(Early) Old Japanese Phonology

The pronunciation of Early Old Japanese (EOJ), the language of the Nara period and slightly earlier, was very simple. There were 13 consonants, 8 vowels, and a very simple syllable structure. Virtually every syllable was CV (i.e. consonant + vowel), and so most words had the patterns CV, CVCV, CVCVCV etc. This meant that no two consonants ever came into contact. No two vowels, even, came into contact within a word. Only a handful of words began with a vowel (i.e. V, VCV, VCVCV etc.); in poetry, it was common practice that, if two vowels occurred in a row, one would drop, or occasionally the two would fuse to create a single new vowel. Often this ‘fusion’ was not actually written, but it is to be inferred. Whenever you see a line with six or eight syllables instead of the usual five or seven, look at it carefully – usually you’ll see two vowels in a row.

The full set of possible syllables in EOJ are:

pa po pu
ba bo bu
ma (mö) mu
ta te ti tu
da de di du
na ne ni nu
ra re ri ru
sa se si su
za ze zi zu
ka ku
ga gu
ya ye i yu
wa we wi wo u
a e ö

The parentheses around () show that this distinction was already on the way out during EOJ. Kojiki is the only text to distinguish and .


Bilabial Alveolar Palatal Velar
Plosive: p t k
Prenasalised Plosive: b d g
Nasal: m n
Fricative/Affricate: s
Prenasalised Affricate: z
Tap/Lateral: r
Approximant: w y

In terms of pronunciation, it’s probable that p, t, k, m, n, r, w, y were pronounced the same as in Modern Japanese. The ‘prenasalised’ sounds b, d, g, z were similar to Modern Japanese, but had a sort of nasal effect at the beginning. Imagine the type of ‘b’, ‘d’ or ‘g’ that you pronounce when you have a badly blocked up nose, and that’s exactly how EOJ, LOJ and MJ pronounced b, d, g. S was probably pronounced as either ‘s’ or ‘ts’.


Whereas later stages of Japanese had five vowels, EOJ had eight distinctions of vowels. Essentially, the language had two sounds which later merged as a single e, two sounds which merged as i, and two sounds that merged as o. The two e‘s are often represented as ê (or just e) vs ë, or e1 vs e2, etc. Almost certainly, by the end of LOJ, we had eight pure vowels. However, the way in which sequences of two vowels fused in poetry, and etymological links between words with different vowels, means that we can hypothesise that the eight vowels derived from a simpler system of four: a, i, o, u. So, for example, ë was a contraction of a sequence of two vowels, *a + *i.

Transcription Probable Sources
a a a a < *a
ê e e1 ye < *ia, ?*e, ?*ie
ë ë e2 ey < *ai
î i i1 yi < *i
ï ï i2 iy < *ui, *oi
ô o o1 wo < *ua
ö ö o2 o < *o
u u u u < *u

We can only hypothesise what the pronunciation of the eight vowels was. We do have evidence, though. Firstly, writers wrote EOJ mostly in a type of phonetic script, using Chinese characters with the appropriate pronunciation, and the contemporary pronunciation of Chinese (‘Early Middle Chinese’= EMC) gives us important clues. The major limitation on this is that we’re not entirely sure how EMC itself was pronounced! Secondly, there are certain common universal tendencies in the way that some sounds change over time, regardless of language. For example, in the history of different languages ai often develops into the sound that we assume ë had. Thirdly, there are patterns between related words in EOJ that are suggestive. For instance, several words appear to have a root ending in a, which appears in compound words, but in ë (or e) otherwise, e.g. ta– vs te ‘hand’, ma– vs me ‘eye’ etc. Many linguists believe that the forms in ë are the result of fusion with an old grammatical particle *i, i.e. ta– + *i > *tai > te, ma– + *i > *mai > me. Can we identify this *i with an EOJ particle i used in only one genre of writing, which might be characterised as marking the direct object or mild emphasis, or with the subject particle i in Middle Korean – which many linguists believe is related to Japanese? Perhaps.


    • ê vs ë: It’s likely that ê was more ‘close’ (in other words, the tongue was closer to the roof of the mouth) than ë. For example, ê is as in French aimer ‘to love’ and ë is as in French j’aimais ‘I used to love’.


    • î vs ï: We believe that î was a proper i as in the Italian or French pronunciation, whereas ï was pronounced with the tongue further back in the mouth, such as the u-sound in Welsh or the sound in Russian transcribed y. The closest equivalents in English would be that î is as in scheme and ï as in skim.


  • ô vs ö: The distinction is probably similar to that between the vowels in French pot and English pot.

An extra problem is the fact that not all eight vowels follow all consonants. For example, although there is a distinction of two e‘s and two i‘s after k, there is only one type of e and only one type of i after t. But is the i that occurs after t i or i? We can’t be sure (although we have a good idea), so linguists following A, B, C or D-type transcription systems use simple, unmarked e, i and o for such cases where there is no contrast. Parallels from trends in other languages suggest undifferentiated e and i (after t, d, n, r, s, z) might be ê and î, and undifferentiated o (after p, b, w) might be ô, but we can’t be sure.


We know that LOJ was a tone-accent language, with patterns of high and low tone over the space of a word. We assume that EOJ also was, but we have no evidence.

Eastern Dialect:

In the Nara period, some poetry (Azuma-uta) was written in a distinctly different dialect to that of the capital. This ‘Eastern dialect’ had some grammatical peculiarities, but its pronunciation is particularly intriguing. Some of its differences are difficult to explain, usually occurring in only one or two words (e.g. yade vs yeda ‘branch’, nino vs nuno ‘cloth(es)’). Some differences, though, suggest something more systematic. For example, there are a number of examples of ô where EOJ proper (and later forms of Japanese) have u (e.g. yöki vs yuki ‘snow’, ninö vs nunö ‘cloth’, – vs –mu ‘will, may’, –tötö vs –tutu ‘whilst’). The yodan-class of verbs sometimes make a grammatical distinction between this ô and u. There are other examples where Eastern dialects tend to have r where ‘standard’ EOJ has y, such as in passive or imperative forms of verbs (e.g. se ro vs se yo ‘do it’). Perhaps this points to an original ly sound (like Spanish ll), but it’s difficult to interpret the data.

Most of the peculiarities of the Eastern dialect occur only in these poems, and aren’t found in later Japanese. The main exception to this is the r vs y distinction. Later Japanese in this case alone adopts the Eastern r rather than the ‘standard’ EOJ y. The grammatical difference between ô and u in verbs appears to survive in the Hachijôjima dialect.

If you have any questions about (Early) Old Japanese phonology, please email Dr. Nic Tranter.

'Simply moving and elegant'