It was a long journey from the cobbled streets of ancient Rome to the windy coasts of Japan, but when they finally arrived, they were here to stay: the 26 letters of the Latin alphabet. Better known here as ローマ字 (rōmaji, Roman letters), or just アルファベット (arufabetto, the alphabet), these so-called 横文字 (yokomoji, “horizontal letters”) are now an indispensable part of Japan and its language.
The Japanese version of the alphabet is used in mainly two ways: for writing foreign — predominantly English — terms, and for Romanizing Japanese. Somewhere at the borderline between these two, there is also a third way the alphabet commonly makes itself seen: in abbreviations and acronyms. In these cases, each letter has to be read separately. And that’s where the trouble starts.
Because although virtually every literate person in Japan — and that’s more or less the whole population minus babies and toddlers — nowadays knows their ABCs quite well, there is some confusion when it comes to the details of how to pronounce them. And that starts with the alphabet’s very first letter. Is it エー (ē) or エイ (ei)?
Later on, the problem reoccurs with letters J and K, for both of which a version with a lengthened vowel (ジェー, jē; and ケー, kē) coexists with a diphthong edition (ジェイ, jei and ケイ, kei). Incidentally, this also is an issue for O, which can be either オー (ō) or オウ (ou).
Next there is C, which is available in its traditional reading, シー (shii), but more recently has come to be pronounced スィー (sii). This reflects a general trend in Japanese to distinguish between the (formerly indistinguishable) syllables si and shi.
D is a step ahead in this respect, because its default reading is already ディー (dii), even though this sound is not a domestic product either. The clash of old and new becomes most obvious in the term CD, normally pronounced neither スィーディー (sii dii) nor シーヂー (shii jii), but シーディー (shii dii).
D does have an alternative reading, though, and that’s デー (dē). The same holds for D’s voiceless brother T, which in Japanese features as both ティー (tii) and sometimes テー (tē).
Two concurring versions are also available for H. Relatively close to the English pronunciation, there is エイチ (eichi), as opposed to the somewhat older pronunciation エッチ (etchi). This latter also happens to be a synonym for sexual activities, so when in doubt, go for the newer reading.
Moving towards the end of the ABC song, there is no way around V. Since Japanese originally has no “v” sound at all, this one is quite a challenge — both to Japanese lips and foreign ears. Its most established version is ブイ (bui), as for instance in Vサイン (bui sain, victory sign). Two other options are ヴィ(vi) and ヴィー (vii), though these seem to exist mainly on paper.
The standard pronunciation for W is ダブリュー (daburyū). Note, though, that in many combinations it’s clipped to ダブル (daburu, double) and is intended to mean just that. Examples are Wチーズケーキ (daburu chiizukēki, double cheesecake) and Wワーク (daburu wāku, “double work” — moonlighting). For internet addresses, with their quite challenging three W’s in a row, the reading can be further clipped to something like ダブダブダブ (dabudabudabu).
The story concludes with the letter Z, obviously, though it’s much less obvious whether it should be pronounced ゼット (zetto), ズィー (zii) or ジー (jii). As in all of the previous cases, in the end you’ll have to decide for yourself.
Because, strangely enough, it seems that nowhere is it explicitly specified how to pronounce the 26 letters of the alphabet in Japan. Though there is a 1991 notice by the Cabinet Office called 外来語の表記 (Gairaigo no Hyōki, The Writing of Borrowed Words), it doesn’t contain any information on the reading of the letters themselves.
The only official source available dates back to 1900 and is called 羅馬字書方調査報告 (Rōmaji Kakikata Chōsa Hōkoku, Report on the Writing of Roman Letters). That this document is not quite up to date can instantly be understood from the reading given for the first three letters; アー (ā), ベー (bē) and チェー (chē).
In the hope of clearing up things a bit, the national broadcaster, NHK, discussed this problem last year at one of its committee meetings, in which a default reading for each of the 26 letters was specified. The minutes of this meeting also provide interesting insight into the problem as a whole. They are now available (in Japanese) here.