How new hypotheses grow

[Larry Trask]

On Tue, 24 Aug 1999 ECOLING at aol.com wrote:

> The question arises then, why is it that the best historical
> linguists of American Indian languages were quite thoroughly
> convinced AT SOME POINT that these languages were related,
> yet the standard kind of evidence was somehow lacking?
> What was it that they found so thoroughly convincing?

A very good question.  In reply, I recommend chapter 2 of the following
book:

Lyle Campbell (1997), American Indian Languages: The Historical
Linguistics of Native America, Oxford UP.

Briefly, Campbell sums up what he calls the "reductionist frenzy" as
follows.

Several (but certainly not most) early Americanists were interested in
finding larger groupings than had been demonstrated.  One of these was
Sapir, the most brilliant and admired American linguist of his day.
According to Sapir's student Mary Haas, Sapir himself regarded his
breathtaking mega-agglomerations as no more than hypotheses to be
investigated, but he nonetheless published them in a way which led
readers to assume he was fully committed to his proposals.  Partly
because of Sapir's eminence, and partly just because his (and other)
proposed groupings appeared in print so often, these groupings came to
be reified.  That is, readers assumed that things like `Penutian' and
`Hokan-Siouan' must be real just because they were mentioned all the
time in the books.

But not all linguists were taken in.  For example, Campbell quotes the
distinguished A. L. Kroeber as follows (p. 75): "Tremendous havoc can be
wreaked when archaeologists or ethologists begin to buil[d] structures
of inference on Sapir's brilliant but flimsy gossamer web of prophecies
as if it were a solid foundation."

In short, the proposed mega-families came to be widely recognized as a
result of authority (Sapir's eminence) and repetition (constant
publication) -- even though there was no hard evidence to back them up.

As a result, Campbell laments, Americanists have been obliged to devote
a great deal of time to disassembling the proposed super-families and to
searching out whatever evidence might exist for grouping *any* of the
languages assigned to them.

This is a salutory lesson in the dangers of reification: because we have
invented a name, we persuade ourselves all too easily that there must be
something "out there" for the name to apply to.

Something similar has happened in my own field.  A few decades ago, a
number of linguists were pursuing the idea of a genetic link between
Basque and the Caucasian languages.  One enthusiast -- I think it was
Martinet, but I'm not sure -- coined the term `Euskaro-Caucasian' for
this putative family.  As a result, the term `Euskaro-Caucasian' crept
into the literature, and it appears in at least one major reference
work.  And yet there is *no* persuasive evidence that Basque is related
to *any* of the Caucasian families, or even that these last are all
related to one another.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk