Typological inference

[Nicholas Widdows]

[ moderator re-formatted ]

<Steve Long wrote>

>> I also think one striking thing about the "[typological] approaches"
>> described is their strength in helping those who must identify language-
>> types and structure in undeciphered text, i.e., "at ground zero."

<Rich Alderson replied>

> 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"?

<and my idea is this>

I would have thought the approaches used to start on Linear B (positing that
two glyphs had the same consonant and had a m./f. vowel alternation) and
cuneiform Persian (matching a guessed X son of-Y to Y son of-Z) could in
fact yield some grammatical and even typological information for a
completely unknown language, if you were lucky; then give clues to meaning,
if very lucky.

If you had a text with phrases such as
	GENEK NIBAKOT
and
	ZUQUD NIBAKOT
and
	GENEK BI ZUQUD NIBAKUR
and the element BI and the endings -OT and -UR recurred often, you'd suspect
you had "and" and singular and plural concord. (If you had BI joining more
than two elements that'd rule out sentence 3 being a passive.)

If the common order was
	NIBAKOT GENEK
etc. then you'd suspect SVO.

If you found either pattern GENEK DEVKULOT ZUQUDU or GENEKA DEVKULOT ZUQUD
recurring, you could infer accusative or ergative.

If you had
	NUKIGAP NULEVE NUZUQ
and
	RATAG RAFUNA RAQOB
without the guessed coordinator BI then you'd guess you had a noun and its
adjectives.

If first Genek then Zuqud then others appeared in a series of formulae of
succeeding date, you could guess the verb meant "succeeded" rather than
"slew"; you could get "son" and a genitive just from the formula;... well,
it's a start.

Nicholas Widdows