Query:r/l

[McConvell, P.]

Discussion recently on FUNKNET raised the question of
alternation between r and l in Spanish and r>l changes. This
was in the context of the general relationship between
historical change and child language/slips etc. I would like
to focus a query on r/l alternation and historical change
crosslinguistically.

Mary Laughren and I have been working on a paper discussing an
innovation r>rl in a particular subgroup of Pama-Nyungan
(Australian). (r is a retroflex glide distinct from rr an
alveolar tap also found in most Australian languages; rl is a
retroflex lateral). We have found the same change happening in
other (only distantly related) Australian families, and also
the opposite change rl>r.

It seems that this may be a rather common type of change in
languages, and we would appreciate references to work
discussing it as a general phenomenon, or regionally.

Secondly, as far as the Australian data goes there seem to be
some generalisations emerging about where one or other (the
rhotic or lateral) is more likely to occur. An rl>r change in
Lardil affects initial and intervocalic segments, but rl is
retained where it is a preconsonantal coda; this seems to
parallel the observation that r>l in Spanish dialects occurs
most freely in codas.

Another tendency in the Australian data for which we have
neither parallels nor a phonetic explanation at present is
that r>rl apparently does not occur, and rl>r occurs more
consistently, in the environment of preceding i. Other
examples or explanations of this apparent affinity or i and r
as opposed to i and a lateral would be welcomed.

[This query has been posted on FUNKNET, LINGUIST and
AUSTRALIAN-LINGUISTICS-L; apologies for duplication]