[Journal of the Simplified Spelling Society, 1988/1 p23,24 later designated J7.]

Sound and Symbol: the Case of Romaji.

John Skelton

John Skelton has taught English and Applied Linguistics in Spain, London, Oman and Singapore, and is now Director of Studies of Aston University's Language Studies Unit. He has travelled widely in Asia and has been learning elementary Japanese for many years.

1 Origin and use.

The Japanese writing system was hailed by the first missionaries to Japan as an invention of the devil (a judgment with which not all students of the language have disagreed over the years), and it was those same missionaries who made the first attempts to render into the roman alphabet the sounds of the language.

The use of romaji, with which this article is concerned, is at once very widespread and very restricted in contemporary Japan. It exists firstly as an aid to foreigners.

Thus for instance the names of tube stations in Tokyo appear in romaji - though as no such concessions are made on overground trains, and Tokyo itself does not have street names, this can often be less help than is necessary. Similarly, brand names often appear in romaji - a necessity in an export-obsessed country. Secondly, romaji have value as fashionable chic, appearing in slogans on the twee T-shirts so in vogue all over east and south east Asia, the English mangled and the sentiments ghastly - no one who has seen a Japanese woman in her mid-twenties proclaim from her T-shirt, "We are little people. We should laugh and play" can sustain undamaged a belief in the dignity of humanity. (See Dougill [1987] for further examples and discussion).

2. Converting sound into symbol.

Romaji, then, is fairly unusual in written scripts in that it is not the major, or a major, vehicle for written communication between speakers of the language. On the other hand, the way it operates to convert sound into symbol, and the linguistic decisions involved, are at once typical and instructive. I should add that in the present paper I shall concentrate on the more widely used Hepburn system: the other principal system, kunrei-siki (as it would write itself), has been officially preferred but is less common and, because on the whole it makes a less detailed attempt to match Japanese sound with the English equivalent, is linguistically less interesting.

All writing systems strike a balance between phonetic fidelity and economy and - when regularised by a Royal Society, an Academia, or by some such smaller but no less dictatorial authority as the anthropologist-between regional variation, if any, and standardisation. In some languages the, writing system may carry further information, as English carries part of its etymology enshrined in its spelling (where /f/ is written <ph> the word is usually Greek, and so on). And all attempts at romanisation refer back to a shifting but related set of sounds, based on those of Latin, to go back no further. Thus in any language written with roman characters the sounds associated with individual letters will be broadly similar, if seldom identical.

3. Systems for writing Japanese.

The situation with romaji however is different from that of the major European languages or from the myriad tribal languages which have never been written in any other way. Japan had a literate culture centuries before the arrival of Europeans, and it is as well to give a few words of explanation about this first of all.

Originally, then, ideographs were borrowed from Chinese, retaining sometimes their Chinese meaning and sometimes their Chinese sound. Many of these ideographs came to be associated with other sounds and other meanings as well. Thus the first character in the Japanese name for their own country, which resembles a capital <E> with a vertical line closing it off down the right hand side, has readings (known as on readings) based on Chinese pronunciations: jitsu and nichi. And it has readings (kun readings) in which the ideograph retains the Chinese meaning but represents a native Japanese word: hi, ka. The on readings need not represent lexical words, kun readings do (though ka here is a bound morpheme). The meaning of this particular ideograph is either day or sun. And of course in the combination romanised as Nihon or Nippon it means sun, but exasperatingly is pronounced /ni/.

Given this complexity the invention of two syllabaries, known collectively as kana, individually as hiragana and katakana, is unsurprising - in any case the structural differences between Chinese and Japanese (they are as different from each other as from English) render Chinese ideographs a clumsy way of capturing the grammatical facts of Japanese. Now the kana themselves are a linguistically self-conscious and sophisticated means of symbolising the range of possible syllable sounds in Japanese. Those who attempted romanisation, therefore, found themselves in the unusual position of having three sources to draw on: the sounds of the spoken language, the way in which it was normally written, and two attempts to offer a simpler and phonemically accurate means of representing it.

4. The letters of romaji.

I turn now to romaji itself. Twenty two different roman letters are used in the romanisation process. Those excluded (in the Hepburn system; kunrei-siki also omits <c, f, j>) are <l, q, v, x>. However <c> only appears in the combination <chi>, pronounced as in cheese, and <f> only in <fu>, as - more roughly this time - in full. The five vowel symbols used are <a, i, u, e, o>, this being the Japanese order; and a basic rule of thumb for romaji pronunciation, which might as well be introduced here, is that vowels are to be sounded as in Italian, consonants as in English. These five vowels cover the five monophthongs in Japanese, and also appear in ten diphthongs.

Such, then, is the basic system. But the learner requires, more or less explicitly, further detailed information.

5. Pronouncing the letters.

a) As in the case of all foreign languages written with the roman script, the learner needs more information about the exact pronunciation - the exact place in the general area of sound conventionally represented by the roman character. Thus a textbook might point out that <w> is like English <w> but pronounced with spread lips, that <f> has a definite /h/ quality and so forth.

One letter, <r>, represents a sound which has no real English equivalent. It may in fact frequently be perceived as having a /d/ quality. That is to say, the general ethos of 'best approximation', of choosing a roman letter representing a broadly similar sound, verges here on the misleading. This is, however, the only sound which is seriously problematical. Compare the problem faced by romanisers of Arabic.

b) There is a general rule that double consonants and double vowels do not have their English equivalents. Both are held for longer, and for roughly twice the articulation of single consonants or vowels. Japanese has numerous minimal pairs where the distinction is made by the opposing of single or double sounds. The distinction is recognised in the kana system, as one would expect.

Where the consonant is a stop, as in gakkoo (school), the articulation is held for a beat before being released, as in Italian. To English eyes the pronunciation of <oo>, as here, or <uu> as in Kyuushuu (the island written in English as Kyushu), as long versions of <o> and <u> respectively, may seem psycholinguistically odd, and it is interesting that many textbooks prefer to write a single <o> with a line over it to show the doubling.

c) There is also a general rule that <i> and <a> are devoiced between voiceless consonants or at the end of a word following a voiceless consonant. Since the verb markers which (very approximately) correspond to a kind of present and past tense, among other uses, are <-masu> and <-mashita>, this devoicing is extremely common both in the real language and in elementary textbooks, and the sounds are typically perceived as <mas> and <mashta>. The devoicing is not marked in roman, and the result is misleading, mispronunciations of such proper names as Takashita by newsreaders relying on romaji being commonplace and understandable.

What seems to have happened here is that, with such sequences as <-masu> and' <-mashita> being normally written in hiragana (as bound morphemes are in Japanese), and therefore with symbols representing /ma/, /su/ and so forth, the romajiare based not on the sounds of the language but on the kana system in which they are encoded. It is an interesting point that romaji is used to reflect a syllabary, and indeed might just as easily be regarded as a syllabary itself.

d) This brings us now to the question of sounds which are allophonically distinct in Japanese, but phonemically distinct in English. This is a difficult relationship for romanisers of any language. As far as Japanese is concerned, it may be said in general that the decisions taken for the Hepburn system do not always appear to be formally consistent, but are psycholinguistically easy to use, while kunrei-siki perhaps achieves greater formal elegance but at the expense of what is known these days as user-friendliness (for instance, <siki>, as above, is pronounced "shiki").

Thus <g> is pronounced as in English go where word initial, but as a velar nasal when medial by many speakers, and the subject postposition, <ga>, one of the most common grammatical particles in the language, is also often pronounced with a velar nasal. This distinction in sound is not recognised in romaji.

6. The kana table.

Contrast this decision with the representation of the sounds of the kana table, which Japanese children learn as English children their alphabet. It begins as follows:


and is read (and chanted) from left to right. The next consonants to be dealt with are, in order, <s, t, n, h>. Of these, only <n> is straightforward. The group yields (for Hepburn):


Some of these sounds resemble closely the English equivalents implied by this romaji transcription - the embarrassing pronunciation in an advertising jingle of the Tokyo department store My City, for instance. Some do not, at least to my ears, though all are potential areas of difficulty for Japanese learners when it comes to sound discrimination in English.

A psycholinguistically interesting point, in this general area, is the case of syllable final <n>, which is in fact the only consonant that may close a Japanese syllable. As in, English, phonemic /n/ assimilates to [m] before a bilabial, and it ought therefore to be written with <n> in romaji. In fact, in some systems it is, and in some it is not: thus one finds both empitsu and enpitsu for pencil.

7. Conclusion.

Such, then, are the decisions taken in the romanisation of Japanese. A restricted set of letters is used to write a larger number of sounds, and where there is a choice to be made between formal accuracy and psycholinguistic case, Hepburn at least has gone for the latter, though it perhaps carries with it an increased risk that learners of Japanese will take Japanese sounds to be more like English than they really are.


J Dougill, 'English as a Decorative Language' English Today, October 1987.

Back to the top.