Elliott Lash wrote:
>  What do you mean by apparent metasthesis? It sounds
>very reminiscent of the historical phonology of
>So, for pre-Silindion we have  *palan 'rock'
>*tho:lon 'river'
>This becomes Silindion  <palna>, <sulno>
>I'm not yet convinced that this is actually
>metasthesis, however. I think that the historical
>changes were:
>  *palan  > *pal@n > *paln= > paln@ >palna
>  *tho:lon > *so:l@n > *so:ln= > so:ln@ > *so:lno >
>So we have both loss of the unstressed vowel then
>vowel harmony, changing /@/ into the vowel of the
>first syllable.

Yes, that works, and would qualify as an "apparent" metathesis too.
Just noticed: both exs. have the same V in both syllables; what about 
something like **palin ?

What happens in Leti and its kin (6 or so lgs.) is also historical-- it's in 
an area where many of the languages have lost final consonants, all or in 
part. But a few, including "Proto-Leti", preserved them by adding a support 
vowel-- in some, it's reconstructible as **-@ while in PL it was a vowel 
echoing the old ultima V. So we have things like *vérat-a 'heavy', *?úlit-i 
'skin' *átur-u 'arrange'  etc. (only *a,i,u could occur in the ultima, and 
stress was/remained on the antepenult). Then, probably in the proto-lang., 
the unstressed ultima V was syncopated, leaving forms like *verta, *?ulti 
and *atru.  From a generative point of view, the underlying forms of the 
modern languages still have the full form + echo vowel, and repeat the 
syncope rule. To some extent the two variant forms have been 
grammaticalized, though it's not entirely clear...e.g. possessive suffixes 
go on the full form (minus echo V); in some cases, use of one or the other 
depends on surrounding sounds-- the languages seem to favor a VCCVCCV... 
rhythm; one writer thinks it's simply (or originally was) a matter of fast 
speech forms. In synchronic analysis, all you can do is call it metathesis.  
There's at least one interesting paper in the Rutgers Optimality Archive 
discussing this from an OT view-- I find it unncecessarily complicated :-( 
The author is Elizabeth Hume, but I don't recall the title.

But wait, there's more: something similar can happen in compounding: Leti 
pipi 'goat,sheep' + duma 'goat, sheep < Ml. domba' can end up as pipdiuma. 
Furthermore, the -u and -i of personal prefixes also metathesizes with the 
verb's initial C: au+la: 'I go' > alua: etc. (the -u- is semi-vocalic)-- 
these look more like anticipatory phenomena. They happen in several other 
languages of eastern Indonesia (and some quite far afield; I once thought it 
might be evidence for subgrouping, but it's more likely an areal thing-- but 
still, mighty strange IMNSHO.)

Anyhow, I'm thinking that in Prevli, the historical development may have 
been something like:
1. either *nátak + echo vowel > nátaka, or perhaps *nátaka to start with.
2. a rule voicing C between unstressed V, which would give **nátaga
3. loss of the post-tonic V, > **natga
(various assimilations will deal with that -tg- cluster, not yet 
decided...maybe ['na?ga])

Perhaps, stress will shift for various grammatical purposes, so we could 
have **natáka, which will result in **ntáka; or a prefix, say **í- > ínataka 
 > **íntak or maybe íntag ['intaG]. As I said, it looks as if it could 
become quite what I've played with so far, the 
basic form of the word tends to become quite obscured, not sure that's a 
Good Thing.  But this is going to be a weird language, at least 

>Another type of faux-metasthesis in Silindion results
>in what looks like an infixed /n/.
No comment on this; except yes, your explanation works, too.