electronic journal of contemporary japanese studies

Article 5 in 2005
First published in ejcjs on 19 August 2005

Search the Web Search ejcjs

How to contribute to ejcjs

Change in Script Usage in Japanese

A Longitudinal Study of Japanese Government White Papers on Labor


Takako Tomoda

School of Languages, Cultures and Linguistics

Monash University

e-mail the Author

About the Author


In the post-war period, official language reforms combined with social and technological change have changed the Japanese writing system. Notable amongst the reforms were officially sponsored limits on the number of kanji. Technological change has been accompanied by an influx of loanwords written in katakana. Both these changes have been subjects of considerable dispute. Since the early 1990s katakana words have received official language planning attention and particular attention has been given to katakana use in government White Papers. This study examines longitudinal change in the proportions of the three Japanese scripts of kanji, hiragana and katakana using samples of White Papers on labor published between 1960 and 1997 which were controlled for subject matter to enable valid longitudinal evaluation. Two counts were conducted and the same trends were found for both counts when variation due to format was controlled for. Contrary to expectations, the proportion of kanji increased while hiragana decreased. The proportion of katakana increased, due mainly to an increased use of loanwords, but the overall level was less than expected. Comparisons were made with other Japanese studies and the effects of language reforms discussed.


The appropriate use of the various kinds of script has long been a topic of scholarly and popular debate in Japan. Official language policies have also placed considerable emphasis on script issues. Whereas in the past attention tended to focus on issues surrounding kanji use, in recent decades the use of katakana has received more attention. It has been claimed that katakana words (many of them loans from English) are overused and their use in government documents has been the target of both criticism and recent official action (Kokugo Shingikai, 2001). Statistical data from both opinion polls and studies of text are frequently quoted and play an important role in guiding decision makers. Textual data does, however, vary according to genre and methodological differences in data collection and processing can make comparisons between studies problematic. Moreover, longitudinal studies are few in number but these are essential if trends are to be revealed and the results of past interventions assessed.

By examining the changes in script usage over a series of samples from government White Papers this study aims to determine which scripts have declined or increased and to examine the role of language policies and other factors that influence the changes found.

The Japanese scripts and how they are used

The traditional Japanese writing system comprises three distinct scripts: the logographic script adapted from Chinese called kanji (漢字), and the phonological scripts of hiragana (ひらがな) and katakana (カタカナ). In addition, repeat symbols, referred to as odoriji, and a range of punctuation marks are used. However, the modern Japanese writing system has also adopted Arabic and Roman numerals, the Roman alphabet (called rōmaji) and additional symbols such as dashes and % symbols. In general, kanji are used for content words of native Japanese origin, referred to as wago, or for various words of Chinese origin which are referred to as kango. Hiragana are used for grammatical features as well as for some content words, most of which are of native Japanese origin, i.e. wago. Numerous words (particularly verbs and adjectives) are written with a combination of these two scripts − kanji for the stem and hiragana for the inflection (e.g. 行きます). The hiragana used in this manner are referred to as okurigana. Katakana have a diversity of uses including content words of foreign origin, as a replacement for writing a word in kanji, for a wide range of proper nouns, for slang terms and onomatopoeia. Alphabetic characters are mostly employed in acronyms of foreign or domestic origin, as well as in numbering sections in a document. Numbers can be written in kanji or Arabic numerals depending upon the type of text. The general rule is that kanji numerals are used when texts are in vertical format and Arabic numerals in horizontal format. Roman numerals are mostly used as section numbering devices.

Reforms to the Japanese writing system

The type and manner of usage of kanji, hiragana, katakana and rōmaji have been the topic of discussion, dispute, agitation and official action since the latter part of the nineteenth century (see Miyajima, 1977; Twine, 1991). Over the post-war period changes in the Japanese writing system have included officially sponsored language planning measures primarily aimed at script usage as well as a range of unplanned changes such as the influx of new loanwords. A comprehensive discussion of post-war language policy and language change is beyond the scope of this article, so the following outline is limited to issues that are likely to have had an impact upon script proportions.

In November 1946, a series of official reforms in the writing system began. The stated aims of these reforms were to simplify the use of hiragana and katakana, reduce the number of kanji in everyday use, and standardize writing styles (see Seeley, 1991; Tsuchimochi, 1993; and Unger, 1996 for details). The first wave of reforms was phased in over a period of more than ten years, however both the reforms and the reform process were subjected to considerable criticism from some quarters. During the 1960s a number of the reforms and reform proposals were re-examined and during the 1970s and 1980s the reforms were rolled back somewhat. The limits on kanji were relaxed and the rules for okurigana use changed in favor of less usage (see Shiota, 1973; and Gottlieb, 1994; 1995 for details). Nevertheless, the present structure of the Japanese writing system remains largely a product of the first series of post-war language reforms.

The post-war era saw an influx of loanwords into the Japanese language. The majority of these loanwords, referred to as gairaigo, are derived from English with smaller proportions deriving from other, mostly European, languages. Loanwords have long been a feature of Japanese and the Japanese language has undergone a number of periods of massive borrowing, mostly from Chinese (see Miller, 1967; Seeley, 1991). In the past, loanwords were generally written in kanji but over the last century the general convention has been to use katakana for new loanwords, particularly for those of non-Chinese origin. This practice was formalized in the immediate post-war period (Habein, 1984).

Whereas in the pre-war era creating calques was a major means of introducing new words into the language, post-war the direct borrowing of words of Euro-American origin has tended to take precedence. Calques generally employ kanji to express the meaning of the new term. For example, 'airport' is calqued as kūkō 空港, which comprises one kanji for 'air' and one for 'port' In contrast, direct loans approximate the pronunciation of the source language and are written in katakana, for example 'modem' becomes modemu モデム. One consequence of this is a relative increase in the proportion of katakana gairaigo and a decline in kango (Sugito, 1989).

