Where did it go?
The mystery of the missing thesis.
I dropped by Blackwell's last night to check out the latest issue of the Transactions of the Philological Society of which I've been a member for about ten years now. I downloaded an article by Don Ringe which was addressing the old problem of the genitive plural in Gothic which is, unexpectedly, -ē rather than -ō. I was wondering what I'd said about it, which is why I was looking for a copy of my thesis.
What actually caught my eye was the old myth of trimoraic vowels as an explanation for different reflexes of the same vowel. I don't know who first suggested the idea, but I believe it's been around for quite some time. Although there are trimoraic long vowels, I believe that such entities as distinctive sounds in a language are very rare. I think they're found in Estonian, but it seems more likely they're a surface (phonetic) phenomenon rather than a phonemic one.
The point about the assumption that there were trimoraic vowels in Proto-Germanic (PrG) is that they have an extremely restricted distribution. Basically, there's no evidence for them outside final syllables, although that's not conclusive evidence to say that they didn't occur elsewhere. Nonetheless, if they had, then they might be expected to leave some sort of specific trace that allows them to be distinguished from bimoraic and monomoraic vowels. In Old English, long vowels in non-initial syllables were shortened while short vowels either merge into a subset of the vocalic inventory or are lost. There's no evidence that vowels in non-initial (i.e., non-root) syllables remained long because they were derived from trimoraic long vowels. As far as I'm aware, there's no evidence from any of the other Old Germanic languages for VVV ~ VV ~ V.
There's also a typological problem with the assumption of word-final trimoraic vowels. The structure of word-final syllables is typically limited in the world's languages. Finnish only allows l, n, r, s, t or a vowel; I believe Classical Greek only allowed ν, ρ, ς or a vowel; in Standard German word-final obstruents are voiceless; in Sanskrit, I think the only segments permitted in a word-final position are p, t, k, h and vowels. (r also??) Chinese only allows a vowel, n or -ng [ŋ] at the end of syllables; Japanese only vowels or n.
Although typology is about tendencies rather than absolutes in language, the introduction of trimoraic vowels in final syllables in Germanic is typologically weak because, as I noted above, such syllables generally have restrictions placed on them that don't affect other syllables.
Another explanation which was put forward for the different reflexes of the same vowel sound word-finally in the Germanic languages was tone. This was put forward on the assumption that because Greek and Lithuanian distinguished acute and circumflex tones, Indo-European (IE) must've done the same.
In other words, there's no sound basis to assume acute vs. circumflex accent on word-final syllables in PrG. Not only that, but the reasons for rejecting the existence of trimoraic vowels in PrG are equally as applicable in this case.
However, I think accent might be behind the different reflexes of PrG *ō in word-final syllables. We know from the effects of Verner's Law that before the stress shift in PrG, stress assignment was lexical, being assigned to the leftmost underlying accented vowel or, in the absence of underlying accents, the initial syllable. That means that paradigms must've had words with final and non-final stress. It is possible that this resulted in some sort of allomorphy, the stressed reflex of *ō remaining unchanged, but the unstressed reflex becoming *ā. The stressed and unstressed paradigms continued in tandem, but end up merging in part to create a hybrid paradigm which prevents the PrG ō-stem endings from completely merging into each other to produce an undifferentiated oblique case ending.
This is no more than a theory, of course. It assumes that in late PrG, unstressed *ō became *ā. Unfortunately, this still means that the behaviour of *ō in word-final syllables is different from non-final ones because I'm not aware that there's any evidence for *ō > *ā non-finally at an early date. Possibly, the special status of final syllables would allow for such a change. One other thing that I should note is that the syllables in question are all closed (i.e., -ōC).
The problem really arises with the change itself. We need to find some reason why PrG *ō in word-final syllables failed to develop uniformly. The proposed answers to the split of *ō in word-final syllables all depend on the existence of something exceptional about such syllables; something that has no effect outside them even if the same conditions are met. The only thing that makes these syllables obviously different is that they're word-final. We have indirect evidence for stress, but that fails to explain anything without assumptions. We have no solid evidence for trimoraic vowels.
Call me Mr Partial, but I still think my explanation has an edge over the other two. I find trimoraic vowels implausible, and acute vs. circumflex accents have been shown to be untenable.
Gothic, to return to Ringe's article, is a little odd among the Germanic languages. My study of the morphology of the Old Germanic languages led me to conclude that the stress shift affected East Germanic first and that the change then affected the North and West Germanic languages. Probably.
That was fun. I haven't written anything like this in far too long.
I dropped by Blackwell's last night to check out the latest issue of the Transactions of the Philological Society of which I've been a member for about ten years now. I downloaded an article by Don Ringe which was addressing the old problem of the genitive plural in Gothic which is, unexpectedly, -ē rather than -ō. I was wondering what I'd said about it, which is why I was looking for a copy of my thesis.
What actually caught my eye was the old myth of trimoraic vowels as an explanation for different reflexes of the same vowel. I don't know who first suggested the idea, but I believe it's been around for quite some time. Although there are trimoraic long vowels, I believe that such entities as distinctive sounds in a language are very rare. I think they're found in Estonian, but it seems more likely they're a surface (phonetic) phenomenon rather than a phonemic one.
The point about the assumption that there were trimoraic vowels in Proto-Germanic (PrG) is that they have an extremely restricted distribution. Basically, there's no evidence for them outside final syllables, although that's not conclusive evidence to say that they didn't occur elsewhere. Nonetheless, if they had, then they might be expected to leave some sort of specific trace that allows them to be distinguished from bimoraic and monomoraic vowels. In Old English, long vowels in non-initial syllables were shortened while short vowels either merge into a subset of the vocalic inventory or are lost. There's no evidence that vowels in non-initial (i.e., non-root) syllables remained long because they were derived from trimoraic long vowels. As far as I'm aware, there's no evidence from any of the other Old Germanic languages for VVV ~ VV ~ V.
There's also a typological problem with the assumption of word-final trimoraic vowels. The structure of word-final syllables is typically limited in the world's languages. Finnish only allows l, n, r, s, t or a vowel; I believe Classical Greek only allowed ν, ρ, ς or a vowel; in Standard German word-final obstruents are voiceless; in Sanskrit, I think the only segments permitted in a word-final position are p, t, k, h and vowels. (r also??) Chinese only allows a vowel, n or -ng [ŋ] at the end of syllables; Japanese only vowels or n.
Although typology is about tendencies rather than absolutes in language, the introduction of trimoraic vowels in final syllables in Germanic is typologically weak because, as I noted above, such syllables generally have restrictions placed on them that don't affect other syllables.
Another explanation which was put forward for the different reflexes of the same vowel sound word-finally in the Germanic languages was tone. This was put forward on the assumption that because Greek and Lithuanian distinguished acute and circumflex tones, Indo-European (IE) must've done the same.
Tangental note: The acute accent in Greek represents aá (i.e., the tone falls on the second mora); the circumflex accent in Greek represents áa (i.e., the tone falls on the first mora). Confusingly, this is reversed in Lithuanian so that the former is the circumflex accent and the latter the acute. I've also seen one article which said that in Modern Standard Lithuanian there is no (longer a) tone accent, although there does seem to be a distinction depending on which mora is stressed.
I think it was the Dutch scholar Alexander Lubotsky who concluded that the lexical accent in Greek couldn't be relied on for the reconstruction of the IE accent, an idea that Morris Halle followed up in an article.
In other words, there's no sound basis to assume acute vs. circumflex accent on word-final syllables in PrG. Not only that, but the reasons for rejecting the existence of trimoraic vowels in PrG are equally as applicable in this case.
However, I think accent might be behind the different reflexes of PrG *ō in word-final syllables. We know from the effects of Verner's Law that before the stress shift in PrG, stress assignment was lexical, being assigned to the leftmost underlying accented vowel or, in the absence of underlying accents, the initial syllable. That means that paradigms must've had words with final and non-final stress. It is possible that this resulted in some sort of allomorphy, the stressed reflex of *ō remaining unchanged, but the unstressed reflex becoming *ā. The stressed and unstressed paradigms continued in tandem, but end up merging in part to create a hybrid paradigm which prevents the PrG ō-stem endings from completely merging into each other to produce an undifferentiated oblique case ending.
This is no more than a theory, of course. It assumes that in late PrG, unstressed *ō became *ā. Unfortunately, this still means that the behaviour of *ō in word-final syllables is different from non-final ones because I'm not aware that there's any evidence for *ō > *ā non-finally at an early date. Possibly, the special status of final syllables would allow for such a change. One other thing that I should note is that the syllables in question are all closed (i.e., -ōC).
The problem really arises with the change itself. We need to find some reason why PrG *ō in word-final syllables failed to develop uniformly. The proposed answers to the split of *ō in word-final syllables all depend on the existence of something exceptional about such syllables; something that has no effect outside them even if the same conditions are met. The only thing that makes these syllables obviously different is that they're word-final. We have indirect evidence for stress, but that fails to explain anything without assumptions. We have no solid evidence for trimoraic vowels.
Call me Mr Partial, but I still think my explanation has an edge over the other two. I find trimoraic vowels implausible, and acute vs. circumflex accents have been shown to be untenable.
Gothic, to return to Ringe's article, is a little odd among the Germanic languages. My study of the morphology of the Old Germanic languages led me to conclude that the stress shift affected East Germanic first and that the change then affected the North and West Germanic languages. Probably.
That was fun. I haven't written anything like this in far too long.
Comments