Q: the 'only six' argument

[Robert R. Ratcliffe]

> ----------------------------Original
> message----------------------------
> Larry Trask wrote:
>
> > So, my question: does anybody believe that any version of this
> > statement is valid?  More precisely, do we have a number N and
> > a set of criteria C such that the existence between two languages
> > of N matches satisfying criteria C is enough to guarantee that
> > the languages must be related?

Wasn't this what Donald Ringe was trying to do? (The Factor of  Chance
in Language Comparison, Philadelphia 1992).

But the statement phrased as you have it is certainly not valid.  First
no number of "matches" (sound correspondences?) can *guarantee* that the
languages are related, only that the probability of their being related
is high. Second "related" has to be understood as historically related
rather than genetically related, because numerical criteria only help to
decide the issue chance vs. non-chance similarity, not which type of
historical contingency (descent from a common source or subsequent
contact) may have produced the non-chance pattern. Third there is no
absolute number valid in all cases, because it depends on the size and
nature of the sample being compared. Specifically in the case of sound
correspondences, the bigger the dictionary or word list the more chance
correspondences can be expected; and the smaller the segment inventories
of the languages compared the more chance correspondences can be
expected. This is because the average expected number of chance
occurences of an event (in this case a correspondence at a given
position in a word) is the probablility of the event (in this case the
relative frequency in the given position of the segments compared
multiplied by each other) times the number of trials (in this case the
number of semantically equivalent words available for comparison).

So if you have two languages A and B, both of which have only ten
consonants evenly distributed in word first position, and you have an
A-B dictionary which has 10,000 entries correlating one word in A with
one and only one semantic equivalent in B, with no synonyms in either
langauge, you'd expect to find about 100 matches between any first
consonant in A and any first consonant in B (chance that x will occur as
first consonant in A: 1/10, multiplied by chance that y will occur as
first consonant in B: 1/10, multiplied by  total places where 1st C of a
word in A can be compared with 1st C of word in B: 10,000). So you
wouldn't be justified in suspecting a historical relationship till you
got a good bit over a 100 matches.  On the other hand if you had two
languages with 25 consonants evenly distributed and a lexicon based on
a1000 word random sample, you'd expect an average of only 1.6 first
consonant matches (1/25 * 1/25 * 1000). So you'd be justified to suspect
a non-accidental, hence historical relationship even with as few as 4 or
5 matches.



 Bobby D. Bryant wrote:

>
>
> In short, I don't think such a formalization of the problem in terms
> of N and
> C is going to work in practice.  At some level you are always going to
> have
> to pile on enough examples to convince your peers, which is of course
> the way
> things have always worked.
>

Piling on enough examples to convince your peers no longer works in
practice, or else the long-distance comparison debates would not have
become as acrimonious as they have. Formalizing the problem seems to me
to be the only way forward. Besides, isn't that where the joy of
research lies-- in ever sharpening and refining our understanding of our
subject matter and of the tools we use to analyze it?




-- -----------------------------------------------------------
Robert R. Ratcliffe
Dept. of Linguistics and Information Science
Tokyo University of Foreign Studies
Asahi-machi 3-11-1,
Fuchu-shi, Tokyo
183-8534 Japan

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/histling/attachments/20000901/0482ac27/attachment.htm>