Tailing off with depth

[ECOLING at aol.com]

I fail to recognize John McLaughlin's response
(see quotes below)
as having any bearing on the challenge I issued.

His response does not refer to a tailing-off in KNOWN cases
with increasing depth.
It refers only to a comparison with Nostratic,
which is an unknown case,
using some method of judging "chance" which
is unspecified.  Such judgements in the past have been
notorious as building in as assumptions the conclusions
they come to.  Most circularly, they ASSUME certain
languages are unrelated which may ultimately not be,
or use a looseness of estimating "lookalikes" worthy of the
worst caricatures by critics of Greenberg.

And just incidentally, I note once again that
Multilateral Comparison has some strengths in this regard
which are still not correctly represented or even understood.
If we take Nostratic as the result not of study of a fixed set
of language families chosen A Priori, but rather as a set of
language families selected by Multilateral Comparison
from among the language families of the world,
then the hypothesis of Nostratic is merely that these language
families appear (by their data, not their geography)
to be likely to be more closely related to each other
than to other language families outside that set.
And the geography would then be a confirmation,
not a fact included in the process of selecting the set
of language families.

Of course, in practice, Nostratic was set up
in an informal process, partly in both ways.
As most pioneer discovery processes quite sensibly use
all information available, including geography,
so there is no "independent"
data set available at the beginning to act as a test.

Neither McLaughlin nor Trask have really taken up
my challenges on getting EMPIRICAL data on the rate of
tailing off of information usable for reconstruction.
I think the reason for this in the field as a whole is that
once the results of the Comparative Method are known,
no one is interested any longer in testing the tools that
got us there.  That is throwing away an enormous part
of our most valuable data, the data on how our tools
work in cases where we KNOW the answer.
This is an empirical question.

In the case of Albanian, Trask refers to what may be
an excess of loanword etc. influences from neighbors,
so that it took linguists using the Comparative Method
enormous effort to conclude that there was a stratum
of vocabulary inherited from PIE via a specific set
of sound changes peculiar to Albanian,
rather than borrowed from IE language neighbors.
[I believe that is what Trask was saying.]

I do not think I have ever seen a discussion of how
Multilateral Comparison might handle situations of
massive loanwords, or might be enhanced to handle
such situations better, short of the full effort of the
Comparative Method referred to above.

Naive Multilateral Comparison in THIS instance might
simply yield the result that the relations of Albanian
seem to be all over the map, sometimes with one
IE family, sometimes with another, in a fashion which
does not suggest subgrouping classification.

Would that be sort of like Germanic
between Slavic and other IE,
but Albanian in a more multipolar way?

These are merely decent guesses.
I still maintain we always need to study explicitly
how each of our tools works.  Albanian may be
a very important test case of a particular kind
for extending our tools to greater time depths.

Best wishes,
Lloyd Anderson

[LA]
> Do it for Indo-European, for goodness sake,
> where we AGREE that the languages all belong
> to one family, and yet MUCH vocabulary
> is NOT represented throughout the family.
> That does not cause us to doubt the reality of IE.
> When pushed by the data pattern, linguists discuss
> dialect chains or even networks or areal subparts of IE.
> So how much of this should we expect for different
> depths of relationship?
> Get the statistics on this, formulated in a simple
> and objective way which can be compared with
> both other known and unknown cases.
> across different levels of depth of dialects,
> families, family groupings and super-families of IE,
> see whether the rate of representation tails off in
> a linear, geometric, or other pattern with increasing
> depth, and what the range of variation is for different
> instances of the "same" time depth (which might
> depend on different social situations, such as
> the relative isolation of Icelandic, vs. Scandinavian,
> vs. mainland Germanic from other closely related
> languages of their family).

[JM]
Donald Ringe has already done something like this comparing Indo-European
with Nostratic ('Nostratic' and the Factor of Chance, Diachronica 12:1.55-74
(1995)).  He found that when looking at the number of subgroups represented
in cognate sets versus what would be predicted from chance resemblances,
Indo-European showed a much-greater-than-chance number of sets with cognates
in multiple subgroups and Nostratic showed the number of sets expected from
chance.