On Tue, 8 Nov 2016 18:52:38 -0500, Anthony Miles <[log in to unmask]> wrote:

>The South American language Tariana has a phenomenon known as H metathesis,  in which a glottal fricative is metathesized to the  initial consonant of the root  regardless of how many syllables that may be in distance. Is this long-distance metathesis attested elsewhere  or is it unique to this language?  

Basque has a rule of identical description: for instance 'sand' after borrowing from Romance developed _arena_ > _areha_ > _harea_.  Ancient Greek had something similar, though the fact that its */h/ was lost intervocalically meant that the input to the rule was limited: still, for instance, *s > /h/-initial verbs gaining the augment /e-/ in the imperfect brought the /h-/ out front, thus 'follow' has present stem _hep-_ and imperfect stem _heip-_ (with spurious diphthong) < _he.ep-_ < _ehep-_.  

>How could such a phenomenon develop?  Is it restricted to the glottal fricative?  If not, are there restrictions, in theory or practice, in the type of consonant that can undergo this phenomenon?  

I forget whose it is, but the typology of consonant metathesis I've seen suggests that this type of long-distance metathesis should be limited to consonants with long, smeared-out perceptual cues, that can overlay the rest of the word without interacting severely with it.  Such consonants are mostly glottals.  If the breathiness, say, that signals /h/ is smeared out over a long part of the word, you could just as easily interpret the same pronunciation as being the realisation of a word where the /h/ was someplace else.  So a language could change by starting to make a consistent choice in all such cases, e.g. that /h/ is always initial, and there's your metathesis.  This is very much the sort of process whose outcome would be likely to be influenced by making a more restrictive canonical root shape, as you allude to below.

The consonants which can do this are basically the same ones subject to long-distance dissimilation, and the reason is similar: if /h/ is smeary, a second /h/ in a word could be masked by the smear of the first one.  For contrast you wouldn't expect a consonant like, say, [j] to participate in either of these sorts of changes, 'cause if you start smearing its realisation out suddenly you've fronted and raised every vowel in the word (that would probably lead to harmony instead).

I suppose there is a sort of long-distance metathesis which this etiology doesn't cover, and that's the case of liquids.  E.g. South Italian Greek systematically moves post-consonantal /l r/ to the initial syllable, so ancient Greek /k_hon'dros, 'pediklon/ 'thick, fetter' become SIG /xron'do, 'pletiko/.  I don't know how to account for that.  (Maybe generalisation from production errors?)

> Since this is a South American language,  it is probably worth considering the scope of nasal harmony in other languages of the continent, and whether H metathesis is the beginning or end of such a historical process.  Moore probably it is unrelated altogether. 

Hard to say given the limited amount you've told us, though I agree if there is a relationship it's more likely in terms of which stem shapes licence each process than a phonetic one.  (But don't forget rhinoglottophiila).