Typology before decipherment?

[ECOLING at aol.com]

[ moderator re-formatted ]

Rich Alderson wrote in reply to Steve Long:

>If a text is undeciphered, it is unclear to me how any amount of typological
>information could be put to use.  Typology makes reference to concepts such as
>"noun", "verb", "adjective", "subject", "object", and so on, which cannot be
>applied to an undeciphered text.  Perhaps you mean something other than
>"undeciphered"?

There are indeed situations in which typology is used before successful
decipherment.  I suspect Chadwick used it for Linear B.
But a case I am more recently aware of is Indus Valley,
which is not generally regarded as deciphered
(no disrespect intended towards anyone).
Like other scripts with what Peter Daniels at least has called
"virtual bilinguals", one can look for indications of numeral systems.
Having recognized numerals with some plausibility, one can then
perhaps correlate plural vs. dual vs. singular as shown by
the numerals with changes in the morphology of suffixes on
words near the numerals,
words which might be nouns counted by those numerals.
Will the morphemes correlating with singular vs. plural be the
very last morphemes in those words?  Or will they occupy the
next-last slot (allowing for different cases)?  Or what?
There is typological knowledge about the relative closeness to the
root or stem of categories of number vs. gender vs. case;
some arrangements tend to be much more common than others.

Typological information
can be both used and abused in such a situation,
either to recognize a common pattern,
or to force an interpretation that might be at odds
with the evidence of the script itself.
More properly, at the current state of our knowledge,
typological catalogs can be used as heuristics.

Best wishes,
Lloyd Anderson
Ecological Linguistics