Hallo conlangers!

On Friday 12 April 2013 20:45:17 BPJ wrote:

> 2013-04-12 02:36, Alex Fink skrev:
> > On Thu, 11 Apr 2013 22:57:00 +0200, Jörg Rhiemeier 
<[log in to unmask]> wrote:
> >> Hallo conlangers!
> [...]
> >> But the impasse with the "Mitian" problem (eight or
> >> perhaps even more stocks in northern Eurasia looking
> >> like relatives, and yet all attempts at proving
> >> their relationship have been inconclusive so far)
> >> may mean that the family tree model has its
> >> problems.
> > 
> > The family tree model may have its problems, but they
> > shouldn't necessarily receive the blame for this.

Indeed not.  I still feel that these language families indeed
form a single genealogical unit.  Look, we have eight stocks
occupying a single contiguous area, with similarities (most
conspicuously the pronoun roots) that are hard to explain
otherwise.  This does not look like a pattern that could have
arisen by chance.  Borrowing also doesn't look very likely
(though I have to admit that it cannot safely be excluded).
This makes a single common ancestor IMHO the most likely
explanation.

Add to this that the area where these stocks originated is
a huge spread zone in Nichols's terms, where people have been
on the move on a large scale during much of known history,
and it is quite plausible that a single lineage expanded,
perhaps from somewhere near Lake Baykal, about inner Eurasia
at the end of the last ice age.

Alas, the time depth presumed in this scenario - about 12,000
years - is so great that the comparative method reaches its
limits.

> >      To
> > make a point I've made here before, a big part of the
> > problem is that people don't appreciate that the time
> > depths of reconstructible branchings usually get
> > further and further apart the further back in time
> > you go, and thus the proportion of common retentions
> > gets smaller and smaller even across each individual
> > step as you go back.

Sure.  If you look at Proto-Indo-European and Proto-Uralic,
for instance, they don't look as similar to each other as,
for example, Latin and Ancient Greek do; it is more like,
two modern IE languages from different branches, if not
worse.  So you get about *twice* the time depth of IE and
Uralic for a putative "Indo-Uralic", if not more.

Another difficulty, of course, is that most of the relevant
families lack ancient written records.  You have to start
with the *modern* languages.  Also, some of them are really
shallow.  Proto-Turkic was spoken only about 2,000 years
ago; Proto-Yukaghir is probably only about half that age.

> >      Indeed, if you assume that
> > you're not dealing with agressive expansionism, of
> > the sort that as far as I can imagine only very-
> > dominant new military technologies or being the very
> > first settlers of the land in question would
> > consistently give you, the time depths should be
> > expected to get *exponentially* larger. The
> > comparative method just runs out of juice.
> 
> Agreed. My main beef with the family tree model is that
> *language contact exists* and that anything can be
> borrowed, which doesn't invalidate the model but must
> be taken into account. I guess that the family tree
> model, the wave model and the punctuated equilibrium
> model all have something to contribute. None of them is
> the whole truth, and none of them is wholly untrue.

Right.  Before a language breaks up, innovations spread
through it in various directions, and you get a pattern
of intersecting isoglosses as can be seen on the dialect
map of any of the major languages of Europe, for instance.
The wave model.  But contact phenomena such as borrowing
or convergence towards a similar phonology or grammatical
typology happens even between *unrelated* languages, there
are numerous examples for that.  

Certainly, some of the similarities we can observe in the
Mitian complex result from language contact.  Perhaps in
an age when the eight protolanguages huddled closer to
each other than in later times.  Of course, Iceland is a
far cry from Greenland when you have to go via Europe and
Siberia, but Ukraine and Chukotka are only about half as
far apart (still, it is a very large distance), and if
you turn back the clock another 5,000 years, the lineages
that would eventually become PIE and Proto-Eskimo-Aleut
may have been only a few hundred kilometers away from
each other somewhere in southern Siberia.

(Another matter is, of course, subgrouping.  Is Altaic a
valid node within Mitian?  Or Indo-Uralic?  Or Uralic-
Yukaghir?  Etc.)

> > [One version of a thought experiment to justify this
> > "exponential" stuff. Since we're assuming no
> > aggressive expansionism, the homeland of the common
> > ancestor of a family of languages should have been
> > around the center of the region its descendants
> > occupy. Now, imagine you've got some big family. As
> > you go back in time one, two, three, four, ...
> > branchings -- suppose they're binary branchings --
> > then you're looking at common ancestors of 2, 4, 8,
> > 16, ... languages. The territories covered by the
> > families of descendants are getting exponentially
> > larger; so the ancestral homelands are getting spaced
> > exponentially further apart; so since languages move
> > at a constant rate,

Not really, it fluctuates; but around a mean that probably *is*
constant.

> >     the time for each branching to
> > happen is getting exponentially larger.]
> 
> Imagine English dies out everywhere but New Zealand and
> then spreads from there again, then try to reconstruct
> the pattern of English loans in various languages after
> 1K years. Makes the whole homeland business moot in
> both the British and American sense!

A nice thought experiment.

--
... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html
"Bêsel asa Éam, a Éam atha cvanthal a cvanth atha Éamal." - SiM 1:1