Re: Stress and vowel length in Tirelat
|From:||Herman Miller <hmiller@...>|
|Date:||Saturday, August 16, 2008, 22:12|
David J. Peterson wrote:
> Herman Miller:
> I still haven't found any clear cases of vowel length being distinctive.
> Oh, you can figure it out. The hard part will be getting a set of live
> native speakers of Tirelat. Once that's done, though, you'll be able
> to cook something up.
> One of the most likely examples, _marat_ "window" vs. _maraat_ "basket",
> could alternatively be treated as a distinction in stress: _márat_ vs.
> So here's what you do. Run an experiment that compares all
> of these:
> 1.) [ma:.'rat]
> 2.) ['ma:.rat]
> 3.) ['ma.rat]
> 4.) [ma.'rat]
> 5.) [ma.'ra:t]
> 6.) ['ma.ra:t]
I'll have to try that out the next time I run across an Impossible Gate
to the Sangari homeworld. My guess is that (2) and (3) will be heard as
"window", (4) and (5) as "basket". The question is what happens when the
clues of length and stress are conflicting, as in (1) and (6).
> My honest guess is that it's vowel length. You don't necessarily
> need minimal pairs to determine this. Take some of your sample
> CV(:)CV(:) words:
> -/mutaa/ "no one"
> -/nuka/ "to return"
> -/riiva/ "sky"
> I don't have stress on these, but unless you have a CVCV words
> contrasting in stress, then it seems like what you have here is a
> language with long and short vowels, and stress that gets attracted
> to heavy things. Think about the speakers, for example. It doesn't
> matter if you don't have /muuta/ = "x" and /muta/ = "y". You
> have *plenty* of examples of the following:
The absence of CV:CV: seems to need an explanation, but it could be a
shortening of vowels in unstressed syllables. Many of the CV:CV in the
current language can be traced back to words that were CVCV in older
versions of Tirelat, so the inconsistent lengthening may be little more
than inconsistent documentation.
I agree that it appears to be length rather than stress that's
significant in the current language. But part of that is due to
lengthening vowels deliberately to get the stress patterns that I
wanted. Stress and vowel length were unmarked in the oldest versions of
the language, but I began to distinguish them in later revisions. For
ba'kaazi "monitor lizard", originally bakazi
vs. 'səlagi "cod", originally solagi
li'naar "dragon", originally linar
vs. 'pivri "welcome", originally pivri
With the three-syllable words, 'CVCVCV is common; CV'CV:CV and CVCV'CV:
are less common, but also exist.
> [Note: There is another option. Perhaps in time x-1, there were
> no long vowels at all in Tirelat. Words of the CVCV: shape,
> then, would be languages where stress was originally on the
> final syllable, and the vowel was lengthened as a result. In order
> for this to work, though, several anomalies would have to be
> explained. First, it wouldn't make sense for there to be CVCV
> forms *and* CV:CV forms, unless there were some sort of
> consonant loss in all CV:CV forms. Further, if all vowels before
> word-final consonants were lengthened, that would be one
> thing. But while we have /linaar/ and /kavaal/, we also have
> /laghal/. Taken altogether, I think the explanation would become
> far more complex than if Tirelat simply possessed long vowels
Some, but not all, of the originally short vowels were lengthened when I
started treating vowel length as phonemic. E.g., the vowels in "gira"
(island) remained short, but the vowel in the initial syllable of "šira"
(weasel) was lengthened. It could be, though, that I was trying to hear
a distinction where there was none to begin with.
I suspect there are elements of both stress accent and vowel length in
this distinction, and that one or the other may be more prominent in
different circumstances. I've been listening to my old sample of Tirelat
speech from Relay 6, which is pretty much similar to the current version
of the language except for the gender agreement. There do seem to be
distinctions in vowel length, but they seem to also be associated with
stress or intonation.