Re: Language reconstruction question

Ian Maxwell wrote:
>So, I was just randomly wondering: How possible is it to reconstruct an
>ancestor language based on only one descendant?
Difficult.  If the language is rich in morphophonemic alternations, then by
"internal reconstruction" you can go a little ways.  For instance, by really
careful analysis and some imagination you might be able to figure out
something like the system underlying German (or English) strong verbs.  You
end up with sort-of "underlying forms" in the Chomsky/Halle sense; they
might "work" and be internally consistent, but they may or may not be
accurate reconstructions.

Looking at Spanish alternations like debér~débo vs. pensár~piénso and the
similar o/ue, might well lead one to conclude that some earlier stage had
two /e/ and /o/ sounds-- and that would be logical, but then to claim they
reflect an even earlier long/short contrast (as we know they did, from
outside evidence) would be speculative.

OTOH a language with few or no alternations, it's all but impossible.  In
Malay, for inst., there's simply no way to know that the /r/ of _bayar_ 'to
pay' has a different source from the /r/ in _layar_ 'sail'

I was under the
>impression that reconstruction was basically done by comparing the
>grammars of various descendants. If it's not possible, theoretically
>there could be older versions of the reconstructed proto-languages that
>we don't (and can't) know about, ne?
>
Right.  The proto-languages that have been reconstructed with fair certainty
only go back maybe 5-6000 years.  But of course, PIE must have developed out
of _something_; it didn't just pop into someone's mind on Friday, 4019
B.C.E.....  Then there's the game the Nostraticists play-- ideally, by
comparing various proto-languages you might see some relations between them.
But IMO it's so vague and so much guesswork that it's barely respectable,
though sometimes interesting and often amusing.