Can someone explain how 'e' and 'wa' are related in some words / 音便? Presumably the 'e' was originally the obsolete ゑ since it's in the ワ行.
Some examples:
- 上（うえ） ←→ 上着（うわ・ぎ）
- 声（こえ） ←→ 声色（こわ・いろ）
- 終える ←→ 終わる
(Bonus question: How do you type ゑ in the IME? I had to copy it from somewhere else.)
There is a whole class of such nouns that exhibit vowel fronting when used as standalone nouns, and not all of them end in -e or -wa. Among other examples, with in-compound forms followed by standalone:
One of the reasons for the theory that a certain class of nouns was followed by the now-obsolete Old Japanese い, an emphatic nominalizing particle, is that the term カムイ appears in Ainu as a likely borrowing from Old Japanese or slightly earlier, clearly manifesting a distinct む and a distinct い sound. These two over time could conceivably mush together into み, much as similar sound shifts have been observed in even modern Japanese (such as たかい becoming たけえ, すごい becoming すげえ, etc.) and in other languages around the world.
As to why only certain nouns evince this particular phenomenon, it is not unknown in other languages for there to be specific noun classes. It's possible that these nouns in Japanese might be a vestige of an earlier stage of the language that had such a specific noun class. Notably, a lot of these nouns (at least, the ones I'm aware of) seem to describe parts of the body, spirits, and other concepts that would be personally important. Polynesian languages have a roughly analogous noun class covering so-called inseparables, and these nouns take a specific version of the possessive particle, "no", contrasting with the "na" particle used for possessed nouns outside of this class.
Regarding transitivity / intransitivity and verb conjugation patterns, see blutorange's extensive post about this subject. There are several patterns, of which -わる / -える is just one.
At some point far in the past (before Old Japanese, at least) these words probably had a single form with a diphthong:
上: *upai
声: *kəpai
The diphthong turned into a single vowel differently in different contexts: word-finally it became /e/, and word-medially the /i/ was deleted. (The *p subsequently turned to /ɸ/, which then became /w/ between vowels and later /h/ elsewhere except before /u/ - this is why you have 原 /hara/ and 藤原 /ɸuʥiwara/.)
The 終わる・終える question is a bit different, and has to do with some transitivity-flipping morphology that no one really understands well.
-us appear on -every- non-compound use of a first-declension Latin noun? The -i could well have been a "null affix".
If you look at the classical spellings for your examples the answer becomes obvious rather quickly:
Let's take a closer look at them individually. I'll be using romaji in my explanations because it'll make the relationships clearer as we go.
In the case of 終わる ←→ 終える, it's the ワ行 equivalent of a rather common transitive/intransitive pair pattern (cf. 上がる・上げる):
ゑin the conversion list. "e" did not give it either.wye.