On Fri, 7 May 2010 09:17:34 +0100, R A Brown wrote:

> Jörg Rhiemeier wrote:
> [snip]
> > As you may remember, Old Albic has both short and long vowels,
> > with all possible combinations of the features [+open], [+front]
> > and [+round], except the null combination which occurs in some
> > affixes but is realized by adopting the features of the nearest
> > vowel.  Thus, the basic vowel qualities are /a e i o ø u y/:
> > 
> >         a  e  i  o  ø  u  y
> > [open]  +  +  -  +  +  -  -
> > [front] -  +  +  -  +  -  +
> > [round] -  -  -  +  +  +  +
> OK - not an uncommon set of vowels.

Indeed not.  I don't think an unusual inventory is the best way to
make a phonology interesting; it is much more interesting what one
*does* with the inventory.  I have seen languages which have odd
phoneme inventories, but do nothing of interest with the phonemes.

> > The first new finding is that long vowels (which occur only in
> > stressed open syllables, having been shortened in all other
> > environments) have two different intonation contours: the
> > *thrusting tone*, which has one peak of stress, and the *slipping
> > tone', which has two peaks.  The thrusting tone is transcribed
> > with an acute accent (á é í ó ǿ ú ý), the slipping tone with a
> > circumflex accent (â ê î ô ø̂ û ŷ); the Old Albic script uses two
> > different length markers.  
> Seems to be some problem with ø-circumflex. Indeed, I cannot 
> find any Unicode representation of this; as far as I can 
> tell on a quick investigation, Unicode has symbols for ø and 
>   ǿ only (with corresponding upper-case forms also). Is this 
> correct?

Yes.  The ø-circumflex uses a combining diacritic (U+0302), which
may cause some problems.

> I assume, both by the use of the acute diacritic and by the 
> use of the term "thrusting", that this is a rising tone. 
> The description of the "slipping tone" sounds like the 
> Mandarin 3rd tone. Is this right?

Not really.  Old Albic is a stress-accent language.  The intonations
of the long vowels are comparable to contrasts found in some local
dialects of German in pairs such as the last two words in the sentence

Sie liebte den Seher sehr.

In some German dialects, the last two words differ only in the kind
of stress.  In German, these intonations are called _Stoßton_ and
_Schleifton_; the terms "thrusting tone" and "slipping tone" are my
own attempts at translating the German terms.  In the sentence above,
_Seher_ has Schleifon, and _sehr_ has Stoßton.  Stoßton is a simple
stress accent on the first mora of the vowel, while Schleifton is a
somewhat weaker accent on the firt mora and a slight secondary accent
on the second mora.  Schleifton occurs where two vowels were contracted
into one, in German as well as in Old Albic.

> I suppose the long thrusting ones could be regarded as 
> having the tense vowel throughout while the slipping ones 
> shift from tense to lax, i.e. you could, for example, use 
> the representation:
> Tr. IPA  Tr. IPA   Tr. IPA
> a   [ɐ]  á   [aa]  â   [aɐ]
> etc.


> Can we think in terms of short vowels being monomoraic and 
> long ones bimoraic? How does the double peaking on the 
> slipping tone work?

Yes, long vowels are bimoraic.  As I said above, slipping tone
involves a main stress on the first (tense) part of the vowel
and a weak secondary stress on the lax part.  Slipping-tone
vowels are just on the border towards being bisyllabic, though
they count as monosyllabic (but, like thrusting-tone vowels,

> > As long vowels occur only in stressed open syllables and are
> > shortened in closed and unstressed syllables, some vowel length
> > alternations occur in the language.  
> ..which means that if it were not for the two different 
> sorts of long vowel, vowel length would not be phonemic. But 
> there is, so we must give them some sort of phonemic status.

No, a stressed vowel can also be short.  In fact, most Old Albic
words do not contain long vowels at all!

> > For instance, the locative
> > case of _mor_ 'shadow' is _môrol_: the underlying stem has a
> > slipping-tone long vowel which is shortened in the base form
> > because the syllable is closed. In the locative case, the /r/
> > following the vowel belongs to the second syllable, and the
> > long vowel, being stressed, is not shortened.
> Yes, I follow this. But as one who has got used to things 
> like _môr_ /mo:r/ "sea" in Welsh, and the monosyllable with 
> circumflexed vowels in Sindarin, this goes against the grain 
> for me.  While I go along with vowel shortening in closed 
> syllables in polysyllabic words and, indeed, in 
> monosyllables where the syllable is blocked by two or more 
> consonants, the shortening in monosyllables before a single 
> consonant might be reconsidered?    ;)

It may be reconsidered.  I am not really sure yet about when exactly
long vowels are shortened.  Old Albic allows for two consonants in
a coda, even though most codas have at most one (and many syllables
are open).  Hence, there is no reason to diasllow long vowels in
closed syllables.  But unstressed long vowels feel awkward to me
here, so I'll probably change the rule such that only unstressed
long vowels are shortened.  Long vowels in monosyllables would thus
not be shortened.

> > [vowel positioning rules]
> All that these six rules seem to me to do is to predict 
> _where_ the vowel will occur, but not what the vowel will 
> be.  It means that speakers of the language would probably 
> manage quite well with a script that indicated only 
> consonants.  But although one unfamiliar with the language 
> would be able to work out where the unwritten vowels should 
> be, I don't see any way s/he could work out what the vowel 
> was.  Or have I missed something?

The rules indeed only specify *where* the vowels are.  The qualities
of the vowels are determined by vowel features that bind to the
morphemes.  All lexical roots have at least one vowel feature bound
to them.  Some affixes have no feature bound to them; in these cases,
the features of the root (or whatever morpheme with vowel features
is nearest) propagate to the affix.  If a morpheme has only one
feature attached (which is the case with most morphemes), that
feature spreads to the preceding morpheme (causing umlaut).

Consider a simple example.  The word _bacas_ 'bread' consists of the
consonants b-c-s; the rules above insert vowel positions like so:
b°c°s.  The root carries the vowel feature [+open], hence, the word
manifests as _bacas_.  (This is an example for the fact that in a
bisyllabic root, both vowels are always the same.)  Now, if we add
the plural suffix -im to it, the following happens: the plural suffix
consists of one consonant: m; which means that the vowel position is
before that: °m.  The suffix carries the feature [+front], hence the
surface form of the suffix is -im.  Now, the vowel feature spreads to
the preceding morpheme, so we get the form _becesim_:

 [+open]  [+front]   [+open][+front]
    |         |          |  / |
  b°c°s      -°m   ->  b°c°s-°m     -> becesim

> > [irregularities and lost consonants]
> No problem - such things happen in natlangs.

Indeed.  Old Albic would be less realistic if there weren't such

> > This means that in a way, the only segmental phonemes of Old Albic
> > are the *consonants* (including the mute consonant *3).  Yet, I would
> > not call Old Albic a "vowel-less" language; 
> Nor would I    :)
> > the vowel features, even
> > if they are autosegments rather than segments, are real, and realized
> > as surface vowels.
> Of course they are.  From what I have gleaned about the 
> autosegmental phonology theory, it appears to postulate more 
> than one tier of phonological represents.  According to SIL 
> "Each tier is made up of a linear arrangement of segments. 
> The tiers are linked to each other by association lines that 
> indicate how the segments on each tier are to be pronounced 
> at the same time." If such an analysis is applicable to Old 
> Albic, doesn't it mean that the consonant phonemes are on 
> one tier and the vowels on another tier? On the 'consonant 
> tier' the _position_ of the vowels are unambiguously 
> defined; on the 'vowel tier' the vowels which fill those 
> positions are defined.

Exactly that.  On one tier, there are the consonants; on another tier,
there are the vowels.  The vowel-positioning rules determine where the
vowels are inserted into the consonant string.

> Clearly Old Albic is not a vowel-less language.  The 
> analysis being proposed seems to be putting consonant and 
> vowel phonemes on different tiers, but not removing vowel 
> phonemes a such.  Or have I entirely missed something 
> somewhere along the line? (I'm not a "razor-witted 
> phonologist"!)

No, you haven't missed anything.  There is no way to predict vowel
features from the string of consonants.  There are minimal pairs
for the vowel qualities, such as _cas_ 'box' vs. _cis_ 'gravel'.

> SIL, however, defines 'autosegmental phonology' thus; 
> "Autosegmental phonology is a non-linear approach to 
> phonology that allows phonological processes, such as tone 
> and vowel harmony, to be independent of and extend beyond 
> individual consonants and vowels."
> This looks to me remarkably similar to the 'Prosodic' 
> approach of J.R. Firth in the 1950s.  If this so, I'm not 
> really clear how this does actually apply to Old Albic.

The mechanisms of umlaut (before /a/, /i/ and /u/) and vowel harmony
(in affixes without vowel features) at least involve vowel features
spreading from morpheme to morpheme.  I am not too firmly saddled in
the terminology of phonological theories; in my usage of the term
"autosegmental", I was inspired by a phonology textbook which
introduced autosegmental phonology as a means to deal with the
behaviour of tones in African languages.

... brought to you by the Weeping Elf