Campbell's def. of "language"

[Larry Trask]

Lloyd Anderson writes:

>  Yesterday I received the latest issue of the Journal "Language",
>  in which there is a review by Doris Bartholomew of
>  Lyle Campbell's book
>  *American Indian Languages: The historical linguistics of Native America*
>  Oxford University Press 1997

>  Campbell defines these terms (and reviewer Bartholomew
>       does not comment, so presumably regards these definitions
>       as quite normal, unremarkable):

>  'dialect' is a variety (regional or social) of a language,
>       mutually intelligible with other dialects

>  'language' is a distinct linguistic entity that is
>       mutually unintelligible with other languages

>  I simply do not believe that Larry Trask is unaware that this is
>       standard linguistic usage.  He is not forced to use it himself,
>       but he is obligated to treat this usage with respect,
>       not to evade discussions based upon it by ridiculing it.

OK; let's talk about this.

First, Lyle Campbell is a friend of mine.  I have the greatest respect for
his work.  I have read his book on American languages from cover to cover,
and I enjoyed it greatly.  It's the perfect work for a non-specialist who
wants an overview of the state of play in American historical and
comparative linguistics.

Now, in his first chapter, Campbell points out a number of the difficulties
he faced in compiling his book.  Like other parts of the world, the Americas
contain many instances of related but distinct language varieties, some of
them closely related, others more distantly related.  Mutual
comprehensibility between related varieties may range, as usual, from 100%
to zero.  Also as usual, specialists have often arrived at very different
classifications of the resulting position.  Just to cite one example, the
large and messy Zapotecan complex has variously been counted at anything
between six and 55 distinct languages (p. 158).  Campbell reports (pp. 3-4)
that specialists have counted the American languages at anything from a
minimum of 400 to a maximum of over 2500.

Now, for his purposes in writing this book, he needs to adopt a policy, and
his policy is spelled out on pp. 7-8.  These are the definitions quoted
above.

Now, it is *perfectly clear* from the context that Campbell is spelling out
a particular policy devised for a particular purpose.  Nowhere does he give
the slightest hint that his definitions are intended to be universal ones.
Nowhere does he suggest that other definitions, other criteria or other
classifications are inappropriate, misguided or wrong.  And *nowhere* does
he assert or imply that his choices are "standard linguistic usage".
Quite the contrary, in fact.  On p. 7, he tells us this: "It is important to
clarify this terminology and to specify how such terms are used in this book
at the outset."

Note the wording: "in this book".

It is simply that Campbell has to do *something* in order to get on with his
book.  And, quite properly, he is explaining his choices.

Also on p. 7, Campbell confesses that mutual intelligibility may be
"difficult...to define or apply in practice".  He says no more about this,
and we are left to conclude that Campbell is relying on his own or his
colleagues' judgements of mutual comprehensibility.

All of this is, or should be, perfectly obvious to any reader of the book.
But Lloyd Anderson -- who apparently has not read the book -- is extracting
bits and pieces badly out of context, declaring Campbell's policies here to
be "standard linguistic usage", and accusing me of "ridiculing" it.

Not guilty.  I've read the book.  Lloyd apparently hasn't, and he should not
be quoting bits of it out of context in such a way.  Nor have I ever
ridiculed the kind of policy adopted here by Campbell.

There's more.  Linguists commonly invoke a number of criteria in deciding
how many languages to recognize and where to put the boundaries.  Only when
no other criteria are available -- written traditions, standard forms,
political boundaries, or whatever -- do we fall back on mutual
intelligibility as the last gasp.  And mutual intelligibility is not such a
great criterion, because it varies along a continuum, and there is no
principled way of drawing language boundaries when we find -- as we often do
-- that mutual intelligibility among a group of related varieties varies
between, say, 20% and 80%.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk