On 3/8/2012 9:08 PM, Alex Fink wrote:

>   One bumps fairly frequently into languages with a phoneme called /c/,
> although sometimes its usual realisation will be [cC)]; but aside from
> Australian languages (with their many areal anomalies, e.g. sibilancy being
> right out) I can't name offhand any language I've seen described as having a
> phoneme /t_-/, rather than /tS/.

I'm willing to accept a few phonemes that might be unnatural for human 
languages, including things like postalveolar stops if it works out that 
way. But without any really compelling reason for anything like that, 
I'd prefer to stick to what's realistic for human languages, so I'll try 
postalveolar affricates in Early Modern Tirelat for now and see how they 
work out.

>> [?r/r?] does'nt strike me as the likeliest source of
>> [r_0]. Any surrounding voiceless fricatives or preceding
>> oral voiceless stops seem more likely -- the latter
>> perhaps after becoming fricatives.
>
> Yeah, glottal stops are voiceless in the opposite of the fashion that most
> other things are voiceless, glottis completely shut rather than not
> vibrating but completely open.  Modal voicing uses an intermediate degree of
> aperture.  As a consequence, devoicing (specially) near [?] doesn't really
> happen.  If [?r] went somewhere special the normal thing would be [r_k],
> proximally at least.  (But creakiness doesn't interact with non-phonation
> features of consonants much either.  Maybe it could>  laryngealisation>
> pharyngealisation or such.)

Okay, I can see how that makes sense. Which voiceless sound would be the 
most likely to vanish without leaving traces of breathy aspiration on a 
voiceless r? Something like /f/, possibly? Or should I stick with /hr/ 
and /rh/?

>>> [re production of /j/ and /w/]
>>
>> I'm thinking some sort of vowel breaking may be involved, like what
>> happened in Spanish with /ie/ and /ue/.
>
> Yes, that could well be.  But the distributional corollary of thàt would be
> that /j/ and /w/ occur with only one or a few nuclear vowels apiece, unless
> later umlauts or similar changes increased the range of vowels you could
> have after them.  Does that square with the facts?

I've been looking at /w/ and I think now that it makes more sense as a 
consonant. The absence of *bw, *pw, *fw, *vw (even though /u/ occurs 
after labial consonants) is one clue. Also, /w/ is somewhat common 
between vowels, in places where there can't be a lost consonant before 
it since the vowel before it is short. (That's assuming that I decide to 
go with compensatory lengthening as the explanation for the long vowels 
in the standard dialect.) It might not have been *w originally (possibly 
*ɣ as I mentioned in another post), but it looks like it was a consonant 
of some kind.

>> One clue is the word "r'jandi"
>> /r_0jandi/, which I'm assuming came from an original *hrjandi or *?rjandi.
>> A 3-consonant initial cluster is unusual for Tirelat (I can't find any
>> examples of one)
>
> You mention on your website (I checked your website since last time!)
> /tkw1r/ -- which is a three-consonant cluster, but with an approximant in
> the third place!  If the glides are recent creations then they were probably
> created after two-consonant onsets just as well as after one-consonant ones,
> which means you should expect initial three-consonant clusters of the form C
> + C + /j,w/.  So I wouldn't be bothered by this.

Well, whether /j/ and /w/ were consonants is the question, so /tkw1r/ is 
just another example of CC + approximant. But that does make it seem 
more likely that /w/ was an approximant instead of something like [ɣ] 
(unless there was a vowel between /t/ and /k/ that was lost), since 
/tkG/ seems like an unlikely initial cluster for EMT.

> Riffing on the theme more loosely, what sort of initial two-consonant
> clusters are out there?
>
>> Well, [D] is just the allophone of /d/ between vowels.
>
> Ah!  Missed that.
>
> But then I see on your website that /d/ and /b/ have intervocalic fricative
> allophones, but not /dz/ and /g/.  Presumably this has something to do with
> the existence of /z/ and /G/ but not of */D/ and */B/ (though note that a
> distinction between /B/ and /v/ is rare!  I'd be a little surprised if /b/
> fricated to something other than [v] in a language with extant [v]).

I've gone back and forth on whether the /g/ vs. /G/ distinction remains 
between vowels. I don't think there are many minimal pairs. Let's see: 
rogu "scissors" vs. roġu "less"; žagi "greedy" vs. žaġi "to dash". 
That's it? I think I could merge intervocalic /g/ and /G/ without much 
trouble.

As far as /b/ vs. /v/, there's one significant minimal pair: kabi 
"overturned" vs. kavi "black". I don't really want to change either of 
those if I can avoid it.

> Seeing that state of affairs, what I would hope is true is that /dz/ and /g/
> are absent intervocalically (except in cases where they morphologically vary
> with non-intervocalic stops), because they were subject to the same
> intervocalic frication as /b/ and /d/ but the results merged into extant /z/
> and /G/.  If you were rather thinking that their frication was _blocked_ by
> the existence of /z/ and /G/, that's more suspect IMO.  (Especially given
> that voiced stops are less stable the further back in the mouth they go.)

/dz/ does contrast with /z/ intervocalically (keża "to announce" vs. 
keza "to look for"). I wonder if this could actually be from a different 
origin, stop + fricative (*kedza)?