The Linguistic Nature of Kanji Reexamined: Do Kanji Represent Only Meanings? Author(s): Sachiko Matsunaga Source: The Journal of the Association of Teachers of Japanese, Vol. 30, No. 2 (Oct., 1996), pp. 1-22 Published by: American Association of Teachers of Tapanese Stable URL: http://www.jstor.org/stable/489563 Accessed: 20/03/2014 13:29 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. 1 STOR American Association of Teachers of Japanese is collaborating with JSTOR to digitize, preserve and extend access to The Journal of the Association of Teachers of Japanese. http://www.jstor.org This content downloaded from 147.251.234.79 on Thu, 20 Mar 2014 13:29:01 PM All use subject to JSTOR Terms and Conditions Journal of the Association of Teachers of Japanese | 1 ARTICLES The Linguistic Nature of Kanji Reexamined: Do Kanji Represent Only Meanings? by Sachiko Matsunaga Introduction Kanji, or Chinese characters, in contrast with well-known phonetic writing systems, are commonly called "pictographs," "ideographs," "logographs," or "morphographs," terms that define kanji as written symbols which solely represent objects, ideas, words, and morphemes, respectively. These terms have been used by those who claim that kanji symbolize meanings independent of sounds (e.g., Wang, 1973; Suzuki, 1975), and those who believe that fluent readers of Chinese and Japanese read kanji without relying on sounds (e.g., Morioka, 1968; Smith, 1985). These beliefs, however, seem to have been falsified as a myth both linguistically (Nomura & Itô, 1978; Itô, 1979; DeFrancis, 1984a, 1984b, 1989) and psycholinguistically (Tzeng, Hung & Wang, 1977; Horodeck, 1987; Perfetti & Zhang, 1991; Cheng, 1992; Matsunaga, 1995). Nevertheless, the belief in this ideographic myth still appears to be strong among certain scholars (e.g., Hansen, 1993) and non-scholars alike, triggering renewed philosophical, historical, linguistic, psycholinguistic, and pedagogical counter arguments. Erbaugh (1995), for example, criticizes the post-structuralists' view which characterizes kanji as "paintings rather than words to be spoken" (p. 265), by saying "'language' in China and Japan [in their view] is somehow reduced to poetry in writing" (p. 265), with formal structures such as pronunciation and rhyme being ignored. Unger (1995) documents how uncritically historians have accepted the ideographic myth, and how such an acceptance has distorted their interpretations of events. Vance (1995) notes discrepancies between meanings associated with the Chinese readings (on'yomi) and the Japanese readings (kun'yomt) for the same kanji, which should not be the case if kanji were ideographs. Tzeng and Hung (1995) reject the hypothesis of linguistic relativity, arguing that reading processes are similar across different scripts. Jorden (1995) emphasizes the pedagogical importance of introducing reading materials whose spoken forms are already familiar to non-native speakers in order to create not mere decoders, but true readers of Japanese texts. This content downloaded from 147.251.234.79 on Thu, 20 Mar 2014 13:29:01 PM All use subject to JSTOR Terms and Conditions 2 | Volume 30, Number 2 These are powerful interdisciplinary arguments, yet what could still be added is a comprehensive argument against the misleading terms used to characterize kanji that are mentioned above. While such an argument is available for Chinese (DeFrancis, 1984a, 1984b, 1989), it is not for Japanese. The purpose of this paper is, therefore, to reexamine the linguistic nature of kanji, while arguing that not only in Chinese but also in Japanese, kanji represent sounds as well as meanings.1 This reexamination will provide negative answers to four questions in turn: "Are kanji pictographs?" "Are kanji ideographs?" "Are kanji logographs?" and "Are kanji morphographs?" Following DeFrancis (1984a, 1984b, 1989), more appropriate labels for kanji in both Chinese and Japanese writing will then be suggested. Are Kanji Pictographs? Historically, it is true that when kanji or Chinese characters were first developed more than three thousand years ago, many of them were pictographs, drawings of concrete objects (Karlgren, 1923/1946; Gelb, 1963; DeFrancis, 1984b, 1989; Coulmas, 1989) with no indication of their sound values (Liu, 1978). As an example, the evolution of the character for ma (in Chinese) and uma (in Japanese) meaning 'horse' is shown below: Shang Great Seal Small Seal Scribal Regular Simplified % % Si % (adapted from J. DeFrancis, Visible Speech: The Diverse Oneness of Writing Systems [1989:96] with permission from University of Hawaii Press). As can be derived from the above example, the evolution of many of the first pictographs is traceable from their earliest forms to the latest form. This fact does not mean, however, that kanji in general are pictographic symbols with no indication of sounds. In modern Chinese or Japanese writing, only one percent of kanji in Chinese (DeFrancis, 1989) and only 11.7 percent of kanji on the Toyo kanji list in Japanese (Nomura & Ito, 1978) originate from the pictographs of 1200-1045 B.C. The current lack This content downloaded from 147.251.234.79 on Thu, 20 Mar 2014 13:29:01 PM All use subject to JSTOR Terms and Conditions Journal of the Association of Teachers of Japanese | 3 of graphic representation is due to the development and the reform of the two writing systems, both of which are characterized by movements toward the representation of the sounds of the two spoken languages. Such movements are exemplified by: (a) the phoneticization of kanji (e.g., the pictograph representing 'wheat' 5|v lai in the protowriting stage, came to be used to represent the sound of the homophonous word 'come' in the real writing stage [DeFrancis, 1984b: 138]); (b) the creation of phonetic compounds, that is, composing new kanji by combining a radical and a phonetic element (e.g., $)j 'ask' [Ch. fang/Jn. ho] = "gf 'talk' [radical] + ~Jj [Ch.fang/Jn. ho]) (Karlgren, 1923/1946:57); (c) the creation of two types of kana (syllabic signs) and the standardization of the Japanese writing system, using kanji and kana in the modern form of kanji-kana-majiri-bun (texts written in a mixture of kanji and kana); and (d) the adaptation of a colloquial style in both Chinese and Japanese to approximate the written to the spoken languages. Once the two writing systems were developed through these stages, kanji could no longer simply be called "pictographs"; as DeFrancis (1984b) puts it: To see that writing has the form of pictures and to conclude that it is pictographic is correct only in one sense—that of the form, but not the function, of the symbols. We can put it this way: QUESTION: When is a pictograph not a pictograph? ANSWER: When it represents a sound, (p. 140) Are Kanji Ideographs? The word "ideograph" is a popular term used by many scholars to describe kanji, expressing the notion that kanji do not represent sounds, but rather ideas. A typical thought of this kind is found in Creel (1936): The Chinese early abandoned the method of writing by means of readily recognizable pictures and diagrams. ... It was in part because the Chinese gave up pictoral [sic] writing that they were able to develop a practicable pictographic and ideographic script, with comparatively little help from the phonetic principle. . . . The course taken in many parts of the world was to conventionalize the picture, reduce it to a simple and easily executed form, and then use it to represent homophonous words or parts of words. The This content downloaded from 147.251.234.79 on Thu, 20 Mar 2014 13:29:01 PM All use subject to JSTOR Terms and Conditions 4 I Volume 30, Number 2 course the Chinese have chosen has also been to conventionalize and reduce, but they then use the evolved element for the most part not phonetically, but to stand for the original object or to enter with other such elements into combinations of ideographic rather than phonetic value. This parting of the ways is of the most profound importance, (p. 91-93, cited in DeFrancis, 1984b: 141) In the above statement, there seems to be a claim of progress in describing Chinese as no longer a pictographic language. Yet as DeFrancis (1984b) has pointed out, Creel's emphasis on "ideographic rather than phonetic value" in his characterization of kahji makes his statement no different from saying that kanji are drawings or pictures in simplified and conventionalized forms. His use of the word "ideographic" is simply a replacement of the word "pictographic" in form, but not in meaning. Although it is an error to emphasize the semantic over the phonetic value of kanji (DeFrancis, 1984a, 1984b, 1989), Many scholars continue to refer to Chinese writing (Karlgren, 1923/1946; Wang, 1973; Kolers, 1977; Liu, 1978; Li & Thompson, 1982; Smith, 1985, 1988) in a similar way to that of Creel (1936). Li and Thompson (1982), for example, label kanji "logographs," by defining them in a similar manner to the one offered for "ideographs" above. They say: . . . the Chinese