reality of PIE

[Larry Trask]

Pete Gray writes:

> Larry said:

>> Without a tolerably unified PIE .....

> Quite!   My point is that we cannot go beyond the "tolerably unified" and
> speak of a single, undifferentiated language.   I thought this was standard
> stuff.   To reconstruct PIE without allowing for some internal variety would
> seem to me - in my innocence, and in light of the IE evidence - somewhat
> doubtful.

But nobody is proposing that PIE was a language devoid of variation.  Every
living language exhibits variation: regional, social, contextual, individual.
And PIE cannot have been different.

It is merely that variation is generally very hard to reconstruct.  Only
occasionally do we find evidence pointing to the existence of variation in our
reconstructed language, and even then such evidence is often hard to interpret
unambiguously.

I can give you a nice example from Basque.  We can reconstruct for Pre-Basque a
certain noun meaning 'interval', but the modern reflexes are inconsistent: some
point clearly to earlier *<gune>, others equally clearly to earlier *<gunne>.
There is no way of resolving this with the data available, and it may be simply
that the form of the word was variable in Pre-Basque.  Or there may be some
other explanation which we can't detect for lack of evidence.

> Indeed, this very variety is what some of the glottalicists rely on - for
> example, in order to link Skt /bh/ with Germanic */b/, both derived from
> dialectic allophones (or "dialectophones") of b/bh.

Sorry; I don't follow.  As far as I know, *everybody* links Sanskrit /bh/ to
Proto-Germanic */b/, both being derived from PIE */bh/.

> That's just one
> example - you know the kind of stuff I mean.   A single unified PIE is
> certainly not what we can reconstruct, except as an artificial abstraction.

But nobody is claiming that our reconstruction contains every detail of the
speech of the PIE-speakers.  That's out of the question.  What we *do* claim is
that we can reconstruct a great deal of PIE.

> A more interesting  and slightly philosophical question is whether we
> believe a perfectly unified pre-PIE is a necessity.   I am arguing that it
> is not - that dialect variation within a language is perfectly normal, and
> the daughter languages may indeed reflect that variation, and even show
> mixing of the dialects (as modern English does).   Of course there are
> examples of a single dialect spawning variant daughters, but I am
> challenging the assumption that all daughter languages must - by
> definition - come from a single undifferentiated original.

Then you are arguing with no one, because no one either denies the reality of
variation or insists that PIE was a language without variation.

>>>> Genetically related languages were once the same language.

> On this, Larry said:

>> The statement above is true ... by definition.

> This begs the question I asked above, and also relies on questions of
> definition - are we talking of a single unifed undifferentiated language?
> That's the concept I am attacking.   It is not true *by definition* that
> genetically related langauges derive from a single undifferentiated
> ancestor.

Assuming that 'undifferentiated' means only 'exhibiting sufficiently little
variation that we may regard the whole as a single language', then I'm afraid
that this *is* true by definition.  That's what we *mean* by 'genetically
related languages'.

> It may be true by definition that they derive from closely
> related forms of that language, but where is your evidence that all must
> come from a single form of that language?   I think it is an assumption open
> to challenge and debate.

Put it this way: the Romance languages derive from a single ancestor, spoken
Latin, in the same way that the IE languages descend from a single ancestor,
PIE.  Nobody claims that spoken Latin was devoid of variation, but only that
spoken Latin was, by any reasonable standard, a single language.

>> Languages which do not
>> descend from a common ancestor are not genetically related.

> Even if they descend from sister languages, which are themselves descended
> from different dialects, which are themselves reflexes - maybe quite complex
> ones - of an earlier dialect continuum - which is itself the result of
> earlier close dialects - etc etc ....  So that there is no single unified
> undifferentiated ancestor?   Or do you believe that there always must be a
> single ancestor without variation?

Again, *nobody* is claiming an absence of variation for any language.

But are you suggesing that a dialect continuum can derive only from an earlier
dialect continuum, and so on, back in time without limit?  I remain to be
persuaded that such a thing is possible.

Anyway, 'dialect' is a two-place predicate: a dialect must be a dialect *of*
something: it can't just be a dialect, *tout court*.

> Perhaps we are again using different meanings of "common ancestor" - yours
> more loose, including variation, and mine excluding it in order to make the
> point that a single dialect-free ancestor may not be necessary.

Actually, I am very close to taking the presence of variation as part of the
definition of 'natural language'.  Not quite sure if I want to do that, but we
never find natural languages without variation.

I also suspect you may be taking regional variation as the only significant
kind of variation.  If you're not, I apologize, but in fact regional variation
is just one of several kinds of variation that languages exhibit.  It is also
the one that is generally easiest to trace back to an earlier absence of
variation.

> Creoles - how can you describe a Creole as descended from a single ancestor?

I haven't done so.

> Doesn't his mean prioritising one of its "parents" over the other?

There is nothing wrong with this conclusion in principle.  We routinely
describe an individual creole as 'English-based' or 'Portuguese-based', or
whatever -- and with good reason.

> Your restriction of
> "related" to mean only "genetically related" means we cannot say, "English
> shows a closer relationship to French than to Italian."

Correct.  We can't, because it's not true.

> Instead we have to
> spell out the nature of that relationship, and say, "English is equally
> related to both French and Italian, but has been more deeply influenced
> by... and so on."

Are you seriously objecting to stating the facts accurately?  Do you really
think it would be a step forward to abandon all distinctions and shove every
kind of connection between languages into a black bag marked 'related'?
Count me out.

> I want to say both sentences have their place, and given the right
> understanding of "relationship", both are true.   You appear to be saying
> that the first is always wrong.  I would say it is only wrong if
> "relationship" is understood purely in a genetic sense.

But that is how we *do* understand the term in historical linguistics.

Basque has borrowed thousands of words from Latin and Romance.  But this does
not make Basque any more closely related to Romance than it ever was.

2000 years ago, Basque was a genetically isolated language (as far as we know).
 Today, Basque is a genetically isolated language whose lexicon has been
heavily influenced by Latin and Romance.  That's all.

> So I ask, is the only relationship two languages can have, a genetic one?
> (Indeed you talk of "inventing non-existent "relationships", and confusing
> these with
> > genetic links?")
> What about Sprachbuende, etc?   There are other relationships - so why deny
> them?   Why not keep the word "relationship" open, and specify "genetic"
> when necessary?

First, because our terminology is established, and changing established
terminology without a profound reason is a bad idea.

Second, because we already have terminology for labeling these other states of
affairs -- 'Sprachbund' being a good example.

I've written several dictionaries of linguistics, and nobody knows better than
I do just what a mess much of our linguistic terminology is in.  It's a relief
to find an area of linguistics in which our terminology is in good order and
generally accepted.  And messing about with this established terminology is
absolutely the last thing I want to contemplate.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk