Sorry for this re-posting, but I forgot to
format the message properly.
----- Original Message
-----
Sent: Friday, August
30, 2002 12:18 PM
Subject: [tied] Re:
Initial stress
> Ah, you're arguing coincidence. But
then the laws of statistics would apply and how could we then get such a large
contiguous area?
Well, first of all, the laws of statistics
do not prohibit clusters of identical events. They even predict them. If you
toss a coin ten times you will not normally get a sequence like HTHTHTHTHT or
THTHTHTHTH (on the average, only one sequence in 512 will be perfectly
alternating). I've just carried out this little experiment,
getting THHHTHTTHH, with a fairly large contiguous "area" of "heads". Now
if I argue that the second occurrence of "tails" is intrusive (due to a late
invasion), who knows if we aren't dealing with a vast area of "heads", covering
half of the string?
Secondly, the existence of te area is not
_entirely_ due to coincidence. For example, the fact that Proto-Germanic was
already initially stressed strengthens the odds in favour of any Germanic
language to have initial stress (unless an innovation occurs, as in English, but
innovations are on the whole many times rarer than retentions). Thirdly, I don't
rule out diffusion or even substrate influence. I simply don't see any solid
evidence for the latter.
Piotr