On 3/7/2012 1:00 PM, Alex Fink wrote:
> On Tue, 6 Mar 2012 20:16:35 -0500, Herman Miller<[log in to unmask]>  wrote:
>> The consonant inventory of modern Tirelat looks something like this:
>> p b	t d	ts dz			k g
>> m	n				N
>> 		r0 r
>> f v		s z	s` z`	x G
>> 		K l
>> w				j
>> Now what I need to do is figure out how it got that way.
> Have you looked through your lexicon for lumpiness in the phoneme
> distributions?  Sounds which occur more than they should be "expected" to
> (probably the result of a merger), sounds which are nearly absent from
> certain environments (probably reflect an old conditional change), ...

It would be nice if I could find patterns like that, but since the 
existing Tirelat vocabulary hasn't actually gone through those 
historical changes that a real language would have done, probably the 
best I could hope for is a few vague hints. It's worth a try though. 
When I looked at tones in Simik, I found some interesting patterns in 
distribution that helped me to reconstruct a basic tone system for the 

> Conversely, to give one example, if /j w/ are frequent and you try to
> explain them away as arising from /i u/ in hiatus, you haven't actually
> explained very much yet, since a distributional oddity remains.  Why did the
> proto-language have so many hiatic clusters of vowels?  Most likely,
> whatever did form them also formed hiatic clusters without /i/ or /u/ in
> them; what happened to those?

I'm thinking some sort of vowel breaking may be involved, like what 
happened in Spanish with /ie/ and /ue/. One clue is the word "ŕjandi" 
/r̥jandi/, which I'm assuming came from an original *hrjandi or *ʔrjandi. 
A 3-consonant initial cluster is unusual for Tirelat (I can't find any 
examples of one), so a reconstruction as *hriandi or *ʔriandi seems more 
likely to me. Another possibility would be a vowel that was lost, 
*hrəjandi or *ʔrəjandi.

>> orig.	later	dial. A	dial. B
>> [vidu]	[viDu]	[viDu]	[viDu]
>> [nitu]	[nidu]	[niDu]	[ni4u]
>> [viru]	[viru]	[viru]	[vi4u]
> That looks to work.  But you can get by with less change and less dialectal
> divergence by doing [d]>  [4] ([>  r]) intervocalically in the second type,
> and after that [D]>  [d] in both.  This avoids the [d]>  [D] in the first.

Well, [D] is just the allophone of /d/ between vowels. One dialect does 
have a /ð/ phoneme, but it corresponds with /v/ in the standard dialect.

Maybe [t] > [4] directly in dialect B without going through [d] (English 
has something like that), then [d] could remain as [d] and dialect A 
would just have [t] > [d].

> AFAIK intervocalic voicing and frication generally apply to all places of
> articulation (even when they don't it's rarely just the dental place.
> Tapping at the dental place alone is common though).  For the initial shifts
> of *[t d]>  [d D] that's no problem; you get mergers into the voiced
> fricatives in other places.  But if your second posited [d]>  [D] in dialect
> A has parallels in other places, then you get rid of all the voiced stops.
> (Is that ok?)

There's a variation between /ɣ/ and /ɡ/ between dialects that may 
parallel the /d/ ~ /r/ variation (or may be unrelated). It's possible 
that Early Modern Tirelat only had voiceless fricatives (not an uncommon 
situation) and that the voiced fricatives of the modern language were 
originally voiced stops. But that would only explain voiced fricatives 
between vowels, not in other places (where voiced stops are still 
present in all dialects).

>> It's also possible that there may have been a voiced lateral fricative
>> /K\/ in EMT, to account for the /l/ ~ /d/ variations between dialects in
>> some words, but there may be other ways to account for that.
> Yeah.  Clusters that resolve differently in different dialects are one.
> Perhaps Proto-Tirelat allowed /dl/ etc. (does modern Tirelat?)

Modern Tirelat has [tɬ] and [dɮ] (spelled tł, dł in romanization), which 
may have been *tl and *dl originally. The devoicing of the l in *tl is a 
regular change (which happens also with *pl and *kl), but the absence of 
/dl/ is unexpected. (I'm wondering if the occurrences of "dł" might just 
be an error in documentation from all the changes that Tirelat has 

> Also, watch out for reconstructing èvery such variation into an original
> phonemic difference; you are liable to get an unrealistic number or
> distribution of proto-phonemes if you do.  If regular, Neogrammarian
> explanations fail, the alternations might just come from dialect mixing, or
> incomplete sound changes, or ...
> Alex