IE versus *PIE

Larry Trask larryt at
Mon Apr 30 09:09:59 UTC 2001

Steve Long writes:

[on Uhlenbeck, Trubetzkoy and Tovar's rejection of PIE and my rejection of
their views]

> Once again, this is not scientifically justified.  Mr. Stirling rightly
> makes the point of falsifiability.  As long as IE languages show any
> serious commonality, there is going to be a genetic relationship shown
> and a hypothetical ancestor will be reconstructible.

Oh, no -- at least, not if we require minimal standards in anything we
accept as a "reconstruction".  I can "reconstruct" a common ancestor for
English and Zulu if you'll allow me all the elbow room I want.  But I have
more exacting standards in mind -- standards which our reconstruction of
PIE undoubtedly meets.

> This does not and cannot logically disprove the presence of other "genetic
> strains" in those languages.

No one is claiming that every single element in an IE language must descend
directly from PIE.

> I think Prof Trask is guilty here of the very reification which he himself
> has noted before on this list.  Something like the idea that a language
> is a single organism in which the parts must all descend from a single
> ancestor.

No; certainly not.  Nobody would be mad enough to defend such a position,
and, as a native speaker of English, and one who has already used a number
of non-Germanic words in this reply, I am hardly well placed to be the
first such madman.

> In fact there is nothing that prevents a community of speakers
> from adopting lexical and structural elements from more than one group of
> earlier speakers.

Of course there is not, and probably every community of speakers does
precisely this.

> The assumption that there must be some kind of core
>  retained from one linguistic community or the other is nothing but an
> assumption.  Nothing makes it necessary.

Agreed, but not the point.  I am not maintaining that no mixed languages
can exist: that position has now been clearly falsified.  I am only
maintaining that the IE languages are not examples of mixed languages -- a
very different matter.

> And just to be clear about terminology, the "relatedness" of historic and
> documented IE languages CAN be asserted with a high degree of certainty.

I agree entirely, though I must make it clear that I understand
"relatedness" as meaning "common origin from a single ancestor", and not as
meaning something substantially vaguer than this.

> Statistically the possibility of what these languages have in common
> happening by accident is highly improbable.

Highly improbable, indeed.  The universe will never exist long enough to
allow anything comparable to the IE languages to come into existence by

> But this is definitely not same as saying that all those individual
> commonalities had to derive from a common source.

Yes, it is.

> There is nothing I have seen that makes it clear that
> those individual commonalities could not have come from more than one
> unrelated source or even distantly related sources.

I am flabbergasted.  What, in my view, makes this position untenable is the
*orderliness* of the connections among the IE languages.  We don't have
mere resemblances, and we don't merely have bags of resemblances.  Instead,
we have systematic patterns linking all these languages in a way which is
impossible to explain except by descent from a single common ancestor.

Consider a parallel from North America.  Athabaskan is a secure and
well-understood family.  Eyak shares a common origin with Athabaskan: this
was proved some years ago by Michael Krauss, who uncovered the systematic
correspondences linking the two.  So far, fine.

Now, Tlingit is commonly suspected of being related to Eyak-Athabaskan, and
it certainly shares a large number of features with EA.  However, in this
case, no orderly patterns can be found linking the two: all we have is a
large collection of similarities, enough to demonstrate some kind of
historical link, but no more.  And, because of the total absence of
systematic and orderly patterns linking the two, no Proto-Tlingit-EA can be
reconstructed -- according to our ordinary ideas of what a reconstruction
should look like, and contrary to Steve's earlier assertion.  Therefore, we
cannot prove that Tlingit is related to EA, in my sense of the term.

There are, of course, several possible explanations, and all have been

(1) Tlingit is unrelated to EA, and the similarities result entirely from
long contact.

(2) Tlingit is a mixed language, resulting from the imposition of an EA
language on an unrelated substrate, with consequent and extensive mixing.

(3) Tlingit is an EA language with a highly unusual history.  In
particular, it broke up into a number of distinct regional varieties, but
some unusual event then caused these varieties to be re-combined --
effectively, koineized -- in an unsystematic manner.

All of these proposals are consistent with the observation that Tlingit
exhibits many similarities to EA but no systematic correspondences.  Any
one of them might be right, but we can't tell.  In any case, as things
stand, we cannot prove that Tlingit is related to EA -- in spite of the
abundant similarities -- and we cannot reconstruct a common ancestor.

The case of IE is wholly different: it does not resemble the case of
Tlingit and EA.

> (As far as what's dead and what's not, and as far as the opinion of the
> linguistic community, I'd be happy to privately share with Prof Trask the
> opinion of some eminent linguists who most definitely do not believe that
> PIE can be plausibly recovered at all.)

Sure; feel free.  I'll be happy to hear about this.  But the wording is
peculiar.  Are you, Steve, suggesting that some linguists believe that PIE
existed but that we cannot substantially recover it?  This is not at all
the position of the three men who I cited earlier and whose view I rejected
as untenable.

> That's not to say PIE didn't exist.  It's just to say that its existence
> is no where as necessary as Prof Trask makes it out to be.  As a
> scientific matter at least.  One's personal beliefs are another matter.

Sorry, Steve -- I can't agree with this for a moment.  The former existence
of PIE is a necessity forced upon us by the data.  The relationships among
the IE languages are just too orderly and systematic to result from
anything other than common origin from a single ancestor.  Other kinds of
origin could produce resemblances, but they could not produce the
systematic relations that we see.

Finally, a very general point on which I perhaps agree with Steve.  We know
it is possible for a language to descend from a single common ancestor in
the familiar way.  The IE languages are splendid examples.  And we know
-- now -- that it is possible for a language to descend from two (or
possibly more) ancestors.  Michif is a magnificent example.  And we also
know that it is possible for a language to descend from no ancestor at all.
Nicaraguan Sign Language is the clearest example I know of.

Given all this, we must be prepared to consider the possibility of all
kinds of intermediate cases.  One of the most interesting intermediate
cases is the Austronesian language Takia, which has borrowed the *entire*
grammar of the neighboring but unrelated Papuan language Waskia, so that
every Takia sentence is a morpheme-by-morpheme gloss of the corresponding
Waskia sentence -- while at the same time Takia has borrowed no morphemes
at all from Waskia.

Linguistic work in the last twenty years has demonstrated beyond dispute
that linguistic descent can be, and sometimes is, far more complicated than
our conventional family-tree model would suggest.  This is fascinating, and
I am confident that the next twenty years will turn up yet more surprises.
But it is out of order to conclude that the IE languages result from one or
another of these more colorful scenarios: the evidence is solidly against
this.  If things were otherwise, our predecessors would never have
succeeded in reconstructing PIE in the first place -- just as specialists
today cannot reconstruct a Proto-Tlingit-Eyak-Athabaskan, in spite of the
obvious commonalities.

Larry Trask
University of Sussex
Brighton BN1 9QH

larryt at

Tel: (01273)-678693 (from UK); +44-1273-678693 (from abroad)
Fax: (01273)-671320 (from UK); +44-1273-671320 (from abroad)

More information about the Indo-european mailing list