The reasons for this shift away from calquing towards direct borrowing have been the subject of considerable dispute but a discussion of these issues is beyond the scope of this article. In brief, it is variously attributed to the post-war limits on kanji (see Maruya, 1978; Suzuki, 1990), the rapid pace of technological and social change (see Ishino, 1983; Iwabuchi, 1993), the spread of English language education (see Honna, 1995; Suzuki, 2001), adulation of Western things (see Suzuki, 1985; Ōno, Morimoto & Suzuki, 2001) as well as other reasons. The phenomenon of an increase in loanwords written in katakana is, however, one that few would dispute − although the degree of increase can vary considerably with the language domain (see Tomoda, 1999). This increase in katakana words has received widespread comment and there have been calls for limits to be placed on the use of these words (Suzuki, 1985; Kajiki, 1996). In the early 1990s, katakana and gairaigo emerged as an issue for official language policy and in the early 2000s steps were taken to limit the use of new katakana words in government documents (Bunkachō, 1994; Tanaka, 2003).

In terms of script proportions, as the number of direct loans in the written language rises, so should the proportion of katakana. Besides adding new vocabulary items to the language, most of which are nouns, new loanwords can also replace existing nouns. For example, depāto デパート (a contraction of  'department store' has generally replaced the kanji calque hyakkaten 百貨店. Since the new loanwords are written in katakana while established nouns tend to be written in kanji, as new loanwords replace existing words, the relative proportion of kanji could be expected to decline.

One gauge of the increase in loanwords written in katakana is the proportion included in dictionaries. This has increased from 1.4% to about 10% of total entries over the last century (Tomoda, 1999). There is, however, considerable variation in the proportion of loanwords written in katakana used in the various Japanese print media. Advertisements tend to contain high proportions of loanwords. Takashi (1990) reported that 22.3% of word tokens in a sample of print advertisements were classed as gairaigo. However, literary works generally contain much lower proportions. An analysis of the monthly magazine Chūō Kōron of 1976 found that 3.7% of words were classed as gairaigo (Miyajima, 1989). The large proportion of katakana in advertising is not only due to the use of gairaigo. Katakana tend to stand out and are consequently the preferred script for use on signs. Ordinary Japanese words can be written in katakana in order to highlight them and product or business names are often written in katakana even when they have no links to foreign words. As a result, katakana can be the dominant script in an advertisement, whilst comprising the smallest proportion of the three main scripts in a sample of general prose.

As a result of the language reforms which limited the number of kanji in general use, the influx of loanwords written in katakana, and the shift away from calquing as the preferred method of creating new words, the following changes in the proportion of scripts could be expected.

  1. With the increase in gairaigo, the proportion of katakana script used should increase. The increase in katakana characters is likely to be greater than the corresponding increase in loanwords since loanwords of European origin tend to be longer than words of Japanese or Chinese origin.

  2. The effect of limiting the use of kanji is likely to result in a decrease in the proportion of kanji. Words that were previously written using those kanji that were affected by the reforms could be expected to change to being written in hiragana or katakana. Alternatively, they could be replaced by other words written in kanji or by loanwords written in katakana.

  3. Since both kanji and katakana are primarily used for writing content words, the tendency to supplement the lexicon with items borrowed directly from English, rather than create calques using the resources of kanji, could be expected to result in an increase in the relative proportion of words written in katakana compared with kanji words.

The above effects would all be expected to result in a reduction in the proportion of kanji and an increase in the proportion of katakana. However, they are by no means the only processes in action. With the advent of word-processors that can deal effectively with Japanese script, the use of kanji has become more convenient and this may have the effect of increasing the proportion of kanji in use (Gottlieb, 1993). Usually this is at the expense of hiragana. For example, the word 'adjacent' can be written in hiragana as となり or in kanji as 隣. Of the two options the hiragana is easier to write by hand compared to more complex kanji but on a computer the kanji is as easy to produce and also takes up less space (see Gottlieb, 1993 & 1998 for discussions).

Considering the variety of processes that could be operating to generate change in written Japanese over various language domains, the need for quantitative longitudinal data from a variety of language domains is apparent. In this study the domain chosen for examination was government White Papers on labor. This domain is a formal one that should reflect official policies on usage, but it is also one that deals with current subject matter and is therefore influenced by socio-economic change. Since the White Papers are issued annually with similar content this domain can provide samples of text that allow valid longitudinal comparison.

The proportions of the three Japanese scripts of kanji, hiragana and katakana and the Roman alphabet in White Papers published between 1960 and 1997 were calculated. These were compared with the results of other studies with particular reference to any change in the proportion of katakana and changes in the level of loanword usage.

Existing estimates of the proportions of scripts in Japanese

A number of estimates of the proportions of scripts used in Japanese have been reported in the literature. These estimates were derived from a range of genres and used quite different sampling methods, so direct comparison between estimates can be problematic. In none of the studies examined was there an attempt to control for subject matter. Since katakana feature prominently in foreign names and in vocabulary relating to new technologies while kanji are prominent in literary genres, the proportions of scripts can vary considerably with the subject matter of a text. Therefore, control for subject matter and genre is a necessary feature of longitudinal studies.

The major surveys of newspapers and of ninety magazines conducted by the National Language Research Institute (NLRI) did not report data for script proportions (NLRI, 1964; 1970). However, the following studies did address this issue. The Kyōdō news agency estimated the proportions of scripts used in newspapers on the basis of a sample of 4,252 sentences selected from newspapers published between July 15 and July 21 of 1971. All characters were counted including symbols and numerals. The following proportions were obtained: kanji 46.5%; hiragana 35.3%; katakana 6.6%; alphabet 0.4%; numerals 1.4%; and symbols 9.8% (see Table A2). When only the four main scripts were considered, the results were: kanji 52.3%, hiragana 39.7%, katakana 7.5% and alphabet 0.5% (Hayashi, 1982).

Nomura (1980) surveyed 27 magazines aimed at general readers published in July 1979 in order to examine the proportion of kanji. The magazines were divided into six classes according to topic area, type of publisher, and readership as follows: seijikeizai-kei 'politics and economy', shimbun-kei 'news magazines', shuppan-kei 'magazines produced by general publishers', taishū-kei 'tabloid magazines', josei-kei 'women's magazines', and dansei-kei 'men's magazines', Nomura sampled 100 sentences per magazine excluding titles, poems, short stories, advertisements, names, tables and other non-prose sections. All classes of characters were counted including punctuation. He found that magazines within a class contained similar proportions of kanji. Magazines in the area of politics and economics had the highest average proportion of kanji at 38.2%, while the lowest proportions were found in women's and men's magazines, at 25.2% and 23.6% respectively. There was a direct inverse relationship between the proportions of kanji and katakana. The average proportion of katakana in magazines concerned with politics and economics was only 6.5% but this increased to 10.4% in women's magazines and 15.4% in men's magazines.

Longitudinal studies are few. Yasumoto (1963) sampled 1,000 characters from each of 100 novels, all by different authors, that were published between 1900 and 1954. He found the proportion of kanji progressively declined while the proportion of hiragana and katakana rose. If this tend continued at the same rate, he predicted that in around 2190 kanji would cease to be used in novels and they would be wholly written in phonological characters.

The monthly magazine Chūō Kōron was sampled at ten-year intervals between 1906 and 1976 to examine changes in the Japanese written language. This sampling involved the random selection of pages to produce a one tenth sample. When the four main scripts were examined for the post-war period the following trends were evident (see Table A1). Kanji declined between 1946 (41.2%) and 1956 (36.2%) but then increased again to 38.0% in 1976. Hiragana increased from 1946 (56.0%) to 1956 (58.2%) but then declined to 55.0% in 1976. Both the katakana and alphabet proportions fluctuated but showed overall increases. Katakana increased from 2.6% in 1946 to 6.7% in 1976 and alphabet from 0.2% to 0.3% over the same period (NLRI, 1987). The authors of this study noted that when the results of the character counts were compared to the results of the word counts over the whole period 1906 to 1976 the following trends were evident. The increase in katakana was due to an increase in gairaigo and the replacement of kanji with katakana. For example, in a number of cases the names of foreign countries and places that were written in kanji in 1916 had come to be written in katakana by 1926. Overall, hiragana tended to increase but the use of native Japanese words, wago, decreased. Following the post-war language reforms, parts of wago that used to be written in kanji came to be written in hiragana. With regard to kanji, the change paralleled that of kango before 1945 but after this the relationship was lost. This may have been due, however, to the subject matter contained in the 1976 Chūō Kōron. The proportion of kanji may have risen in this year since the magazines used were concerned with stories from the Meiji era (NLRI, 1987).

Loanwords in government documents

Government documents have generally been considered a domain in which language use is conservative. Nevertheless, the use of loanwords and other new terms written in katakana appears to have increased in recent decades. Japanese government departments have received criticism for their excessive use of loanwords written in katakana (see Mogami, 1984, 1991; Suzuki, 1990; Loveday, 1996). A survey conducted by NHK (Japan Broadcasting) found that out of a sample of 11,835 names of local government projects 25.2% contained gairaigo (Mogami, 1984). Suzuki (1983) argued that government departments should use words that people can easily understand but that recently introduced gairaigo were unclear and caused confusion. Ekuni (1993) said that he could accept the use of many katakana words in advertising but not the high level of use of gairaigo in government. Government use of katakana words has been blamed for their increase in the news media. When such words are used in reports, the media has no choice but report them and this introduces these words into the language of ordinary people (Mogami, 1984; Sekine, 2003).

The Asahi daily newspaper reported that in 1988 a glossary of katakana terms, which had been compiled for internal use in the planning department of the Tokyo Metropolitan government, had become unexpectedly popular. The reason given for this was the difficulties being experienced by bureaucrats and politicians in understanding written texts due to the increase in new katakana terms (Asahi, 1988).

In 1989 the Minister for Health and Welfare (later to become prime minister) Koizumi Junichirō criticised the use of katakana words and said proper Japanese words that older people could understand should be used in aged care. He set up a Terminology Rectification Committee to replace katakana words in publications produced by offices under control of his ministry (Kajiki, 1996). At the national level, in 2002 the Foreign Loan Words Committee (Gairaigo Iinkai) was established. It was charged with producing lists of replacement terms for katakana gairaigo that had been used in White Papers (Tanaka, 2003).

On the basis of the above examples, it appears that there has been a considerable increase in the use of katakana and gairaigo in government. Nevertheless, it is not clear whether there has been a general increase in these words in government publications, or whether this increase has been limited to specific high-profile areas such as project names.

Aims of this study

This study examines longitudinal change in the use of scripts in the formal domain of government White Papers using samples that are directly comparable in both subject matter and style. The primary aim is to determine whether there has been an increase in katakana and decline in kanji and gauge its extent. Government White Papers on labor, called rōdō hakusho, were chosen for the following reasons:

  1. Rōdō hakusho are issued annually and cover very similar lists of topics, and therefore are a text medium in which the subject matter is relatively stable over time.

  2. Since they refer specifically to events and socioeconomic conditions of the previous year, they are unlikely to include extended passages that reflect the writing styles of the more distant past.

  3. Labor conditions is a field in which technological change would not be expected to have resulted in a large influx of new terms associated with new technologies.

  4. Unlike newspaper or magazine articles, White Papers on labor contain very few personal, place or product names.

  5. Being government documents they should conform to public service guidelines for writing and consequently variations in style from year to year and the use of faddish terms could be expected to be minimal.

  6. Finally, rōdō hakusho could be expected to be free from katakana words associated with the reporting of foreign news or advertising.

For the above reasons, White Papers on labor provide a domain in which the effects of subject matter, literary style and technological change are controlled for.

The secondary aims were: 1. to examine the data published by previous researchers and reorganise this data to allow meaningful comparison between these studies and the present study; and 2: to determine the extent to which any changes were attributable to language policies and to discernable stylistic factors.


On the basis of the findings in the existing literature three hypotheses were proposed:

  1. The overall proportion of kanji would decline over time.

  2. The proportion of katakana would increase.

  3. The proportion of hiragana would increase.

Method and materials

Four rōdō hakusho were selected on the basis of availability and similarity in content. The same section was examined in each. These sections dealt with the living conditions of workers. The selections examined were as follows:

  1. 1960 (Showa 35) Rōdōsha no seikatsu jōtai 'the living conditions of workers' pp. 201-213;

  2. 1976 (Showa 51) Shōhisha bukka to kinrōsha kakei no dōkō 'trends in consumer prices and the household budgets of workers' pp. 22-31;

  3. 1986 (Showa 61) Bukka, kinrōsha kakei no dōkō 'trends in prices and the household budgets of workers' pp. 67-82; and

  4. 1997 (Heisei 9) Bukka, kinrōsha kakei no dōkō 'trends in prices and the household budgets of workers' pp. 68-83.

The White Papers of 1976, 1886 and 1997 were directly comparable in terms of content and sub-headings whereas the organization of the 1960 White Paper was rather different. Consequently, the most comparable section, in terms of content, of the 1960 White Paper was selected.

For each White Paper two counts were done. These used the same criteria for judgment. In the first count all characters in the sections specified above were classified according to the type of script in which they were written and the totals recorded page by page. The aim of this count was to control for variability between the samples due to differences in subject matter. Since there was considerable variation from year to year in the length of the sections, a second count was done on samples of equal length. The aim of this count was to control for sample length. In some years more detail was included in the prose, while in others it was summarized in tables and graphs. In both counts, graphs and tables were excluded as were their titles and other associated script. All characters in the prose sections were counted by the same scorer. Since the principal aim was to examine changes in the language over time, the use of the same scorer controlled for any variation due to subjective decisions about the status of a particular item. There was, however, very little scope for subjectivity since the decision making criteria were determined beforehand.

The focus of each count was upon the status of each written character. Scores were determined for the total numbers of kanji, hiragana, katakana, Arabic numerals, Roman numerals, % symbols, letters of the alphabet, and ~ symbols. Sentence endings and other punctuation marks were not scored. Katakana bars (ー) were counted as one katakana, repeat marks were counted as kanji or hiragana depending upon the case, and numerals written in kanji (including zero) were counted as kanji. In addition the number of numerals written in kanji was scored separately. This sub-score did not include kanji numerals that were parts of words, such as in ippō 'on one hand' but included numerical terms such as futari 'two people'.

The frequencies of each kind of script were entered into Statistical Package for the Social Sciences (SPSS). Chi-square tests were conducted to determine whether any changes in the proportions of the different scripts from sample to sample were significantly different from that expected as a result of chance variation.

Results and discussion

Count 1: Variable sample length

The results reported in this section derive from samples of White Papers on labor, which have been controlled for topic matter. The total number of characters sampled in each edition varied greatly. The same section of the 1986 White Paper contained almost twice the number of characters as the 1976 one. The total numbers and proportions of each of the scripts and symbols counted are presented in Table 1.

Table 1: Frequencies and proportions of each character type in White Paper samples (Count 1).

The large difference in the frequency of Arabic numerals included in the 1960 White Paper and in the papers from 1976 onward was mainly due a change in the style of printing between these two dates from vertical to horizontal. This meant that many numbers that were written in kanji in 1960 were written in Arabic numerals in 1976. The following categories of numerals were written using kanji in 1960 and using Arabic numerals in 1976: section numbers, table numbers, years, months, and percentages. In the 1960 White Paper, sub-sections and lists of points were numbered using Arabic numerals or katakana, whereas only Arabic numerals were used in the 1976 White Paper. Roman numerals were used in both the vertical and horizontal formats but nearly always as components of tables. None appeared in the sections of connected text that were counted. Quantities were always written in kanji in the 1960 selection but in the 1976 White Paper all quantities were written in Arabic numerals. In the 1986 White Paper some small quantities, for example futari, tended to be written in kanji, while larger ones always used Arabic numerals. The effect of these changes in the way numerals were written was an inflation of the overall number of kanji in the 1960 White Paper relative to the later ones. Due to the effect on the overall proportion of kanji, numerical kanji were scored separately.

In the 1960 White Paper 6,151 characters where counted but this included 52 symbols and 11 Arabic numerals. No Roman numerals or alphabetic characters were used. The three Japanese scripts accounted for 98.98% of characters. 52.87% were kanji, 45.65% hiragana and only 0.46% were katakana. Of the kanji, 333 were numerals, so when these were subtracted the proportion of kanji fell to 47.46% of total characters. Of the 28 katakana found in the sample, 6 were section markers while the rest were used in writing words (see Table 1).

The 1976 White Paper contained 4,828 characters of which 92 (1.90%) were symbols. 11.54% of characters were Arabic numerals, there were no kanji numerals and there were no alphabetic characters or Roman numerals. The three Japanese scripts accounted for 86.56% of all characters: 44.88% kanji, 39.15% hiragana and 2.53% katakana (all in words).

The 1986 sample was the largest with 8,362 characters of which 7,504 (89.74%) were in Japanese script in the following proportions: 48.54% kanji, 38.65% hiragana and 2.55% katakana. Symbols comprised 1.71% and Arabic numerals comprised 8.51%. Three letters of the alphabet appeared as the acronym VTR.

In the 1997 sample, Japanese characters comprised 90.21% of the 6,229 characters sampled, with kanji accounting for 50.36%, hiragana for 36.36% and katakana for 3.48%. There were 50 alphabetic characters (0.8%) and Arabic numerals accounted for 7.16%.

The change from vertical to horizontal printing style between 1960 and 1976 was associated with a decline in the proportion of Japanese script used. However, this proportion then increased through to 1997 even though it did not reach the 1960 level. The change from writing numerals in kanji to Arabic form accounted for much of this change between 1960 and 1976. The overall proportion of hiragana declined throughout the period while that of katakana increased. However, the level of katakana increase was less than the hiragana decrease so the relative proportion of kana declined overall. The use of letters of the alphabet became evident in the 1986 sample and had increased considerably in the 1997 sample. In both samples the items in which alphabetic characters appeared were acronyms.

Since the principal aim of this study is to examine longitudinal change in the three Japanese scripts, three further calculations were done.

  1. The three Japanese scripts plus alphabet but excluding Arabic numerals and symbols were totaled.

  2. Proportions were calculated for each of the Japanese scripts and alphabet as percentages of the total number of characters written in Japanese and alphabetic script, excluding Arabic numerals and symbols.

  3. To control for the change in numeral use between 1960 and 1976, kanji numerals were subtracted from the total kanji scores to give a non-numerical kanji score. This allowed the kanji proportions to be more validly compared as well as controlled for variations in the level of reporting of figures in text compared to in the tables (see Table 2 below).

Table 2: Frequencies and proportions of non-numerical Japanese and alphabetic characters in White Paper samples.

When the proportions of Japanese scripts and alphabet were adjusted for numeral and symbol use, a steady rise in the kanji portion over the years 1960 to 1997 of 4.57% was found. There was a corresponding decline in the hiragana portion over the same period of 8.79%. These trends were not simply associated with the change in format since both were evident from 1976 to 1997. The katakana portion rose from 0.49% percent to 3.83% between 1960 and 1997, a rise of 3.34%. In addition, the use of letters of the alphabet increased from zero in 1976 to 0.88% in 1997. Since both katakana and alphabetic characters were used principally for writing loanwords, these two scores can be combined to give an indication of the proportion of script taken up in rendering loanwords. The katakana score in 1960 included katakana used in numbering so only 0.38% (22) were used in writing katakana words, whereas 4.71% of katakana and alphabetic characters were used in words in 1997, a rise of 4.33%.

Due to the small number of alphabetic letters, these were excluded from statistical analysis. The Chi-square test for independence was conducted on the three Japanese scripts only. Taking all four samples of text into account the Chi-square statistic was highly significant (χ2 = 205.9, df 6, p < 0.001). The standardized residuals indicated that the increase in katakana from 1960 to 1997 was the largest contributor to this result, so a further test was undertaken on kanji and hiragana alone. Again the result was highly significant (χ2 = 61.4, df 3, p < 0.001) and an inverse relationship was evident in the standardized residuals. Figure 1 is a graphic representation of the change in the proportions of script types.

Figure 1.

Count 2: Equal sample size

The effect of using samples that were controlled for both topic matter and number of characters was examined by equalizing the size of the samples. Since the 1976 sample was the smallest at 4,828 characters, scores were calculated for the same total number of characters in each of the other samples. This was done by counting the first 4,828 characters in each of the corresponding sections of the other White Papers. The resultant samples contained equal numbers of characters and covered similar topic areas. When all characters are considered, the effect of equal sample size compared to count 1 on overall kanji proportions was very small with the same trend being evident. In the 1986 sample the proportion of hiragana was reduced by 2.9% and the proportion of katakana rose by 0.6%. When compared to count 1 the proportions of numerals and symbols were affected in all of the three White Papers. However the variations did not exceed 1.5%. Since numerals tended to be unevenly distributed in the texts some variation was expected. When the numerals are removed and the main three Japanese scripts plus the alphabetic characters are considered separately for count 2, the sample sizes lose their equality and range from 4,179 to 4,773 characters. The resulting totals and proportions for count 2 are presented in Table 3.

Table 3: Frequencies and proportions of non-numerical Japanese and alphabetic characters in White Paper samples (Count 2).

For the White Papers of 1960 and 1997 there was little difference in the proportions of the four main scripts between the two counts. Both of these samples were of roughly equal size in count 1, so they were each reduced by about a quarter for the second count. In contrast, the sample of the 1986 paper used in count 1 was almost twice the size of that in count 2. This resulted in greater variation between the two counts particularly in the proportions of hiragana and katakana. The shorter sample (count 2) resulted in increased proportions of kanji and katakana and reduced proportions of hiragana and alphabet. Nevertheless, the trends which were evident when equal subject matter was the criterion (count 1) were still evident when equal sample size was the criterion (count 2), albeit less obviously. The proportions of kanji, katakana and alphabet increased from 1960 to 1997 while the proportion of hiragana decreased.

When the three Japanese scripts were considered the variation across the samples was found to be significant (χ2 = 195.4, df 6, p < 0.001) with katakana showing the greatest range in the standardized residuals. For kanji and hiragana alone a significant effect was again found (χ2 = 56.4, df 3, p < 0.001) as was an inverse relationship. Figure 2 shows the script proportions for count 2.

Figure 2.

The statistics for the two counts demonstrate the same trends and indicate that the changes found in script types over the samples are highly unlikely to be the result of chance.

Summary of results

There was an increase in the non-numerical kanji proportion from sample to sample. However, this increase was relatively small (4.57% in count 1), indicating that kanji use was fairly stable between 1960 and 1997 in White Papers on labor. This result was unexpected, since it was predicted that the proportion of kanji would decline as the proportion of katakana rose. Instead the hiragana portion declined by 8.79%. As expected, katakana use increased but only to 3.83%.

Comparisons with other studies

In order to enable meaningful comparisons between the present study and other studies, published data from earlier Japanese studies was re-systematized and adjusted to the counting system used in this study (see appendix for complete figures). When compared to the Chūō Kōron sample of the same year, a much lower proportion of katakana was found in the 1976 White Paper. Also, the kanji proportion of the White Paper was far greater even though the vertical format of Chūō Kōron would have placed numerals into the kanji portion, while these would not have been included as kanji in the horizontally formatted White Paper (see Table 4 and Table A1 below). It should be noted, however, that the Chūō Kōron study was primarily concerned with word types and not script proportions. By their own admission, the authors of the study indicated that the script proportion data may have suffered from inadequacies. In addition, Chūō Kōron and the White Paper on labor are very different kinds of documents. Therefore, the most that can be drawn from this comparison is that the kanji proportion can vary by 14% between text types published in the same year. Compared to the newspaper sample of 1971, the proportion of kanji in the 1976 White Paper was only slightly lower despite the vertical format of the newspapers (see Table 4). The difference in the hiragana proportions appeared to be mostly accounted for by the higher level of katakana in the newspaper sample. However, due to the absence of separate data for numerals in the newspaper sample such a conclusion can only be tentative (see Table A2).

Table 4: Comparison of studies for 1970s samples.

Of the previous studies of script proportions in Japanese, that of Nomura (1980) would seem the most comparable to this study. Of the categories of magazines examined in his study, those concerned with politics and economics (seijikeizai-kei) could be expected to be the most similar in content and style to White Papers on labor (see Table A3). The grouped data placed these magazines in an intermediate position between the 1976 White Paper and Chūō Kōron with regard to proportions of kanji and hiragana. Since Nomura has published detailed data, it is possible to calculate script proportions for individual samples using the same method as used in this study, thereby allowing more meaningful comparison (see Table 4). Since the seijikeizai magazines used both horizontal and vertical formats, it was necessary to separate the data for this category. Only one of the four magazines in this category was written in horizontal format, Shūkan Daiyamondo (Weekly Diamond). However, its 11.4% of katakana and 38.7% of kanji indicated that it was very different to the 1976 White Paper and to subsequent White Papers. Both Ekonomisuto (Economist) and Tōyōkeizai (Eastern Economics) were written in vertical format and covered similar subject matter. Their kanji proportions were very similar, particularly when non-numerical kanji were considered, but remained well below those of the White Papers (see Table 5).

Table 5: Seijikeizai-kei magazines 1979 adjusted to 4 main scripts.

Since Nomura's data allowed investigation of the variation in script proportions between samples from different magazines published in the same year, further investigation was undertaken to determine the range of script proportions and facilitate comparison between studies. As data was available for Chūō Kōron in 1976, a comparable magazine was sought from among Nomura's samples. Shūkan Shinchō (Shinchō Weekly) was selected and found to contain very similar proportions of kanji and hiragana to Chūō Kōron and somewhat more katakana (see Table 6).

Table 6: Shūkan Shinchō 1979 and Chūō Kōron 1976 adjusted to 4 main scripts.

At the lower end of the kanji spectrum, four magazine samples from the Men's and Women's magazine categories were selected for further analysis (see Table 7). All were written in the vertical format but only one, Shūkan Josei (Weekly Woman), tended to use kanji numerals. The magazine with the lowest proportion of kanji (23.8%) had the highest proportion of katakana (22.4%), while that with the highest proportion of kanji (31.1%) had the lowest proportion of katakana (11.9%). This inverse relationship was evident across the four magazines. The highest proportion of alphabetic characters was also found in this group at 1.6%.

Table 7: Men's and women's magazines 1979 adjusted to 4 main scripts.

The effects of language policy on script usage

The reasons behind the changes found in the White Papers are likely to lie in a combination of official language policies, departmental guidelines on writing style, the stylistic preferences of the writers and editors, and the advent of wide-spread computer use. Of these factors, official policy and the associated guidelines for government publications are the most amenable to analysis.

The kanji list issued by the National Language Council (Kokugo Shingikai) in 1946 (Tōyōkanjihyō) limited the kanji used in government documents to 1,850. In the preface, it also recommended that certain types of words be wholly written in hiragana rather than kanji. These included: auxiliary verbs (jodōshi), adverbs (fukushi), conjunctions (setsuzokushi). In 1949 the Council issued reforms to be applied to government documents that recommended they be written in horizontal style. A style guide for government documents (Kōyōbunsakusei no yōryō) was issued in 1952 that specified the use of plain, easy to understand language including the replacement of difficult kango, mainly by wago. Further reforms in 1959, increased the use of okurigana, for example, verbs such as owaru (finish) were written with two hiragana following the kanji rather than one (i.e. 断わる instead of 断る) but recommended adverbs be written in kanji (Shiraishi, 1960).

The 1960 White Paper did not completely conform to these policies and guidelines. It was still written in vertical style (this changed to horizontal in the 1961 White Paper) and used a few kanji outside the list, for example shojō (爼上). With regard to hiragana use, it went beyond the guidelines. For example, words such as izen (still), mottomo (most) and tsugi (next) were all in hiragana as were a number of the verbs that had been the subject of the 1959 okurigana reforms (e.g. okonau, kuraberu). This preference for hiragana was, however, not always consistent, for example onaji (same) was sometimes written おなじ and other times as 同じ.

In 1973, the Council issued new guidelines that allowed the option of using fewer okurigana (Gottlieb, 1995). It was expected that these would be followed in the 1976 White Paper, however, its okurigana use followed the 1959 guidelines. Adverbs were in kanji (e.g. 最も) and the kanji were generally within the Tōyō kanji list (桁 was an exception).

The number of kanji was increased to 1,945 with the issue of the Jōyō Kanji list in 1981 (Seeley, 1991). New style guidelines were issued the same year which recommended the 1973 okurigana rules, but gave the 1959 usage as an option, and specified kanji be used for adverbs (Bunkachō Kokugoka, 1982). In the sample from the 1986 White Paper there was no evidence of the additional kanji being used but this was hardly surprising since there were also no vocabulary items found that required their use (many of the additional kanji were for words such as 'cat' or 'monkey'. Okurigana use generally conformed to the 1959 guidelines but there were a few examples of the 1973 revisions (e.g. 引上げ). Adverbs were mostly in hiragana.

In the 1997 Paper, there was no evidence of the additional kanji that were added to make the Jōyō kanji list but again this was likely due to the vocabulary used. Some kanji outside the list were used (e.g. 剥落). Okurigana use was mixed with both longer and shorter forms appearing. Kanji use for adverbs was similarly inconsistent (e.g. 更に and さらに) but many were written in kanji.

Besides these policy-related changes some stylistic factors were evident. Words that were written in wago in 1960 had been replaced by kango synonyms in 1997. Typical examples included ichijirushii (いちじるしい) 'remarkable' being replaced by kencho (顕著), kuraberu (くらべる) 'compare' by hikaku suru (比較する), medatta (めだった) 'conspicuous' by kenchona (顕著な), and ochikomu (落ちこむ) 'go down' by geraku suru (下落する). Also, more sub-headings were used in 1997. The effect of this was that conjunctions used in bridging passages, written mainly in hiragana, were replaced by sub-headings written mainly in kanji.

It should be noted that was not possible to quantify the effects of particular policies on the relative proportions of kanji and hiragana. It was, however, possible to identify policy changes that favored increased kanji use and determine whether these had an effect. The 1981 reform had the effect of increasing the number of kanji for general use from 1,850 to 1,945 and removed the ceiling on kanji usage by downgrading the force of the guidelines on kanji use (Gottlieb, 1995; Seeley, 1991). In terms of effect on the samples used, this change had no impact. Almost all kanji across the samples were within the earlier Jōyō kanji list and both the 1960 and 1997 White Papers used kanji beyond the official lists on occasion. The changes to okurigana usage, and to the use of hiragana for adverbs and conjunctions reduced the number of hiragana but a considerable time-lag was found between the release of a policy or guideline and its implementation. The 1960 sample was closer to guidelines issued in 1946 than those of 1959. In 1976 the 1959 style was followed − not that issued in 1973. Even in 1997 the 1973 okurigana changes were only partially followed. Nevertheless, these changes did favor an increase in kanji at the expense of hiragana. With regard to adverbs, kanji use was evident in 1976 and more so in 1997. Conjunctions generally remained in hiragana throughout. Again these changes favored an increase in the proportion of kanji. Stylistic change in the form of a shift towards greater kango use also contributed to the relative increase in kanji and indicated a move away from the earlier policy of using a plain spoken style − a shift that can also be discerned in official style guides (see for example Bunkachō, 2003).

