--- In cybalist@..., tgpedersen@... wrote:
> The only non-structured part is thus that of Manansala's list, and
of course my own (implicit) comparison between reconstructed PAA and
PIE forms with Manansala's heaps. Your description of my "use of
ahistorical collections of modern words (following Manansala, in the
case of Austronesian)" is thus, shall we say, incorrect.
However, considering the central importance of AN in your arguments,
the absence of a historical interpretation of the AN data is, shall
we say, a serious methodological deficiency.
> In the given case, we have
EIEC:
*H2eH-
(pres. *H2éHor) "burn, be hot"
*H2éHo:s
ha:s (acc. ha:ssan) "soda ash, potash; soap;
(pl.) ashes" Hittite
*H2eH-seH3 "hearth" [should be -s-eH2 -- Piotr]
Orël & Stol'bova
82:
*?es- "fire"
> Are you saying that if someone doesn't provide rules with which to
relate these two roots, then they are not related? And if someone
should claim that this might be a case of borrowing one way or the
other, and no date of borrowing is provided, then no borrowing has
occurred? Because that is what your words mean, as they stand.
Please note the epistemological difference between reality as such
and our _knowledge_ of reality. Until the rules are presented and a
cogent demonstration is put forward, there is no serious proposal to
discuss and we can only engage in premature and futile speculation.
As regards the possibility of borrowing, a convincing scenario must
be offered, including an answer to the questions "when", "how",
and "in which direction" -- otherwise there is no telling whether a
borrowing could have occurred or not.
Morphological analysis is important too. In the given case, we have
PIE *h2ah1- (pronounced [xah-] or similarly) with productively formed
derivatives, e.g. *h2ah1-(e)s-/*h2ah1-o:s -- an IE s-stem noun
(*h2ah1-s-ah2-, as in Latin ara, is a further derivative of that).
This *-s- is a stem-forming element whose function in IE is
understood quite clearly (it isn't just a vague and semantically
empty "extension"). It is not part of the root. On the other hand, *?
es- looks unanalysable in AA (I omit the question of the general
credibility of Orël & Stolbova's reconstructions). If we compare like
with like, i.e. a root with a root, the set of equations *h2 = *?,
*h1 = *s (no discussion provided) looks unconvincing by itself,
whether we are talking of common descent or borrowing. These PIE and
AA terms are in actuality neither more nor less similar to each other
than just about any pair of CVC morphemes.
One could speculate that a loan like *?es- could have become *xahs-
in PIE and that a truncated *xah- could then have arisen by
backformation. But why should that have happened, to ask the most
obvious question (formal difficulties apart)? Why borrow a "fire"
word and make it mean "ashes" if you already have your own "fire"
and "ash" terminology? The similarity is too vague to make such risky
and complex scenarios desirable. Borrowing from PIE into PAA is
hardly likely either, especially considering that PAA is generally
dated to a more distant past.
> Let me give you an example: everone and his brother agrees that
Greek "pyrgos" and German "Burg" etc are related and that they were
borrowed probably from AfroAsiatic. Probably, a date can be fixed on
to that loan too. But does that mean that if no one had been able to
put a date on the loan, if all we had was a similar-sounding word in
some AfroAsiatic language, we would be forced to accept a null
hypothesis that they were native IE words? I don't think so.
I'm apparently neither everyone nor everyone's brother, since I doubt
the equation "pyrgos = Burg", and find no ground for accepting the
theory that *bHerg^H- comes from Afroasiatic. A null hypothesis is by
definition what we assume to be the case until a compelling case has
been made for a less parsimonious competing proposal. *bHerg^H- is a
good PIE reconstruction, based on rigorous comparison involving
Hittite, Old Indic, Armenian, Germanic, Celtic, etc. I'm agnostic
about its remoter history. "Similar-sounding" is anything but a
secure criterion of relatedness, especially if you work with
ambitious time scales. A lay person (I mean everyone and his brother)
may be lured into finding such hackneyed comparisons as "day = dies"
or "deus = theos" self-evident; it takes a suspicious linguist with
his comparative tools to show that these equations are illusory.
> ... You might of course set up a (Neo-Grammarian?) rule of the game
that if roots can't be connected at once with a set of rules they are
not related, but the strict application of that game rule would,
given yours and Miguels admission that none of your rule sets cover
the whole reality of the development of Slavic, entail that Slavic is
not a branch of IE but (by application of a prudent rule of null-
hypotheses) language isolates like Basque and Burushaski.
Regular correspondences account for so much in the development of
Slavic that the inevitable messy residue is justly deemed to be
irrelevant to the question whether Slavic is IE. Its existence only
provides linguists with further puzzles to be solved, and as time
goes by we do solve some of them and gradually achieve a better
understanding of the whys and wherefores of the messy part. The fact
that 100% rigid regularity can be neither attained nor even expected
(since "mess generators" are built into the universal mechanism of
language) doesn't mean that the requirement of regularity has been
compromised, or that "no method is perfect, so anything goes". So
far, the regularity principle has proved to be the most effective
heuristic tool in historical linguistics and constitutes the backbone
of any serious relationship proposal.
Piotr