On Sun, 2 Dec 2018 at 00:31, Tristan Mc Leay <[log in to unmask]> wrote:
> I know there's many ways to get rid of Cl- clusters, but so far I haven't settled on one but I do like changing phonological constraints by just not using/discarding words that don't fit the pattern of the language.

Weirdo! :-p

The idea creeps me out, really.  I first encountered it in a
discussion of Russian (wow I'm rubbish at citations), the author
positing that words with heavy, counter-sonority clusters like /mgla/
'haze, darkness' had been lost historically at greater than average
rates, and I was bothered then too.

Let's suppose that in our fieldwork we document two languages X and Y.
Language X has a strict CVC structure while language Y allows initial
C+/l/ clusters.  We're trying to do historical reconstruction, and we
have found a section of the lexicon that seems to be shared between X
and Y, but (try as we might with clever correspondences) this set
contains no words showing clusters in Y.  Well, one could hypothesise
(1) This set of words is a stratum of borrowings from X into Y.
That cleanly accounts for the absence of the clusters.  One might also posit
(2) X and Y are related, and this is the inherited vocabulary;
clusters in Y are borrowed from a third source.
-- supposing there was a third source with intimate enough contact
with Y to enlarge its phonotactics.  But this
(3) X and Y are related, and this is a subset of the inherited
vocabulary; clusters in Y are also inherited, and X has selectively
disappeared all the words in question.
I've only seen as a reductio ad absurdum.  Look, your proposal for the
language family can't be right, 'cause if it were we'd have to imagine

Or, there is a mechanism whereby words with difficult _inflection_ are
selectively lost.  Like how the Romance languages have for 'carry'
descendants of the very regular _portō, portāre, portāvī, portātum_,
and none of them have gone for odd suppletive _ferō, ferre, tulī,
lātum_.  That one makes sense.  For language learners, encountering a
word-form reinforces in memory all the forms regularly related to it;
so if you have two synonyms, one regular and one irregular, the forms
of the regular word will achieve more mutual reinforcement than those
of the irregular word, so the regular word will be more cognitively
accessible, and over generations that could add up to loss.

But how's that supposed to go for phonology of inherited words?
Phonology is supposed to be discrete and categorial: if a word is
(phonotactically) valid, it's valid, end of!  What's the mechanism
that would lead speakers to start replacing?