PIE vs. Proto-World (Proto-Language)

[Larry Trask]

On Wed, 21 Jul 1999, Adam Hyllested wrote:

> On Fri, 16 Jul 1999, Ralf-Stefan Georg wrote:

>> "Having consonants" may be a
>> universal feature of human language, "Being related to some other lg."
>> simply is not.

> A fascinating thought: Imagine "Being related to some other lg." WAS
> actually a universal feature of (spoken) human language, wouldn't it
> then be the only true universal?  In other words, if all spoken
> languages should prove to be derived from one common source, would
> we then be able to distinguish true language universals from other
> features shared by all languages ONLY because of the fact that these
> features were present in "proto-world"? Language typology would
> still be a great help in the work of reconstructing proto-languages,
> but would it prove anything about the nature of spoken human
> language?

An interesting point.  This is the problem of `founder effects'.  A
founder effect is a feature of an ancestral language which is not
cognitively necessary -- that is, it is not a true language universal --
but which happens by chance to persist in all descendants of that
ancestral language.  Clearly founder effects have the potential to
distort our understanding of true linguistic universals.

This idea has received some discussion in the literature, though not, I
think, a great deal.  Johanna Nichols has appealed to founder effects to
account for the significantly non-random distribution of certain
structural and other features among the world's languages.  Nichols has
no interest in "Proto-World", but I can cite another case which is
relevant.

When I started my PhD some years ago, no example was known of a language
with object-initial basic word order (OVS or OSV), and it was widely
suspected that such languages were impossible.  Since then, a few
O-initial languages have turned up, practically all of them in Brazil.
If we hadn't managed to study those languages before they vanished, we
might have gone on believing forever that O-initial languages were
impossible.

Well, they are indeed possible, but they appear to be rare.  But why are
they rare?  Geoff Pullum once indulged in some musings on this point.
Maybe O-initial languages really are cognitively or functionally more
difficult than other types.  Then again, maybe they are not, and the
observed distribution is an accident, an artefact of human history:
speakers of non-O-initial languages have simply been lucky enough to
flourish at the expense of speakers of O-initial languages.  If the
second is the case, then we are looking at a kind of founder effect.
But who knows?

There exists a principle, called the Exit Principle, which says this:
any feature of a language which is not cognitively necessary can change.
This implies that every such feature *will* change eventually, but who
knows how long `eventually' might be?

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk