Transliterating Japanese, when and why.

I'll start with a summary: transliteration should only be done when the intended audience is expected not to care about the original language, or as a last resort when the textual medium used does not allow the representation of the original language. So with that covered, let me explain this in a few more words.

The common name for the "script" used for transliterating Japanese is called romaji, rōmaji or roomaji, which is itself a transliteration of the Japanese word ローマ字, meaning "roman lettering". There are many different schemes for 'romanising' (the act of transliterating into roman script) Japanese, and broadly these fall apart into two forms:

1. Romanisation used when Japanese needs to be transliterated for an unknown non-Japanese audience that is familiar with roman script.
2. Romanisation used when Japanese needs to be transliterated for a known non-Japanese audience that is familiar with roman script.

While the only difference between these two forms is whether the audience is known, this difference is pivotal to the choice of romanisation. In the first category we finds romanisation systems which favour reflecting the japanese written system over spoken Japanese, whereas the second category favours the pronunciation over written Japanese.

Which to use, then?

The answer to this question should be fairly self evident: it depends on the audience. If it is not known which language the audience is familiar with, such as Japanese publications intended for international publication, then clearly it makes the most sense to rule out phonetic confusion by sticking to a romanisation that doesn't reflect pronunciation. The best know romanisation system that does this is the 訓令式 ("kunreishiki") romanisation, the official standard for converting the Japanese syllabic writing system into roman lettering. Because of this, the romanisation is counter-intuitive for various syllables, prefering the compositionally consistent "si", "ti", "tu", "hu" and "zyo" (amongst others) to reflect respectively し, ち, つ, ふ and じょ while ignoring that to for instance an English speaker these sound like respectively "shi", "chi", "tsu", "fu" and "jo".

Whenever this romanisation is presented, the reader must do a bit of mental conversion to turn the presented text into what it is supposed to sound like, because this kind of syntactic romanisation relies on the idea that all Japanese syllables should fit in a table with consonants on one side, vowels on the other, and every syllable represented as intersection, so that the syllable in the 't' column at the 'i' row is written as 'ti' - even though loose consonants do not exist in Japanese, and the syllables do not conform to this rule.

If the audience is known, a localised romanisation can be used to spare the reader the constant need to do this mental conversion. The Japanese words じょうよう can be phonetically romanised as "djoojoo" for German speakers, but as "johyoh", "jouyou" or "jōyō" for English speakers. One of the problems we see with this approach is that there are three possible romanisations for English, caused by the fact that there is no 'o' sound in the English language that is similar to the one used in Japanese, so 'sounding it out' becomes a compromise between readability and correctness: "johyoh" is not quite readable, but will sound roughly right, while "jōyō" is very readable but will sound wrong.

When to use romaji

This is another question one needs to ask before using romanisation schemes. Before checking whether a 'universal' or 'localised' romanisation should be used, one really needs to ask whether romanisation should be used at all. Here's a list of reasons you might need to use romanisation schemes:

1. The audience could care less about the actual syntax and formants in Japanese.
2. The medium you are relying on does not allow the use of actual Japanese.
3. You underestimate the intelligence of your audience.
4. You cannot be bothered.

These last two seem rather silly, but I wouldn't have added them if they didn't pop up as reasons all over the place. The first reason applies to for instance people who read world news; readers of news items concerning Japan or Japanese endeavours do not care about the Japanese language, they just want to know who did what. As long as the romanisation is uniform throughout the material they are presented, they couldn't care less about which romanisation scheme is used. Typically for this target audience a localised phonetic romanisation scheme is the best choice.

The second reason is when one for instance works with older computers that do not support foreign languages, or modern computers in which getting Japanese input to work means disabling or ruining other functionality of the computer. In this setting the choice of romanisation scheme should usually be accompanied by an explanation of the scheme used, so that people understand why they are not being presented with actual Japanese, and can adjust to the scheme used. The choice of scheme still depends on the audience, a phonetic romanisation being prefered to a syntactic one to make life easier for the reader.

The third reason is not as silly as it initially sounds. There are many educational works written with the intention to teach you Japanese, without actually teaching you how to write Japanese. To make matters worse, often these works will not use a phonetic romanisation - which would make sense if you do not intend to teach writing - but use a syntactic romanisation "because it mimics the Japanese writing system while being more readily accessible to learners familiar with [insert audience's language here]". There are two major problems with this rationalisation:

1. It pretends to reflect the Japanese writing system without actually being Japanese, and
2. It forces the audience to first learn how to properly sound out the romanisation scheme.

If one has to learn how to sound out a romanisation scheme first, one might as well learn the actual syllabic writing system (learning to associate unique pronounciations to 46 pictures is hardly a major effor for people who intend to learn an entirely new language), which in turn negates the first problem: any syntactic romanisation will fail to fully capture the Japanese writing system because of the inherent overlap that romanisation brings with it. Using the syllabic writing system used to write out loanwords and to add emphasis to text, one can for instance write チ and ティ, pronounced "chi" and "ti" respectively. While a phonetic romanisation will be able to differentiate between these two, a syntactic romanisation is forced to write both as "ti". In effect, the use of a system that "represents the Japanese writing system" is actually preventing learners from telling the difference.

The fourth reason, finally, is also not to be taken lightly - when one has the option to use Japanese properly, but you cannot be bothered to use it for whatever reason (say, the time it takes to write an answer to a question in japanese takes longer than to quickly write it in a romanised form) the choice of romanisation is essential. If you cannot be bothered to use Japanese, pick a romanisation scheme that can be read without having to bother with lots of conversion and questions on how to properly pronounce what you just wrote. Use a phonetic romanisation (effectively: if you don't want to make the effort to get the Japanese across properly, do not expect other people to want to make the effort to decode what you wrote either).

The executive summary

Use romanisation when it makes sense: idealy only use it when you must, and even then pick a scheme that does not present readers with unnatural phonetics. If you're wondering whether to use phonetic or syntactic romanisation, ask yourself whether the rest of your text is in a particular language. If it is, the answer should always be phonetic. If you're writing educational matererial and use romaji because it's easier to read, stop and rewrite everything in proper Japanese (with furigana accompanying every kanji) and either teach people how to read hiragana before you properly begin, integrate it throughout the first few bits of your text, or inform people they need to know it already.