the meaning of "genetic relationship"

[Isidore Dyen]

----------------------------Original message----------------------------
The following is a belated response to the problems that you raised. I
hope that it helps or at least does not add to the existing confusion.
 
In discussing the meaning of a term like 'genetic relationship' it helps
to distinguish between our theory of what it is and (on the other hand)
what we do to demonstrate why we should believe that a genetic
relationship exists between two or more languages. Theoretically (and by
definition) two languages are interrelated (related to each other) if they
separately continue what was once a unitary (but not necessarily uniform)
language. To demonstrate that it is likely that two languages continue the
same unitary language it is necessary to show that they exhibit systematic
correspondences, better called collateral correspondences to distinguish
them from lineal correspondences that a language shares with its earlier
stages. It is these separate lineal correspondences that form the
collateral correspondences that are used in the reconstruction of forms of
the original unitary language.
 
Collateral similarities not organized as systematic
correspondences are not acceptable evidence of separate continuations of
an earlier unitary language. We do not know enough about the distributions
of phonemic similarities. Presumably if we did, we might be able to
provide an estimate of the value of the Greenbergian collections.
Systematic correspondences adhered to rigorously, allowing nevertheless
for deviant correspondences attributable to metathesis or analogic change
among some others, provide an approach to a qualitative collection of
evidence for a genetic interrelationship. The collection must be of
sufficient magnitude and/or quality to exclude any possibility
of being attributed to chance or borrowing. Similarity correspondences
between homosemantic or homeosemantic pairs of items do not escape the
likelihood of being due to chance. The hypothesis of interrelationship is
an explanation of collateral correspondences, or, if you wish, is
an inference from the collection of items paired by satisfying the demand
for collateral correspondences.
 
Greenberg's organization of African languages into four families was
successful because it could be shown that their respective linguistic
material exhibit collateral correspondences. The
fact that he was led to his conclusion through what he calls 'mass
comparison' based on apparent similarities is beside the point, though it
suggests the possiblity (not the likelihood, which is required in
scientific determinations) that there might be a relationship.
However speaking about languages as 'wholes' is not loose talk. A
language, technically as opposed to a dialect, is a bounded chain of pairs
of mutually intelligible dialects. It has a boundary that it shares with
each other language since none of its dialects is mutually intelligible
with any of theirs. In this sense it is a whole. What you speak of as
'genetically related' parts--with the implication that some parts are not
'genetically related'--are more commonly called 'cognate' or 'shared
inheritances'. The remainders are composed of individual inheritances and
innovations, the latter including borrowings.
The reason interrelated languages are treated as wholes is that each
represents a separate continuation of the original unitary language via a
succession of native speakers, their separation occurring at the moment
the last cross-pair of mutually intelligible speakers had vanished.
 
It is an assumption of the comparative method that different languages do
not mix (under natural circumstances). Creoles are either aberrant
dialects if they are still part of a chain with other dialects or
different languages if they are not. In the latter case its first native
speaker was not mutually intelligible with any dialect of any of the
languages that contributed linguistic material to it. (Of course it is
conceivable that today we might be able to concoct a dialect chain to
connect two contemporary languages, but it is hard to see what purpose it
would serve and therefore why we should carry that possiblity into our
method.)
 
 
 
On Tue, 18 Mar 1997, benji wald wrote:
 
> There is a point about "genetic relationship" that I think is worth
> considering, because I think both sides on various controversies about it
> tend to ignore it.  To begin with, we can take Ruhlen Merritt's fallacious
> argument that reconstruction presupposes genetic relationship, which he
> takes to mean that genetic relationship has already been "established"
> before comparative reconstruction can begin.  For him it is established on
> the basis of the kinds of Greenbergian mass comparisons which have figured
> (or been attempted to figure) most recently in this list in the discussion
> of the relationship between Dravidian and the African families -- and,
> admittedly, in Greenberg's division of African languages into four genetic
> families, now generally accepted -- but not without further testing and
> refinement.  Against the mass comparison method, other historical linguists
> have inevitably argued about the confounding effects of borrowing and
> chance resemblances.  My thought, as follows, is that when we talk about
> genetic relationships among LANGUAGES, rather than parts of the lexicon,
> morphology etc etc, both sides obscure something.
>
> Thus, first, against Merritt's argument.  He's absolutely wrong.  Mass
> comparison gives the basis for a genetic HYPOTHESIS.  Comparative
> reconstruction TESTS that hypothesis.  Without it nothing has been proven,
> not genetic relationship, borrowing or chance resemblance.
>
> Next, to the extent that a comparative reconstruction is successful it does
> NOT demonstrate that the "languages" involved are genetically related, but
> only that those PARTS of the languages which are reconstructed are
> genetically related.  Of course, it provides confidence that other parts of
> those languages are also genetically related, but, again, that is only
> DEMONSTRATED when reliable comparative reconstruction is performed on those
> other parts.  Otherwise, it remains only a possibility.  Creoles and mized
> languages show that genetic relationship of some parts of a set of
> languages do not always presuppose that other parts of the same languages
> are necessarily GENETICALLY related.  And indeed, it is well-known that all
> languages borrow as well as genetically inherit.
>
> Thus, speaking about genetic relationship among "languages" as "wholes" is
> loose talk.  The internal structure of trees intending to show branching
> genetic relationships are always a problem because different parts of a set
> of languages are not always related in the same way.  Innovations begin in
> different areas and have different spreads according to the time of contact
> and subsequent events.  This is well-known from dialect geography, but adds
> confusion to arguments about genetic relationship.  Sometimes, it does
> little harm, but when we are in the mass comparison stage, it can result in
> much futile argument.
>
> Having said this, I admit that the poor quality of Winter's data and
> arguments have been worth pointing out.  But I see no reason to postpone my
> thoughts until more competent proposals flare up.
> --  Benji
>