The fact that linguistic structure is
immensely complex does not mean that it is unknowable. It only means that an
"overall similarity" metric taking into account "all aspects" of a language
cannot be defined in a meaningful way. Imagine asking a biologist a question
like, "Is a dolphin more similar to a fish or to a bat?". You wouldn't be
surprised if the answer were "It depends", would you? "Similarity" is not very
important in "historical biology". The general body plan of a dolphin is like
that of a fish, but on other levels of description there is any number of
"hidden" homologies between dolphins and bats, which biologists regard as more
fundamental and more significant than the deceptive superficial similarity
between dolphins and fishes. No metric is needed here. It is enough if we
are able to reconstruct the position of the three groups in their
common family tree. Rather than classifying organisms according to any
possible arbitrary schema, we prefer a historical taxonomy, according to which
two species are "closer" if they share a more recent common ancestor. The least
arbitrary classification is one that reflects phylogeny. The kind of "closeness"
that is of primary interest in this kind of taxonomy is genetic.
Historical inguistics uses a similar
principle. We have a fairly reliable method which allows us to do the
following:
(1) Decide "beyond reasonable doubt" if
languages A and B are demonstrably related in the genetic sense.
(2) Identify regular correspondence classes
between A and B.
(3) Formulate a detailed relationship
hypothesis for A and B, including a partial reconstruction of their common
ancestor (Proto-AB) and of two chains of changes (two convincing "historical
narratives"), one leading from proto-AB to A and the other from Proto-AB to
B.
As genetic relationship is a transitive
relation, other languages can be compared with A and B, and if they
are recognised as being related as well, the family will grow. With a
little bit of luck, we may be in a position to reconstruct the internal
structure of the family (its hierarchical levels of "branches" and
"subbranches"). This happens if for two or more languages the initial part of
their reconstructed histories is identical (for example, all the Slavic
languages share the sequence of changes leading from PIE to Proto-Slavic,
though their more recent histories have diverged.
What else would you demand of a
reconstructive method? Of course we cannot prove _directly_ that PIE existed and
had the properties ascribed to it, but comparative reconstruction is based
on correspondences that cannot conceivably be due to chance because of their
pervasiveness and regularity, and the method requires us to control the data so
that secondary similarities due to borrowing are excluded from the
reconstruction. It is the rigour of the method that makes PIE a logical
necessity and that allows historical linguists to treat its reconstructible
traits as something real. But many aspects of PIE are forever beyond our reach.
Even if you had an entirely objective way of measuring the overall
similarity of two "complete" language systems, PIE is not and will never be
reconstructed completely.
To sum up, it makes perfect sense to ask if
Vedic Indo-Aryan is more closely related to Mycenaean Greek than to Hittite --
this question can in principle be answered by applying reliable methods. All
three languages are equally close relatives of PIE, being its lineal descendants
after the same length of time (give or take a practically insignificant
difference). The question which of them is most _similar_ to PIE can only be
answered if you explain what you mean by "similarity" and how you propose
to measure it in practice. I am afraid the answer will depend on your
subjectively preferred metric. You may have heard of lexicostatistics, which
uses the statistical study of vocabulary to "measure" linguistic proximity.
Unfortunately, the "hard figures" it produces are of no real use
in reconstructing linguistic history: vocabulary diffuses easily and at
unpredictable rates, introducing errors that cannot be controlled.
Piotr
----- Original Message -----
Sent: Monday, February 04, 2002 12:10 AM
Subject: Re: [tied] Why is PIE more centum than satem?
*****GK: It seems to me even less evident that it
can't. The danger
here is this: taking later, evolved,
"deeply modified" forms of the putative
original as
starting points for retrospective reconstruction,
and
accepting that these forms are indeed "immensely
complex objects", one
risks reconstructing something
that is totally unreal, an "immensely complex"
PIE
hypothesis that only exists in the imagination of
admirably qualified
(but perhaps totally mistaken)
linguists. The same logic holds whether you
are moving
in one direction or the other. The same difficulties
which
prevent (or render idle) an adequate response to
the broached issue, should
lead to the conclusion that
PIE is an artificial construct of a totally
different
order than the real and attested language families
that we know.
In that case, I would agree that the
issue is idle, since we are comparing
real languages
with something that cannot be demonstrated to have had
a
real existence. If PIE is a kind of retroactively
constructed Platonic Form,
I would consider comparison
of Hittite, Sanskrit, Greek etc. to it as a
completely
futile exercise in terms of historical relevance. But
if it is
assumed that PIE as a total system is a truly
historical linguistic
phenomenon, and if one is
confident about its "reconstruction" then the
broached
issue ("which language/linguistic family is
closer/closest as a
total system to this original")
becomes relevant and interesting,
difficulties
notwithstanding.******