reality of PIE as dialect network

[Larry Trask]

Lloyd Anderson writes:

> The following comments by Larry Trask on comments by Peter Gray
> reveal yet again that the use of absolutely discrete categories
> yes-or-no may be a model not fitting the facts, and attempts to force
> such a terminological usage on us may be counterproductive,
> by rendering such a model incapable of application to messy reality.

> [PG commenting on someone else]

>>>> No.  Relationship is an absolute.  ....
>>>> Genetically related languages were once the same language.

>>> Sorry, Bob, I can't agree, and I suspect you're in a minority these
>>> days (though I may be wrong!).

> [LT, with clarifying inserts [ ] in the first sentence]

>> You [PG] are [wrong], I'm afraid.  The statement above is true not just
>> because all linguists believe it: it is true by definition.  Languages which
>> do not descend from a common ancestor are not genetically related.

> Not so, the matter is not so simple.

Yes, it is.  (This is beginning to sound like a British Christmas
pantomime. ;-) )

> Sufficiently massive borrowing *does* constitute a kind of
> genetic relation,

No, it doesn't.  This is rather like arguing that I can become more closely
related to you by copying your behavior and your clothes.

English has borrowed no words that I know of from Pitjantjatjara, and the two
are not related.

English has borrowed one word ('sauna') from Finnish.  This doesn't make them
genetically related.

English has borrowed a handful of words from Chinese.  This doesn't make them
related.

English has borrowed quite a number of words from Algonquian.  This doesn't
make them related.

English has borrowed a couple of hundred words from Italian.  This doesn't make
them related -- or, at least, it doesn't make them more closely related than
they were before.

English has borrowed several thousand words from Old French and modern French.
This doesn't make them related.

Lloyd, are you trying to tell us that there exists some number N, such that,
as soon as a language borrows N words from another language, the two suddenly
become related?  If not, then what *are* you trying to say?

> and the more sophisticated researchers today
> do recognize that all of these kinds of genetic relation do occur
> simultaneously, in various different combinations and mixtures.

OK.  Which "more sophisticated researchers" have you got in mind here?
Can we have some names and publications, please?  *Who* exactly has argued
that borrowing on any scale constitutes a genetic relationship?
Answers, please.

> That does not mean we cannot distinguish the kinds.
> And with careful work and also some luck, we can also use the manifest
> results of language cross-breedings

Now we have cross-breeds?  Lloyd, just which languages would you regard as
cross-breeds?  Cross-breeds of what?

> to conclude something about
> the circumstances of the language contacts and social contacts
> which led to them.

We would all like to do this, but that doesn't mean we have to ruin our
established terminology in favor of a pink haze of undifferentiation.

> If two language clusters are in intimate contact
> (whether ultimately descending from some proto-world or not)
> long enough that their interaction creates a complex dialect
> network, then that dialect network *is real*
> (referring here to Trask's phrase that PIE is real,
> which Peter Gray did not in any way deny),

But this state of affairs is *not* what we call a 'dialect continuum'.
It's more like an extreme -- and so far hypothetical -- variety of
Sprachbund.

> yet it may be impossible in the time frame of that dialect net
> or in any time frame somewhat preceding it
> to say that there is a single point uniform ancestor,
> from which all descendants evolved.

No common ancestor, no dialects.

The Balkan languages do not constitute a dialect continuum merely
because they are in a Sprachbund.

> The same may be true of a single language having spread
> across an area with a number of other languages which
> become substrates of different parts of the proto-language cluster.

I'm getting a bit edgy about this repeated incantation of 'substrates',
and I may shortly fire off a comment or two.

> It simply may be a more useful model to think in terms of
> an ancestor with some regional variations which do *not*
> go back to a common origin, in either of the real sorts of situations
> just mentioned (and others).

> This in no way denies that there should *also* be single origins
> for some common elements in such situations, nor does it deny
> that much significant IE morphology *does* go back to a common
> singular origin in PIE.

This sounds to me like a version of Dixon's punctuated-equilibrium
scenario.  Well, we've discussed that before, and we'll doubtless do so
again.  But note: Dixon explicitly *denies* that his scenario is
relevant to IE, and he accepts the reality of PIE.

> Nor, more importantly than either of the above, which are conclusions,
> does it deny that it is useful to try to lead various attested forms back
> to common origins in PIE, to discover more cognate forms and
> structures than are known at any given time.

> All of these models and techniques can operate simultaneously,
> with more benefit that if we limit ourselves to only one,
> as long as we keep in mind the limited capabilities of each
> technique we use, that *every* technique is biased towards
> certain sorts of answers rather than others, biases which may
> be more harmful or helpful depending on the particular nature
> of the context being investigated.

I don't object to anything in this last paragraph.

But I do object to the repeated assertions that there is something
fundamentally wrong with our current understanding of IE and of PIE.
If you want to seek evidence that proto-languages and family trees
are less than universally applicable, IE is one of the very worst places
to look.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk