Neuroimaging studies with printed words have well established that word recognition provokes orchestrated brain activity that occurs within a number of neuro-anatomical subsystems such as orthography (letter and letter combination) and phonology (speech sound) [1–3]. At present, central questions of controversy are where these subsystems are represented in the brain and how they are associated with one another. One powerful tool for exploring the cerebral loci of phonological processing and the transformation of a written word's orthographic form into phonological form is the regularity effect, which refers to the activation difference between reading loud of regular words following usual letter–sound correspondence rules (e.g., save, gave, wave) and reading loud of irregular words with ambiguous letter–sound correspondences (e.g., have). Previous efforts to map written English words’ phonological functioning as measured by the regularity effect have identified several distinct brain regions including left inferior frontal gyrus [4,5], left superior temporal cortex , right inferior frontal areas  and bilateral motor cortex . Recent studies with other languages suggest that language discrepancies in orthography and the reliability of letter–sound mapping constitute important sources of constraint on the cortical organization of phonological information [8–10].
The present study was aimed to extend this line of research by using written Chinese, a logographic system that differs markedly from alphabetic languages such as English and Italian. Written Chinese uses characters as a basic writing unit, which have a square configuration and map onto morphemes (meaning) rather than phonemes (minimal sound units as represented by English letters) in the spoken language. A Chinese character's phonology is defined at the monosyllabic level, with no parts in the character corresponding to phonological segments such as phonemes. Thus, regular or quasi-regular letter–sound rules that exist in all alphabetic languages are impossible in Chinese . However, Chinese is a tonal language using four different tones as suprasegmental phonological features that change the pitch of the syllable and are semantically contrastive. For example, the character SYMBOL is pronounced /ma3/ (the number following the pronunciation refers to tone), meaning horse. It differs from the character SYMBOL (pronounced /ma1/, meaning mother) in tone, though both pronunciations share the same consonant and vowel. A majority of characters are used as words, which are analyzable into internal sublexical structures composed of a phonetic component that might hint at characters’ pronunciation and a semantic component that might suggest characters’ meaning. Previous analysis of Chinese corpus indicates that < 30% of characters are pronounced identically as their phonetic components . More importantly, it is never the case in Chinese that the phonetic component is a segment that maps onto a substring of the word's phonological form in the way that a letter (or letter combination) maps onto a substring in an alphabetic system. For instance, in beech, the b corresponds to /b/, a phoneme, and the latter is a segment of the word's phonological form /biytc/. In the Chinese character SYMBOL (pronounced /li3/, meaning reason), its phonetic component SYMBOL (pronounced /li3/ as well, meaning inside) does not correspond to a piece of the character's phonological form. Rather, it is the syllable that may be the same at character and sub-character levels. Despite the ambiguity of Chinese orthography that complicates mapping of phonetic components to character sounds, a recent study  suggested that phonologic codes of phonetic components are accessed rapidly in naming whole characters. This finding has been bolstered by neuropsychological evidence from brain-damaged patients .
Because written Chinese is notably different from alphabetic languages in orthography (square shape) and phonology (no letter–sound correspondence but with tone), research with Chinese is important to advance our understanding of the universality and particularity of the organization of phonological subsystems in the brain. The present study utilized whole-brain, event-related (ER-) fMRI to observe the transient brain response  associated with Chinese reading. Past studies of regularity effects with alphabetic words all used a block design, in which regular words were exposed in one block, and irregular words in another block. Regularity effects obtained in such a design could arise from strategic processing of words. In this study, we used ER-fMRI, in which presentations of regular and irregular Chinese single character words were randomized. This design permits us to examine the neural substrate relevant to automatic processing of sublexical phonological information.
MATERIALS AND METHODS
Ten male right-handed volunteers gave their informed consent to participate in the experiment, which was approved by a local ethics committee at the University of Texas Health Science Center at San Antonio (UTHSCSA). All subjects were native Chinese Mandarin speakers from Mainland China, ranging in age from 29 to 39 years and living in the USA for no more than 6 years.
Apparatus and procedure:
This study was performed on a 1.9 T GE/Elscint Prestige whole-body MRI scanner (Elscint Ltd., Haifa, Israel) at the Research Imaging Center at UTHSCSA. Functional scans were obtained by using a single-shot T2*-weighted gradient-echo echo planar imaging (EPI) sequence, with slice thickness = 6 mm, in-plane resolution = 2.9 × 2.9 mm, and TR/TE/θ = 1000 ms/45 ms/90°. The field of view was 373 × 210 mm, and the acquisition matrix was 128 × 72. Twenty contiguous axial slices were acquired to cover the whole brain in two runs, with 10 slices acquired in each run. Scanning was event-related, with acquisition of images synchronized to stimulus presentation. For each slice, 610 images were acquired with a total scan time of 610 s. The anatomical MRI was acquired using a T1-weighted, three-dimensional, gradient-echo pulse-sequence. This sequence provided high-resolution (1 × 1 × 1 mm) images of the entire brain.
Materials and behavioral performance:
We used 20 regular words whose phonetic components are pronounced the same as the whole words and 20 irregular words whose phonetic components are pronounced differently from the whole words. Based on previous findings of greater regularity effects for higher frequency words than for lower frequency words , highly familiar Chinese words were used in this study. Visual complexity and onset were matched across the two sets of words. Figure 1 illustrates examples of experimental items. The stimuli were shown through a LED projector system. Each word was presented for 250 ms, followed by a fixation cross for 19.75 s. The subject was required to read aloud the viewed word. Stimuli were presented in two imaging runs of 20 trials each.
We used MatLab and in-house software for image processing , including corrections for head motion and global MRI signal shift. Skull stripping of the 3D MRI T1-weighted images was performed using Alice software. These images were spatially normalized to the Talairach brain atlas using a Convex Hull algorithm .
Functional images were sorted into regular and irregular word groups. For each voxel, 10 trials were averaged to generate a 20 s time course with a temporal resolution of 1 s. Timing of each slice was corrected by the sequential order of multi-slice acquisition, using a linear interpretation. By assuming that the timing of the hemodynamic response is similar to that of previous results from a BOLD ER-fMRI study , the data from the 4 to 7 s after stimulation were set as the activation state, and those from 16 to 19 s as the resting (baseline) state. Activation maps were calculated by comparing images acquired during each task activation state (reading aloud of regular and irregular words) with those acquired during the resting state, using a Student's group t-test. The first 10 images acquired in each trial were discarded to ensure that the MR signal reached the steady state. Both T1-weighted anatomical images and activation maps were spatially normalized into Talairach space using the Convex Hull algorithm. The averaged activation maps across the 10 subjects with a t value threshold of 3.14 (p < 0.01) were then overlaid on the corresponding T1 images. For each condition, Talairach coordinates of the center-of-mass and volume (mm3) of the activation clusters were determined based on the averaged activation maps.
The fMRI images averaged across subjects for the regular words vs fixation comparison and the irregular words vs fixation comparison are shown in Fig. 2. Significant areas of activation are summarized in Table 1. A large common neuroanatomical network was activated by the two types of Chinese stimuli, which includes left inferior frontal gyri (Brodmann areas (BAs) 44/9, 45/47) and its right homologue, bilateral premotor cortex, left and right middle frontal gyri (BAs 9, 10/46, 6/8), left superior frontal cortex, left anterior cingulate cortex (BAs 32, 24/23), bilateral superior temporal gyri (BAs 22, 38), left inferior parietal lobule and its right homologue area (BA 40), left precuneus (BA 7), right postcentral gyrus (BA 43/3, 40), left lingual gyrus in the occipital cortex (BA 17). Cerebellum was also active.
Direct comparisons of regular and irregular words indicated that left medial frontal gyrus (BA 6), right supero-medial frontal gyrus (BAs 8, 6), bilateral superior parietal lobule (BA 7), bilateral cuneus in the visual system (BAs 18, 19) and thalamus were active during reading irregular rather than regular words. There were no interesting areas unique to reading of regular words.
Based on previous work on English regularity effects [4–6] and Chinese reading [18–20], we selected left infero-frontal cortex and its right homologue areas, left middle frontal cortex (BA 9), bilateral (pre-)motor areas including supplementary motor area, bilateral temporal gyri (BA 22 and its vicinity), and anterior cingulate cortex as regions of interest to determine the regularity effect. Compared to regular words, reading aloud of irregular words resulted in significantly greater brain activation (by volume) in a one-tailed t-test in the following regions (Fig. 3): left infero-frontal cortex covering BAs 44/9 and 45/47 (t(9) = −2.37, p = 0.02), left motor cortex (t(9) = −2.99, p = 0.009), left middle frontal gyrus at BA 9 (t(9) = −2.02, p = 0.037), left temporal lobe (t(9) = −2.16, p = 0.03) and its right homologue area (t(9) = −2.23, p = 0.026), and left anterior cingulate cortex (t(9) = −2.38, p = 0.02). Right inferior frontal regions showed a marginally significant effect (t(9) = −1.81, p = 0.052). There was no reliable effect in right motor cortex (t(9) = −1.01, p = 0.17.
Our results from reading in Chinese replicate and extend previous findings of reading in English and other alphabetic languages. Converging with the studies of alphabetic languages [4–10], this study found regularity effects in left inferior frontal regions (BAs 44/9, 45/47), left (pre)-motor cortex including supplementary motor area, and left superior temporal lobe, suggesting their contribution to orthographic to phonological transformation that is presumably general across languages. The left inferior frontal and temporal cortice may be related to a general phonological analysis of stimuli, while the motor cortex is responsible for phonological output at the articulatory stage. The greater activations observed in these areas for irregular words arose from the competition of phonological codes and/or articulatory gestures at the phonetic component level and at the whole word level.
The left middle frontal cortex at BA 9 was shown to mediate phonological processing. In the existing literature, this region has not been identified by most of the studies with alphabets. Our previous study using written Chinese, however, has demonstrated that, relative to a crosshair baseline, peak activation during word generation was located in this area . Thus, the BA 9 is a very important area in reading logographs. The regularity effect as seen in this region indicated its participation in orthography-to-phonology transformation.
Another important finding of this ER-fMRI study is that the right hemisphere is heavily involved in the reading aloud of Chinese logographs. The regularity effect was marginally significant in right inferior frontal regions, and was highly reliable in right anterior superior temporal cortex. These right hemisphere sites have been implicated very rarely in past work with alphabetic words (but see ). The right inferior frontal gyri are known to service Chinese reading  and mediate episodic memory processes by which one retrieves spatial features of perceived objects [21,22]. We believe their participation in orthography to phonology mapping is relevant to the unique square shape of Chinese words that demands intense fine-grained analysis of spatial properties. Regarding the right superior temporal lobe, this area is known to be relevant to perception and analysis of pitch and tone . We assume that its activity here is associated with tonal representations and processing of Chinese logographs. The regularity effect observed in this area indicated the involvement of a general phonological analysis subsystem to which the tonal information is attached.
Our demonstration of regularity effects in the anterior cingulate cortex is important. It supports the hypothesis that the anterior cingulate cortex plays a prominent role in the executive control of cognition and the online monitoring and evaluating of performance by detecting cognitive states such as response competition [24,25]. Greater activation in this area for irregular words than for regular words is assumed to stem from the activation of the phonological code of phonetic components in an irregular word, which competes with the activation of the phonological code of the whole word. Further, it may be also relevant to the competition between articulatory gestures of the word and its phonetic component. As an evaluation module, the anterior cingulate cortex detects such kinds of competition. Given that the present study used ER-fMRI in which presentation of regular and irregular words was randomized, access to phonological information of phonetic components is hypothetically automatic. In such cases, the anterior cingulate cortex is involved in language and sound organization when strategic processes are less engaged .
The right superior frontal and parietal regions (BAs 8/6, 7) and the bilateral cuneus (BA 19/18) in the visual system mediated the reading of irregular, but not regular, words. Previous research indicates that these areas participate in allocation of attention resources and analysis of visuo-spatial information of objects [21,22]. Their involvement in the processing of irregular words implies that evaluation of the appropriateness of the phonological code of phonetic components is associated with a further analysis of orthographic units of printed words.
Our fMRI findings are important for elucidating the process underlying the reading of words; in particular, Chinese characters. The regularity effect detected with high frequency Chinese stimuli provides strong neural evidence for the most recent eye movement results . Consistent with the assumptions of the cognitive model of reading Chinese , our fMRI results indicate that during the reading loud of a familiar Chinese word, its phonological information is processed obligatorily at sub-word componential level. According to our findings, the reading aloud of a Chinese word hypothetically involves four sub-processes: (1) general phonological processing, which activates phonological information predominantly by left infero-middle frontal and bilateral temporal cortices; (2) tonal analysis, which is mediated by right superior temporal gyri and is characterized by the nature of Chinese; (3) articulatory production, which is serviced by the left motor cortex; and (4) evaluation of phonological codes and production gestures, which is mediated by the anterior cingulate cortex. Through the evaluation sub-process, the consistency of phonetic components’ and whole words’ phonological codes is decided. When there is a discrepancy between activated phonological representations or articulatory gestures, such as in irregular words, competition occurs, which leads to additional analysis of visual-orthographic and phonological information of printed words. Right superior frontal and parietal gyri and bilateral cuneus are recruited for this further checking of orthography to phonology mappings.
This ER-fMRI study intended to examine the neural correlates of phonological processing in reading words aloud, in particular, Chinese characters. The regularity effect was employed as a vehicle for achieving this goal. Greater activations for irregular words compared to regular ones were observed in left infero-frontal regions, left middle frontal area at BA 9, left motor cortex, bilateral anterior superior temporal gyri, and anterior cingulate cortex. Right superior frontal and parietal regions (BAs 8/6, 7) and bilateral cuneus in the visual cortex contributed to the reading of irregular, but not regular, words. The function of these cortical regions is discussed in connection to the current models of reading. Parts of the neural circuitry involved in reading Chinese are similar to what has been reported for alphabets. However, reading aloud of Chinese words is more closely associated with the activity of the right hemisphere sites including infero- frontal and superior temporal gyri. This is attributed to the unique visual-spatial and tonal analyses demanded by Chinese. The regularity effect seen in left BA 9 extends our past finding, indicating its contribution to orthography to phonology mapping. Our results agree with the idea that different native languages may shape different neural systems.
This research is supported by grant 28502500 from the University of Hong Kong Foundation for Educational Development and Research.
1. Petersen SE, Fox PT, Posner MI. et al
. Nature 1988 331, 585–589.
2. Demonet JF, Chollet F, Ramsay S. et al
. Brain 1992 115, 1753–1768.
3. Vandenberghe R, Price CJ, Wise R. et al
. Nature 1996 383, 254–256.
4. Herbster AN, Mintun MA, Nebes RD. et al
. Hum Brain Mapp 1997 5, 84–92.
5. Fiez JA, Balota D, Raichle M. et al
. Neuron 1999 24, 205–218.
6. Rumsey JM, Horwitz B, Donohue C. et al
. Brain 1997 119, 739–759.
7. Pugh KR, Shaywitz BA, Shaywitz SE. et al
. J Exp Psychol Hum Percept Perform 1997 23, 299–318.
8. Hagoort P, Indefrey P, Brown C. et al
. J Cogn Neurosci 1999 11, 383–398.
9. Paulesu E, McCrory E, Fazio F. et al
. Nature Neurosci 2000 3, 91–96.
10. Price CJ, Green DW, von Studnitz R. Brain 2000 122, 2221–2235.
11. Tan LH, Perfetti CA. Read Writ 1998 10, 165–200.
12. Pollatsek A, Tan LH, Rayner K. J Exp Psychol Hum Percept Perform 2000 26, 607–633.
13. Patterson K, Suzuki T, Wydell T. et al
. Neurocase 1995 1, 155–165.
14. Buckner RL, Bandettini PA, O'Craven KM. et al
. Proc Natl Acad Sci USA 1996 93, 14878–14883.
15. Xiong J, Gao JH, Lancaster JL. et al
. Hum Brain Mapp 1995 3, 209–223.
16. Lancaster JL, Fox PT, Downs H. et al
. J Nucl Med 1999 40, 942–955.
17. Rosen BR, Buckner RL, Dale AM. Proc Natl Acad Sci USA 1998 95, 773–780.
18. Chee M, Tan E, Thiel T. J Neurosci 1999 19, 3050–3056.
19. Tan LH, Spinks JA, Gao JH. et al
. Hum Brain Mapp 2000 10, 27–39.
20. Klein D, Milner B, Zatorre RJ. et al
. NeuroReport 1999 10, 2841–2846.
21. Lepage M, Ghaffar O, Nyberg L. et al
. Proc Natl Acad Sci USA 2000 97, 506–511.
22. Haxby JV, Ungerleider LG, Horwitz B. et al
. Hum Brain Mapp 1995 3, 68–82.
23. Zatorre RJ, Evans AC, Meyer E. et al
. Science 1992 256, 846–849.
24. Posner MI, Dahaene S. Trends Neurosci 1994 17, 75–79.
25. Carter CA, MacDonald AM, Botvinick M. et al
. Proc Natl Acad Sci USA 2000 97, 1944–1948.