate the angular gyrus, whereas when the meaning of words is accessed, activation is observed in the left posterior temporoparietal cortex, including the angular gyrus (Price, 2000). It may be necessary to employ not a simple rest, but a stimulus that is meaningless but not unlike a target word as a control state in order to activate the angular gyrus. In this regard, it is of interest to note that activation of the angular gyrus has not been found consistently in functional imaging studies. Some reports have found activation of the angular gyrus in reading (Jessen et al., 1999). In a positron emisson tomography (PET) study of Japanese kana reading, Law et al. (1991) found a significant increase of blood flow in the left angular gyrus. However, there are also quite a few reports of studies in which this gyrus was not activated in reading tasks (Howard et al., 1992; Price et al., 1994, 1996a, 1996b; Herbster et al., 1997; Hagoort et al., 1999). In Japanese kana reading, Sakurai reported that this gyrus was not activated in PET measurement (Sakurai et al., 2000). Using the fMRI method, Fujimaki et al. (1999) reported that kana reading activated Broca’s area, the insula, supramarginal gyrus, and posterior superior temporal area, but not the angular gyrus. In a magnetoencephalography study on sin-
8
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
Fig. 4. The activation in the left inferior occipital gyrus revealed by contrasts between K-word reading and J-word reading in the early stage (K1 ⫺ J1). The top left image shows the axial image (z ⫽ ⫺6) and the top right one shows the rendering image on the lateral surface of the left hemisphere. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, k ⬎ 10 voxels). The bar graph shows percent signal changes in each condition during the first and second fMRI sessions. Abbreviations are the same as in Fig. 3.
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
9
Fig. 5. The activated right superior occipital gyrus revealed by contrasts between K-word reading and J-word reading in the late stage (K2-J2). The top left image shows the axial image (z ⫽ 40) and the top right one shows the rendering image on the lateral surface of the right hemisphere. The threshold of significance for effects of conditions was set at P ⬍ 0.05 (corrected, k ⬎ 10 voxels). The bar graph shows percentage signal changes in each condition during the first and second fMRI sessions. Abbreviations are the same as in Fig. 3.
10
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11
gle kana character reading, Koyama et al. (1998) reported that only one of six subjects showed activity in the left angular gyrus. At present it is difficult to reconcile these different results. Since different languages employ different script systems, cross-cultural studies such as ours, which utilize a different script system as a control, may be helpful in separating graphemic, phonologic, and semantic aspects of character/word processing. Can the activation in the angular gyrus be related to the activity of the neural representation of the visual word form? Because our study did not employ the paradigm that can isolate the stage of visual word-form detection from that of oral reading/comprehension, we cannot address this issue. However, it can be said that the areas usually activated when processing word form have mainly been located in the lateral occipital lobe away from the angular area (Petersen et al., 1990). Another possible interpretation of the angular activity would be in terms of working memory, since the supramarginal gyrus has been found to be activated in tasks that demand the temporary store of linguistic information (Paulesu et al., 1993). Our present paradigm did not require working memory for its performance. In addition, our review clearly showed that the regions reported to have been activated are much more anterior and do not extend caudally to the angular gyrus (Fujii, 1998).
Acknowledgments This work was supported by a grant-in-aid for scientific research from the Ministry of Education, Science and Culture of Japan and a grant from the Japan Society for the Promotion of Science.
References Chee, M.W.L., Tan, E.W.L., Thiel, T., 1999a. Mandarin and English single words processing studied with functional Magnetic Resonance Imaging. J. Neurosci. 19, 3050 –3056. Chee, M.W.L., Caplan, D., Soon, C.S., Sriram, N., Tan, E.W.L., Thiel, T., Weekes, B., 1999b. Processing of visually presented sentences in Mandarin and English studied with fMRI. Neuron 23, 127–137. Chee, M.W.L., Weekes, B., Lee, K.M., Soon, C.S., Schreiber, A., Hoon, J.J., Chee, M., 2000. Overlap and dissociation of semantic processing of Chinese characters, English words, and pictures: evidence from fMRI. Neuroimage 12, 392– 403. Damasio, A., Damasio, H., 1983. The anatomic basis of pure alexia. Neurology 33, 1573–1583. Dejerine, J., 1891. Sur un cas de ce´ cite verbale avec agraphie, suivi d’autopsie. C. R. Soc. Biol. 3, 197–201. Dejerine, J., 1892. Contribution a l’e´ tude anatomo-pathologique et clinique des diffe´ rentes varie´ te´ s de cecite verbale. C. R. Soc. Biol. 4, 61–90. Friston, K.J., Holmes, A.P., Worsley, K.J., Frith, C.D., Frackowiak, R.S.J., 1995. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Mapp. 2, 189 –210.
Friston, K.J., Holmes, A.P., Price, C.J., Buchel, C., Worsley, K.J., 1999. Multisubject fMRI studies and conjunction analyses. Neuroimage 10, 385–396. Fujii, T., 1998. Neural correlates of working memory. J. Psychol. Rev. 41, 157–171 [in Japanese]. Fujimaki, N., Miyauchi, S., Pu¨ tz, B., Sasaki, Y., Takino, R., Sakai, K., Tamada, T., 1999. Functional Magnetic Resonance Imaging of neural activity related to orthographic, phonological, and lexico-semantic judgments of visually presented characters and words. Hum. Brain Mapp. 8, 44 –59. Geschwind, N., 1965. Disconnexion syndromes in animals and man. Brain 88, 585– 644. Hagoort, P., Indefrey, P., Brown, C., Herzog, H., Steinmetz, H., Seitz, R.J., 1999. The neural circuitry involved in the reading of German words and pseudowords: a PET study. J. Cogn. Neurosci. 11, 383–398. Herbster, A.N., Mintun, M.A., Nebes, R.D., Becker, J.T., 1997. Regional cerebral blood flow during word and nonword reading. Hum. Brain Mapp. 5, 84 –92. Hirayama, K., Kawamura, M., 1993. MRI-Cerebral Topodiagnosis. Igakushoin Press, Tokyo, [in Japanese]. Howard, D., Patterson, K., Wise, R., Brown, W.D., Friston, K., Weiller, C., Frackowiak, R., 1992. The cortical localization of the lexicons: Positron Emission Tomography evidence. Brain 115, 1769 –1782. Illes, J., Francis, W.S., Desmond, J.E., Gabrieli, J.D.E., Glover, G.H., Poldrack, R., Lee, C.J., 1999. Convergent cortical representation of semantic processing in bilinguals. Brain Lang. 70, 347–363. Iwata, M., 1984. Kanji versus Kana: neuropsychological correlates of the Japanese writing system. Trends Neurosci. 8, 290. Iwata, M., 1985. Neural mechanism of reading and writing. Tokyo University Press, Tokyo. Jessen, F., Erb, M., Klose, U., Lotze, M., Grodd, W., Heun, R., 1999. Activation of human language processing brain regions after the presentation of random letter strings demonstrated with event-related functional magnetic resonance imaging. Neurosci. Lett. 270, 13–16. Klein, D., Zatorre, R.J., Milner, B., Meyer, E., Evans, A.C., 1994. Left putaminal activation when speaking a second language: evidence from PET. NeuroReport 5, 2295–2297. Klein, D., Milner, B., Zatorre, R.J., Meyer, E., Evans, A.C., 1995. The neural substrates underlying words generation: a bilingual functionalimaging study. Proc. Natl. Acad. Sci. USA 92, 2899 –2903. Koyama, S., Kakigi, R., Hoshiyama, M., Kitamura, Y., 1998. Reading of Japanese Kanji (morphograms) and Kana (syllabograms): a magnetoencephalographic study. Neuropsychologia 36, 83–98. Law, I., Iwao, K., Fujita, H., Lassen, N.A., Miura, S., Uemura, K., 1991. Left supramarginal/angular gyri activation during reading of syllabograms in the Japanese language. J. Neurolingusitics 6, 243–251. Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9, 97–113. Paulesu, E., Frith, C.D., Frackowiak, R.S.J., 1993. The neural correlates of the verbal component of working memory. Nature 362, 342–345. Perani, D., Paulesu, E., Sebastian, N., Dupoux, E., Dehaene, S., Bettinardi, V., Cappa, S.F., Ferruccio, F., Mehler, J., 1998. The bilingual brain: proficiency and age of acquisition of the second language. Brain 121, 1841–1852. Petersen, S.E., Fox, P.T., Snyder, A.Z., Raichle, M.E., 1990. Activation of extrastriate and frontal cortical areas by visual words and word-like stimuli. Nature 249, 1041–1044. Price, C.J., Wise, J.S., Watson, J.D.G., Patterson, K., Howard, D., Frackowiak, R.S.J., 1994. Brain activity during reading: The effects of exposure duration and task. Brain 177, 1255–1269. Price, C.J., Moore, C.J., Frackowiak, S.J., 1996a. The effect of varying stimulus rate and duration on brain activity during reading. Neuroimage 3, 40 –52. Price, C.J., Wise, R.J.S., Frackowiak, S.J., 1996b. Demonstrating the implicit processing of visually presented words and pseudowords. Cereb. Cortex 6, 62–70.
H.-S. Lee et al. / NeuroImage 20 (2003) 1–11 Price, C.J., 2000. The anatomy of language: contributions from functional neuroimaging. J. Anat. 197, 335–359. Sakurai, Y., Momose, T., Iwata, M., Sudo, Y., Ohtomo, K., Kanazawa, I., 2000. Different cortical activity in reading of Kanji words, Kana words and Kana nonwords. Brain Res. Cogn. Brain Res. 9, 111–115. Talairach, J., Tournoux, P., 1988. Co-planar Stereotaxic Atlas of the Human Brain. Thieme, New York. Vanderplas, J.M., Garvin, E.A., 1959. The association value of random shapes. J. Exp. Psychol. 57, 147–154. Yamadori, A., 1975. Ideogram reading in alexia. Brain 98, 231–238.
11
Yamadori, A., Ikumura, G., 1975. Gentral (or conduction) aphasia in a Japanese patient. Cortex 11, 73– 82. Yamadori, A., 1986. Category specific alexia and a neuropsychological model of alexia, in: Henry, S.R.K., Hoosain, R. (Eds.), Linguistics, Psychology, and the Chinese language, Hong Kong University Press, Hong Kong, pp. 255–264. Yamadori, A., 1998. Aphasia in ideograph readers: the case of Japanese, in: Coppens, P., Lebrun, Y., Basso, A. (Eds.), Aphasia in Atypical Populations, Laurence Erlbaum Associates Press, Mahwah, NJ, pp. 143– 174.