writing system is unique among modern writing systems in being semantically, rather than phonologically grounded. That is, in Chinese each 'character' or LOGOGRAPH, represents a semantic or grammatical unit. It does not convey phonological information except in certain composite logographs where the pronunciation of the composite is similar or identical to one of its component logographs. Thus, even in those cases, the phonological information conveyed by the composite logographs is based on other logographs, whose forms provide no clue to their pronunciation. (Li & Thompson, 1982:77) Contrary to Li and Thompson's statement above, the linguistic data provided by DeFrancis (1984b) show that "the phonetic elements have a great deal to do with the sounds of Chinese characters" (p. 108). DeFrancis statistically examined the construction features for kanji, and found that nine-tenths of kanji in Karlgren's Analytic Dictionary of Chinese and Sino-Japanese (1923) contain a phonetic element as well as a semantic element. Moreover, among 500 kanji which he sampled from Chen's This content downloaded from 147.251.234.79 on Thu, 20 Mar 2014 13:29:01 PM All use subject to JSTOR Terms and Conditions Journal of the Association of Teachers of Japanese | 5 (1928) list of 4,719 different characters that were found to occur in a frequency count of almost one million characters of running text, 91(18 percent) were found to be independent phonetics, 394 (79 percent) were phonetic compounds, and only 15 (3 percent) had no phonetic aspect at all. Furthermore, among the 394 phonetic compounds, the phonetic element had much greater importance than the semantic element in helping to identify the linguistic element associated with an individual kanji. In 66 percent of the cases the phonetic element represented specific sounds, while in 52 percent (based on the author's calculation according to DeFrancis' [1984b: 129] data) the semantic element did no more than suggest general categories of meaning, such as liquid, fire, and so on. What these results mean, according to DeFrancis (1984b), is that "a reader with a knowledge of the phonetic component in Chinese writing has two chances out of three of guessing correctly the pronunciation of any given character he is likely to encounter in reading" (p. 108). In other words, for native readers who can derive meanings from sounds, being fluent speakers, the phonetic element plays a more important role than the semantic element in determining the meaning of a kanji, since the radicals provide, for the most part, little more than a very vague hint for the reader to use in order to reach the meaning of the whole character (DeFrancis, 1984b). The notion that kanji are "ideographs" representing only ideas should therefore be rejected in Chinese writing. The ideographic myth is popular in connection not only with Chinese writing but also with Japanese writing. In order to see how, the way in which the Japanese writing system works needs to be described briefly. The two languages, Japanese and Chinese, being completely unrelated to each other, have in common only the fact that they both use kanji in writing, and that many lexical items called kango, or Sino-Japanese words, were borrowed from Chinese into Japanese. This is due to the historical fact that the Japanese not only adopted Chinese characters to write native Japanese words, but also borrowed Chinese lexical items by assigning readings which are approximations of the Chinese pronunciations. There are two ways in which kanji were adopted to write native Japanese words: one is by borrowing kanji with the Chinese sounds irrespective of their meanings (e.g., tilya + ffi ma = jQffi yama 'mountain' [Seeley, 1991:190]); the other is by assigning the Japanese sounds to the meanings of borrowed kanji (e.g., yama (or [JL| 'mountain'). The former This content downloaded from 147.251.234.79 on Thu, 20 Mar 2014 13:29:01 PM All use subject to JSTOR Terms and Conditions 6 | Volume 30, Number 2 syllabic usage of kanji is called man' yogana and was later simplified to two types of kana, the modern Japanese syllabic signs, namely hiragana and katakana: hiragana are used mainly for grammatical inflections and particles; katakana are used for Western loan words and for special emphasis, which is especially likely in onomatopoeia. The latter meaning-based usage of kanji is called kun. As mentioned earlier, however, Sino-Japanese words are read not in Japanese readings {kun) but in Chinese readings. This usage of Chinese readings of kanji is called on. Thus, in cw-reading, the character (_Lf meaning 'mountain' or 'Mt.' is read as san, and in £w/z-reading, it is read asyama, as illustrated in the following two sentences, with underlines indicating the parts written in kanji: [ll \Z.