As we saw in the Hashimoto Shinkichi feature, Nara-period man’yōgana for kana that sound the same in modern Japanese — ki, ke, ko, and so on — were sorted into two groups, Type A and Type B, that never mixed. So what differed between the two sounds? This article traces the path to the leading answer reached by Arisaka Hideyo and others.
Note that the contents here are explained as well-known facts of Japanese linguistics. They contain no quotation from Arisaka’s own papers (which are not publicly viewable online).
First, line up where A/B shows up
The kana with A/B sorting that Hashimoto showed are these 13.
e, ki, ke, ko, so, to, no, hi, he, mi, me, yo, ro
If we rearrange these by the “columns” of the fifty-sound table, we notice something.
| Column | Kana with A/B sorting |
|---|---|
| i column | ki, hi, mi |
| e column | ke, he, me (the “e” is a slightly different case, a distinction between the a row and the ya row) |
| o column | ko, so, to, no, yo, ro |
| a column | (none) |
| u column | (none) |
The A/B sorting appears only in kana of the i, e, and o columns, and not in a single kana of the a or u columns.
If it were a consonant difference, this wouldn’t happen
Suppose the A/B difference had been a difference in the consonant “k” at the head of “ki.” The “k” at the head of “ka” and “ku” should be the same consonant, so “ka” and “ku” ought to have A/B too. In fact, though, “ka” and “ku” have no sorting.
On the other hand, the kana that do have A/B sorting — “ki,” “ke,” “ko” — are exactly those with the vowels “i,” “e,” and “o.” The most natural way to explain this lopsidedness is: the A/B difference lay not in the consonant but on the vowel side.
Five becomes eight
The vowels of modern Japanese are five: a, i, u, e, o. If the A/B difference lay on the side of the vowels i, e, and o, then Nara-period Japanese would have had these vowels.
| Modern vowel | Nara-period vowels |
|---|---|
| a | a (one) |
| i | two: i (Type A), i (Type B) |
| u | u (one) |
| e | two: e (Type A), e (Type B) |
| o | two: o (Type A), o (Type B) |
1 + 2 + 1 + 2 + 2 = 8. This is the so-called “eight-vowel theory.” The Type B vowels are often written in notation as ï (i-B), ë (e-B), and ö (o-B).
But it isn’t “settled”
The eight-vowel theory is a leading one, but not the only answer. Views have also been put forward that read the A/B difference not as a difference of the vowel itself but as a difference in the consonant (for example, palatalized versus non-palatalized consonants) or in the sounds around the vowel. What the Type B vowels actually sounded like — for instance, whether they were central sounds, further back in the mouth — is still debated today. What is certain is the fact that two systems were written separately, and the lopsidedness that the difference shows up only in the family of i, e, and o. The eight-vowel theory is one of the most widely accepted hypotheses to explain that fact.
In the next article, we look at a law of vowel combinations within words that supports this hypothesis more strongly still.
Summary
| Point | Detail |
|---|---|
| Observation | A/B sorting appears only in kana of the i, e, and o columns, not in the a or u columns |
| Inference | If the head consonant differed, it should appear in the a and u columns too — so the difference lay on the vowel side |
| Eight-vowel theory | i, e, and o each split into A and B, making the vowels 5 + 3 = 8 |
| Caution | Not the only answer; the actual sound of the Type B vowels is still debated |
In a nutshell — a sorting that appears only in the family of ki, ke, and ko suggests the difference between the two sounds lay on the vowel side. The theory that Nara-period Japanese had eight vowels is widely accepted as a hypothesis that explains that lopsidedness.