Richard:
> > That's yet another aspect to the issue. A similar case is retroflexes
in
> > the Indic (i.e. indo-Aryan) languages. <snip>
> > Scanning the literature, one could be tempted to suspect
> > there's
> > something in the water that causes retroflexes :-)
>
Mark:
That is another problem. I see substratum. It occurs in India along with
Dravidian speakers.
Richard:
That's the traditional view. The problem is that the oldest Sanskrit shows
very little in the way of Dravidian loans. In David MacAlpine's
Elamo-Dravidian hypothesis, it was proposed that Dravidian retroflexes arose
from r + dental, much as in Swedish or one Yorkshire (?) dialect of English
and as supplemented in later Indic languages.
> > The issue of non-Germanic cognates of 'bow' and 'bow' (both the
> > homonyms) is
> > complicated. The Germanic forms point to PIE *bHeugH-, but the other
> > languages (Greek, Latin and Sanskrit, at least - I haven't checked
Pokorny
> > for other languages) point to PIE *bHeug.
Mark:
> There are thrre roots *bheug. But there is a real word bUk in Turkic and
> it means 'bend'. And there is
> a real word el (hand), elig (hand), bilek (wrist, arm), and pilek
> (five). Why can't *bheug simply be
> from Hunnish. Weren't germans in contact with them? Werent' the Celts in
> that region?
Richard:
*bHeug 'flee' is evidenced by Latin fugio: 'flee' and Greek pheugo: 'flee',
both real words. What's more, Old English bu:gan (whence English 'bow' as
in 'bow down'), besides meaning 'bend', also occasionally meant 'flee',
Probably two homophones, but maybe our ancestors saw some connections
between the words.
I think it's just coincidence, but of course the Huns also invaded India,
and we have Sanskrit bhuj 'bend, bow'. On-line Pokorny (Leiden University)
seems to be off-line, so I can't check where the *bHeug roots manifested
themselves. I fell however that these words must predate the Hunnish
invasions. Hengist and Horsa were in Kent in 449, and Attila the Hun was
born in 406 and died in 453.
Having said all this, a relationship between PIE *bHeug and Turkic bük looks
plausible for the consonantal elements; I can't comment on the vocalic
elements.
[On the 'el-' of 'elbow':]
> > The etymology is complex, but:
> > (a) Words for parts of the limbs seen to change their reference easily.
> > (b) Ablaut seems to have left PIE full of surface irregularities,
> > which have
> > been resolved differently in the daughter languages. A modern
> > parallel is
> > Polish.
> >
> > In short, I see no significant problem with the normal etymology.
Mark:
> In how many branches of IE does it show up?
Richard:
Lots. Celtic (Irish, Welsh), Germanic (English, Old Norse, German and
probably Gothic), Latin, Greek and Tocharian all show forms that can be
referred to *h1eh3len-. Forms with a -(V)k suffix instead of -en appear in
Latvian, Lithuanian and the Slavonic languages. It might just possibly be a
different word. It may also occur in Indo-Iranian, but it may also reflect
another root. There's an informative thread starting at
http://groups.yahoo.com/group/cybalist/message/23716 .
Mark:
> What is wrong with Greek/Latin ul-, like English being related to
> Akkadian QATUM,
> Turkic kol (arm), all going back to *qathum? Much simpler. And this is the
> NOSTRATIC LIST.
Richard:
Well, it wouldn't relate to Turkish el 'hand' or Chuvash pelik 'five' then,
would it?
I'm leaving objections to those more knowledgeable in Nostratic matters.
Incidentally, what's Greek ul-?
Mark:
> I give you a set of integers: {1,2,2,3}. They came from one of them.
> What do you do? Most linguistics
> books are silent as if it is magic.
>
> 1. Average of some kind: e.g. pick *2
> 2. Mode: *2
> 3. Median: *2
> 4.Majority vote: *2
> 5. assume increase: *1, therefore *1>2, and *1>3 or *1>2>3
> 6. assume decrease: *3, therefore *3>2, etc.
> So when I see something like :"language X has {p,p,f,f,f} so obviously
> *p" I ask (and I used to
> ask loudly, to linguists on mailing lists and sci.lang" "what is the
> f*cking rule?"
Richard:
I've heard that being polite helps.
The immediate response is that f > p is much than (conditional) p > f.
Unconditional f > ph is not unknown - Ahom, Shan and, I think, Armenian
spring to mind. For Ahom it appears to be an alignment with the Indic
L-language, Assamese. Why it extends to Shan, which is closely related, I
don't know.
Mark:
> No book ever has any rule, or any explicit algorithm. I think I read
> every one.
Richard:
The reasons for a language undergoing one sound change rather than another
are still a mystery. It has been theorised that sound changes will tend to
make the phoneme systems more 'regular', but irregularities can happily
sustain themselves for centuries. Isn't the asymmetry in the Turkish vowel
system an example? In principle vowels can be classified by three
features - [+/-]high, [+/-]back and [+/-] round, but [a] is lower than all
the other vowels.
> Suppose now we have 2 language families {1,2,2,3}, and {2,3,3,4} (e.g.
> IE and AA),. Suppose
> we reconstruct PIE and PAA using any of rules 1-4, we get 2 and 3 and
> then using one of them
> have to choose 2.5 or 2, or 3 or something. But suppose we look at the
> raw data e..g. {1,2,2,2,3,3,4}
> Now it looks like we should select *2.
>
> But what if we had some reason for selecting rule 5 or 6? What rules can
> they be?
>
> That is what I am working on and those are the parts that I am posting
here.
Well, the rules in deducing the changes are to prefer 'naturalness'.
However, either that concept is not fully understood, or is only a
probablistic rule. One can get some pretty weird changes. See Robert
Blust's article on the Austronesian languages in the collection Philip Baldi
edited on langauge change in the Austronesian languages. There are two
versions - the full volume, aimed I presume at university libraries, and a
student edition. Robert Blust's article is in both versions.
Naturalness can only get you so far. The continuing arguments over the
phonetic forms of the PIE stops are evidence of that. It's difficult to
computerise a technique when the practitioners disagree. You can aim to
minimise the weirdness of the rules for the sound changes given by equally
explanatory sets of rules starting from isomorphic proto-language
definitions, but different metrics give different results. (For example, is
linear regression always the best approach?) You should also consider the
naturalness of the starting point.
How is you automation of theoretical physics progressing?
>
>
> >
> >
> > Mark:
> > For example, it is said that Altaic had an initial-p that changed to a
> > bilabial fricative and disappeared
> > but apparently along the way it also became h in some places. One of the
> > classics is Doerfer's *pOkUrz (ox)
> > from which he gets OkUz, OkUr, hOkur, hOkUz etc. Now it so happens
> > that this
> > word looks too much
> > like pecus (IE cattle) to be an accident. So why cannot the same thing
> > happen to *parsh? *pash? And
> > what if it had an even earlier form which could have given rise to eat.
> >
> > Richard:
> > The biggest problem I can see in relating Doerfer's *pOkUrz and PIE
> > *pek^u-
> > is that the <s> of Latin pecus is not part of the root. Germanic and
> > Indo-Iranian show only a stem in peku-. Latin has (citing just the
> > forms in
> > my pocket dictionary):
> >
> > 1. pecu: 'flock of sheep', stem pecu-, neuter.
> > 2. pecus 'cattle, herd, flock; animal', stem pecor-, neuter.
> > 3. pecus 'sheep, head of cattle, beast', stem pecud-, feminine.
> >
> > Only no. 2 has the right stem. Note that the final consonant has
> > developed
> > from /s/, with the nominative and accusative singular retaining /s/
> > becuase
> > it was not followed by a vowel.
Mark:
> Look at pecor and pecud. * pecudh. From *dh I also derive both r and z.
> Now you have actually buttered my bread.
Richard:
I strongly suspect the -d- and -r- (possibly once -s-) extensions are purely
Latin innovations. They are not unreasonable extensions from an animate
nominate singular 'pecus', which occurs in Latin, although the 4th
declension noun is neuter. As Pokorny is down, I can't double check the
claims, alas.
How do you explain the multiple development in the same language? Any
similar examples of /d/ and /r/ alternating? (Perversely, we do have
several example of /l/ in Latin where we would expect /d/.)
Richard:
> In any case, the point is that we have to go beyond IE since this is
> Nostratic. And now I think I explained why I do what I do.
>
>
> >
> > Incidentally, a loan of a pre-PIE animate nominative singular *pakuz to
> > Altaic might appeal to some people, but I don't think the timing is
right.
>
Mark:
> I said Doerfer did *pOkUrz not me.
>
> How does one reconstruct sounds?
>
> One way is to put both sounds in it e..g you want r and z, so you
> assume rz. Another is to select
> something like a centroid (e.g. average of some kind). I select *dh for
> r, z, d etc. But there could be
> some other directionality involved. These are complicated. I have
> reasons for selecting what I
> selected. I am not finished with the writing because I have so much to do.
Richard:
You start by assuming that sound changes are overwhelmingly regular. If by
'dh' you mean a voiced fricative you have precedent for dh > z, dh > d in
Semitic, e.g Hebrew v. Aramaic, and possibly in Arabic dialects. However,
having both in one language is unusual.
Mark:
> What I think I will do over the next 3 weeks is try to put together the
> Akkado-Turkic cognates
> where Akkadian has lost consonants still retained in Turkic.
>
> I have a question for everyone.
>
> Where can I publish such a paper?
Richard:
You could try 'Mother Tongue' - see
http://www.people.fas.harvard.edu/~witzel/aslip.html . You may have to
persuade them you're not someone who believes that all languages descend
from Turkish.
Richard.