🚀 go-pugleaf

>>unusual.  *NO ONE* of my acquaintance does strictly binary comparisons, or
>>suggests that that is the way to do comparative studies.
>
>	I certainly agree there, but what G,R, and Sh are suggesting
>is that that's the way it works out in practice -- that a lot of
>comparisons of distant Native American languages have been done using
>the two-at-a-time approach, for instance. Or else comparisons of
>Indo-European with only Semitic or Afro-Asiatic or Uralic. Or at
>least, so they charge.

It's important to distinguish here (as Greenberg, at least, sometimes
doesn't) between comparison of two *languages*, which is seldom the best
strategy if there are other alternatives, and comparison of two
*families*.  Most of what long-range comparison has been done of New
World languages has been "binary" in the second sense -- e.g. Siouan
and Iroquoian, Uto-Aztecan and Kiowa-Tanoan, etc. -- or of an isolate
with a family, e.g. Yuchi and Siouan, Kutenai and Salish, etc.  This is,
more often than not, the right way to go about things.  Even if there
is merit in large groupings such as, for example, Greenberg's
Almosan-Keresiouan, there can be no doubt that such aggregations have
internal structure, i.e. that Siouan is more closely related to some
subunits of A-K than to others.  A comparison of Siouan and Iroquoian is
an attempt to explore the hypothesis that these are more closely related
to one another than either is to anything else, i.e. that they do
constitute a genetic unit at some level, and are therefore an
appropriate subject for comparative reconstruction.

Scott DeLancey