*st%)0 Yama ni noboru. '[F 11] climb a mountain.' 9<|i-[ fC^-2>o Fuji-san ni noboru. '[F 11] climb Mt. Fuji.' When kanji are labeled as "ideographs" in Japanese writing, the following arguments are typically made (e.g., Kato, 1989), many of which are shared by Morioka (1968), Sakamoto and Makita (1973), Suzuki (1975, 1975/1982), Iwata (1983), Backhouse (1984), and Shibatani (1990): 1. One can get meanings of words written in kanji immediately and directly. 2. Kanji differentiate homophonous words. 3. Kanji indicate subtle differences in meanings of words. 4. One can correctly guess the meanings of words written in kanji upon encountering them for the first time. 5. Kanji have a function of creating new words. 6. Kanji being mixed with kana in Japanese writing provides visual signals for the readers to pick up meanings from kanji, and thus makes skimming of a text easier. (Kato, 1989:9-17, summarized and translated by the author) As for argument 1, counter evidence is abundant. For example, Perfetti and Zhang (1991) and Cheng (1992) found that subvocalization occurs, if not prior to, then simultaneously with meanings in Chinese at the word level. Tzeng, Hung and Wang (1977) and Horodeck (1987) provided strong evidence for subvocalization at the sentence level in Chinese and Japanese, respectively. Furthermore, Matsunaga (1995) confirmed This content downloaded from 147.251.234.79 on Thu, 20 Mar 2014 13:29:01 PM All use subject to JSTOR Terms and Conditions Journal of the Association of Teachers of Japanese | 7 Horodeck's result using an eye-tracking methodology at the discourse level in Japanese. What these results indicate is that fluent readers do not get at meanings of words without using the sounds of kanji under normal conditions (i.e., reading Japanese and Chinese for comprehension). Although one may argue that this psycholinguistic evidence does not prove that one cannot get at meanings of words immediately and directly, the evidence in support of the direct access of word meanings (i.e., lack of subvocalization) is weak, and limited to the identification of familiar single-character words (Hatta, 1978; Seidenberg, 1985).2 Normal Chinese and Japanese texts do not consist solely of familiar single-character words. Argument 1 stated above thus lacks empirical support. As for argument 2, it is true that kanji can differentiate homophones, as seen in: (a) native-Japanese words, $L 'look (at)' and 'examine,' both of which are read as mini (Kato, 1989:9); and (b) in Sino-Japanese words, ^Et shibo, shikyo, toshi and 'Wc sh oshi, meaning 'die,' 'pass away,' 'be frozen to death,' and 'be burned to death' respectively (Iwata, 1983:184). As for the native-Japanese examples, again, while no one can deny that kanji differentiate the meanings of these homophones, contexts can do the same job as kanji in normal writing. Thus, the latter example of miru W>%), which appears only in the context of medical treatment or examination, need not be written in kanji, so long as the meaning of the verb is determined by the context. As for the Sino-Japanese examples, Iwata (1983) argues that when compound words consist of two characters XY and XZ, or YX and ZX, the difference in meaning is determined by the difference between Y and Z. This rule indeed applies to the two words, $0fc and 'M^E- However, as pointed out by Nomura (1988), in the words ^E"t and ^Ei, the difference in their meanings cannot be easily determined by the two characters t 'disappear' and ^ 'leave.' Thus, argument 3 cannot be said to be valid as strongly as it is claimed to be, not only in native-Japanese words but also in Sino-Japanese words. After all, "it is not the kanji themselves that are doing the work of differentiation. The associated morphemes, written in kanji or not, are doing the work" (T. J. Vance, personal communication, February, 1994). Argument 4 is the so-called "semantic transparency" of kanji (Suzuki, 1975, 1975/1982; Shibatani, 1990). In order for one to understand this notion of "semantic transparency," it is necessary to recall the fact that in Japanese, there are many cases in which both Chinese readings (on) and Japanese readings (kun) are assigned for the same kanji. The character 7fC, for example, has the