Re: [tied] Old English "a-spylian"

From: Richard Wordingham Message: 17482
Date: 2003-01-09

--- In, Piotr Gasiorowski
<piotr.gasiorowski@...> wrote:
> ----- Original Message -----
> From: <alexmoeller@...>
> To: <>
> Sent: Wednesday, January 08, 2003 8:11 PM
> Subject: Re: [tied] Old English "a-spylian"
> >> Post-Classical perlavo: (attested) means 'wash thoroughly', so
*experlavo: would have meant 'wash out thoroughly'.
> > and this form shouldnt give a romanian "spala" since there is no
loos of the cluster "rl" in words like "urla", "tsurloi", etc. You
should have had an "esperla", with much indulgence an "sperla" and
not "spala"
> <spãla>, to be precise. The assimilation of originally medial *e to
the *a of the next syllable and its pretonic reduction to <ã> seem
normal to me (experts please correct me if I'm being naive), as does
the simplification of *ex- (cf. *expantica:re > spânteca).
Etymological -rl- _at a prefix-root boundary_ was assimilated to -ll-
already in Latin (note the absence of Romanian rhotacism here!). We
can therefore assume *expellavo: 'rinse'.

Has anyone any good suggestions as to when -rl- stopped assimilating
to -ll- ?

A fair number of Alex's problems seem to arise because of the
difficulty of clearly defining the various sound changes. It occurs
to me it would be useful if they could be defined in a practicably
executable format. The best mechanism I know of is Mark
Rosenfelder's Sound Change Applier, at . It does have some limitations -

a) only 200 rules (Latin to French needs about that many in the
normal formulation of sound changes. A normal rule with a
conditioned exception seems to need 3 rules - alter segment with
conditioned non-change, say to X; basic change; change X back to
original change.)

b) It can't copy segments, e.g. for gemination (other than by having
one rule for each consonant affected)

c) It can't handle optional changes (there can be a lot of them)

d) It isn't designed for piping, e.g. PIE > Latin > Romanian.

e) Supersegmentals are fiddly, but they can be handled. (For
instance, Latin stress *can* be inserted via the rules and then used
as an input to subsequent rules.)

However, it is still a useful resource as it stands. It's available
both as source code and as an executable.