The proportion of katakana in the White Papers increased from sample to sample but this increase was smaller than had been expected. When the subject matter and numeral usage were controlled for, the increase was 3.34% from 1960 to 1997 with only a 1% increase in the period 1986 to 1997. Apart from a few instances of katakana being used for numbering in the 1960 White Paper, katakana were found to be used for content words. Most of these were loanwords, the only exception being the Sino-Japanese word tansu (タンス) 'chest of drawers' which is now conventionally written in katakana (the kanji i.e. 箪笥 are not on the lists). Consequently, the increase in katakana script was directly related to loanword usage. We found no evidence of a massive influx of loanwords and our expectation that White Papers on labor would be at the lower end of the scale of loanword usage in print media was confirmed. No loanwords were used in titles or subheadings and there were many pages that contained no katakana. The majority of loanwords were found in sections dealing with the quality of life and referred to common household items and well-known terms. Of the economics-related terms, the majority were compounds comprised of kango with a katakana portion, such as mainasuyōin (マイナス要因) 'minus factor', purasuyōin (プラス要因) 'plus factor', and sābisuryōkin (サービス料金) 'service fee'. A few terms may have posed comprehension difficulties. These included anbaransu 'unbalance' and piiku 'peak' in the 1976 sample; raifusaikuru 'life cycle' and taimuragu 'time lag' in the 1986 sample; and rifōmu 'reform' and rīsu 'lease' in the 1997 sample. The main innovation over the period was the introduction of alphabetic acronyms. The first of these was VTR, which appeared in the 1986 sample. In the 1997 sample, four acronyms (D.I., WPI, CSPI, CPI) appeared and the proportion of alphabetic characters was highest at 0.88%. Since these acronyms all appeared with explanation they could not have been a source of confusion. When the katakana and alphabet proportions are combined, the proportion of script used to write foreign-derived lexical items in the 1997 sample rises to 4.71%.

It was not possible to discern any effect of the use of word-processors. From 1984 their use increased rapidly (Gottlieb, 1998). So it is likely that the writers of both the 1986 and 1997 White Papers used them. However, the use extra-list kanji predated this and I could find no examples where kanji that should not have been used, according to the guidelines, otherwise appeared in the text.


Of the samples of text examined in this and the other studies, White Papers on labor contained a high proportion of kanji, comparable in level to that of newspapers. The level of katakana was lower than that of the other samples and this seems to have been due to the lower level of gairaigo use compared with magazines or newspapers. The level of katakana use has increased over the last few decades but the increase in these White Papers was comparatively small. An unexpected finding was the relative decline in hiragana and increase in the proportion of kanji. This trend was, however, in accordance with the Chūō Kōron study and points to a relative decline in kanji until the 1950s followed by an increase. In the White Papers this change was probably due to a combination of: 1. policy changes that favored the use of fewer okurigana and the use of kanji for words previously written in hiragana (but was not due to the increase in the number of kanji on official lists); 2. a shift in style away from plain speech towards a literary style that employs more kango; and 3. changes in layout that produced more subheadings and shorter prose sections. What was also notable about the effects of official language policies and guidelines was the extent to which the White Papers did not conform and the long time-lag between the release of a policy and its effects.

Over the various studies of text examined, there was considerable variation in script proportions between the types of print media. Therefore the findings of this study cannot be generalized to other written genre. Similarly, a longitudinal study of the change in Japanese script usage in one genre cannot be regarded as representative of the language as a whole. What is needed is longitudinal data from samples from a number of genre that are comparable in terms of subject matter and include controls for confounding factors such as horizontal or vertical format and numeral use. If the same phenomena found in this and the Chūō Kōron study are evident, this would suggest that these changes are ones that characterize written Japanese in the post-war period.


Table A1: Totals and proportions of script types in samples from Chūō Kōron.

Table A2: Totals and proportions of script types in newspapers 1971 (tokens).

Table A3: Total counts of characters in magazines in the seijikeizai class.


Asahi Shinbun (1988). Katakanago hanran kanawan (I cannot bear the flood of katakana words). 13 May, p. 29.

Bunkachō (1994). Kotoba ni kansuru mondōshū: Keigohen [Shinkotoba Shirīzu No 2](Questions and answers on language: Polite speech). Tokyo: Ōkurashō Insatsukyoku.

− − (2003). Kōyōbun no kakiarawashikata no kijun (Standards for writing in government documents). Tokyo: Daiichihōki.

Bunkachō Kokugoka (1982). Genkō no kokugohyōki no kijun (Current standards for writing), new edition. Tokyo: Gyōsei.

Ekuni, S. (1993). Gairaigo ni tsuite (Regarding loanwords). In Kawadeshobō (Ed.), Kotoba dokuhon gairaigo. Tokyo: Kawadeshobō Shinsha. 125-131.

Gottlieb, N. (1993). Written Japanese and the word processor. Japan Forum 5(1): 115-132.

− − (1994). Language and politics: The reversal of postwar script reform policy in Japan. The Journal of Asian Studies, 53(4): 1175-1198.

− − (1995). Kanji Politics: Language policy and Japanese script. London: Kegan Paul.

− − (1998). Keeping up with the Tanakas: The social construction of word processing technology in Japan, 1982-1995. Japanese Studies 18(2): 153-163.

Habein, Y.S. (1984). The history of the Japanese written language. Tokyo: University of Tokyo Press.

Hayashi, O. (Ed.) (1982). Zusetsu nihongo (Japanese language in tables and charts). Tokyo: Kadokawa shoten.

Honna, N. (1995). English within Japanese society: language within language. Journal of Multilingual & Multicultural Development, 16 (1&2): 45-62.

Ishino, H. (1983). Gendai gairaigokō (Thoughts on contemporary loanwords). Tokyo: Taishūkan.

Iwabuchi, E. (1993). Gairaigo. In Kawadeshobō (Ed.), Kotoba dokuhon gairaigo. Tokyo: Kawadeshobō Shinsha. 7-15.

Kajiki, M. (1996). Kotoba chōsa no kongo no kadai (Challenges for future language surveys). Hōsōkenkyū to Chōsa, Dec: 17-21.

Kokugo Shingikai (2001). Kokusai shakai ni taiōsuru nihongo no arikata (Optimum response of the Japanese language to the global society). Science of Humanity, Bensei, 33, 97-114.

Loveday, L. (1996). Language contact in Japan: A socio-linguistic history. Oxford: Clarendon Press.

Maruya, S. (1978). Nihongo no tame ni (For the sake of Japanese). Tokyo: Shinchosha.

Miller, R.A. (1967). The Japanese language. Tokyo: Charles E. Tuttle Co.

Miyajima, T. (1977). Gengoseisaku no rekishi (History of language policies). In Y. Suzuki (Ed.), Kokugo kokuji mondai no riron. Tokyo: Mugishobō. 7-19.

− − (1989). Changes in the language of a magazine. In S. Mizutani (Ed.), Japanese quantative linguistics. Bochum: Studienverlag Dr. Brockmeyer. 31-49.

Mogami, K. (1984). Oyakusho wa katakanago ga osuki? (Do government offices like katakana words?). Hōsō Kenkyū to Chōsa August: 20-38.

− − (1991). Hakusho no naka no gairaigo (Loanwords in White Papers). Hōsōkenkyū to Chōsa, June: 34-45.

NLRI (National Language Research Institute) (1964). Gendaizasshi kyūjusshu no yōgoyōji (Vocabulary and Chinese characters in ninety magazines of today) Vol. III. Tokyo: Shūeishuppan.

− − (1970). Denshikeisanki niyoru shinbun no goi chōsa (Computer study of newspaper vocabulary). Tokyo: Shūeishuppan.

− − (1987). Zasshiyōgo no hensen (Changes in the language of a magazine). Tokyo: Shūeishuppan.

Nomura, M. (1980). Shūkanshi no kanjiganyūritsu (Chinese character content of weeklies). Keiryōkokugogaku (Mathematical Linguistics) 12(5): 215-222.

Ōno, S., Morimoto, T. and Suzuki, T. (2001). Nihon, Nihongo, Nihonjin (Japan, Japanese and Japanese people). Tokyo: Shinchōsha.

Seeley, C. (1991). A history of writing in Japan. Leiden: E. J. Brill.

Sekine, K. (2003). Shinbunkiji no nakano katakanago (Katakana words in newspaper articles), Nihongogaku, 22(7), 30-39.

Shiota, N. (1973). Nihon no gengo seisaku no kenkyū (Research on language policy in Japan). Tokyo: Kuroshio Shuppan.

Shiraishi, D. (Ed.)(1960). Tōyōkanji, gendaikanazukai, okurigana no tsukekata (Guide to the tōyōkanji, modern kana usage and okurigana). Tokyo: Ōkurashō Insatsukyoku.

Sugito, S. (1989). Lexical aspects of the modernization of Japanese. In F. Coulmas (Ed.), Language adaptation. Cambridge: Cambridge University Press. 116-126.

Suzuki, S. (1983). Gyōsei to gairaigo (Public administration and loanwords). Gengo Seikatsu 329: 29.

− − (1990). Kanjijōhōryoku saihakken (The rediscovery of the informational power of kanji). Tokyo: Sotakusha.

Suzuki, T. (1985). Yōgo no genjō to shōrai (The current condition of Western words and the future). Nihongogaku 4: 4-17.

− − (2001). Eigo wa iranai!? (We don't need English?). Tokyo: PHP Shinsho.

Takashi, K. (1990). A functional analysis of English borrowings in Japanese advertising: Linguistic and sociolinguistic perspectives. Ph.D. dissertation, Department of Linguistics, Georgetown University.

Tanaka, M. (2003). 'Gairaigo iikae teian' nitsuite (On the proposal for replacing loanwords). Nihongogaku, 22(7): 50-60.

Tomoda, T. (1999). The impact of loan words on modern Japanese. Japan Forum, 11(1): 231-253.

Twine, N. (1991). Language and the modern state. London: Routledge.

Tsuchimochi, G. H. (1993). Education reform in postwar Japan: The 1946 U.S. education mission. Tokyo: University of Tokyo Press.

Unger, J. M. (1996). Literacy and script reform in occupation Japan: Reading between the lines. New York: Oxford University Press.

Yasumoto, B. (1963). Kanji no shōrai (The future of kanji). Gengoseikatsu 137, 46-54.


I would like to thank Dr Bruce Stevenson, Dr Pam Reynolds, Dr Kais Hamza and Brian May for advice and assistance with statistical analysis. This project was completed with the assistance of research grants from the Faculty of Arts, Monash University.

About the author

Takako Tomoda is a lecturer in the School of Languages, Cultures and Linguistics at Monash University, Australia. She graduated from Seijo University and completed her Master's degree at the University of Arizona. Her research interests include language policy, language change and language contact.

e-mail the Author

Back to Top

Copyright: Takako Tomoda
This page was first created on 19 August 2005. It was last modified on 30 January 2006.

ejcjs uses Dublin Core metadata in all of its pages. Click here to enter the Dublin Core metadata website The Directory of Open Access Journals includes ejcjs within one of the most comprehensive online databases of open access journals in the world. Click here to enter the DOAJ website.

The International Bibliography of the Social Sciences includes ejcjs within one of the most comprehensive databases of social science research worldwide. Click here to enter the IBSS website

The electronic journal of contemporary japanese studies is permanently preserved at research libraries worldwide by the LOCKSS electronic data storage system. Click here to be taken to the LOCKSS homepage.

This website is best viewed with a screen resolution of 1024x768 pixels and using Microsoft Internet Explorer or Mozilla Firefox.
No modifications have been made to the main text of this page since it was first posted on
If you have any suggestions for improving or adding to this page or this site then please e-mail your suggestions to the editor.
If you have any difficulties with this website then please send an e-mail to the webmaster.





Search Now:
Amazon Logo
Search Now: