Re: CHAT: query: where to start?

Tim Smith wrote:

> Speaking of odd kinds of morphology: in the conlang project that's
> currently getting the largest share of my attention (which doesn't have a
> name yet), the way aspect is marked on the verb is that the imperfective
> always ends in a nasal, and the perfective always ends in the corresponding
> voiceless stop.  For example:
> [...]
> In other words, the point of articulation of the final segment is part of
> the stem, but the manner of articulation is the affix (if it can still be
> called an affix when it's not even one whole phoneme).  This sort of came
> to me in a flash of intuition, but I don't think I've ever heard of
> anything like it in a natlang.  Does it sound at all plausible?
Remember, though, that morphology is processed only after phonology is.
That means that since you're making meaningful distinctions with place of
articulation, those places of articulation must also be making phonemic
contrasts; you have minimal pairs here.  So, it's not so much that manner
of articulation *is* the affix as, diachronically, the different morphemes were
once disambiguated in some other way.  For example, it could have been
that at one point there was a single nasal consonant that connoted the imperfective
aspect, with the other consonant adjacent to it, perhaps at a different place
of articulation.   Say,

*pela-n-t            *tera-n-k
*pela-t               *tera-k

Then a sound change or two come along and wipe out that neat morphology, which
result in assimilation of some kind:

*pela-n-t > *pela-n-n > pela-n        *tera-n-k > *tera-N-k > tera-N
*pela-t > pela-t                                *tera-k

See, isn't that easy?  All it requires in the first case is one simple shift of /t/ to
assimilate in manner of articulation to /n/;  in the latter case, you need to assume
that /n/ shifts to [N] allophonically before [k] (a very normal soundshift), and
then that final /k/ drops off.  Or, indeed, you could just say that the /n/ becomes
[N] allophonically before [k], and then another soundchange drops all final
voiceless consonants, which would get rid of both final /t/ and final /k/, in both
words.   So, in other words, it's not hard to imagine that kind of thing coming
about.  Synchronically, it would indeed *look* like the rule you stated (at least,
based on the information you have provided here).

Maybe I've muddled myself or something, but I think that's how you could do it.

======================================
Tom Wier   |     "Cogito ergo sum, sed credo ergo ero."
======================================