--- In cybalist@yahoogroups.com, "Davius Sanctex" <gruposdavius@t...>
wrote in Message 22854:
[Peter] Minimal pairs are not enough to establish the existence of a
[Davius] We can try to reformalate the principle in the following
way: "If
we have a minimal pair, the two word must have different phonological
unferlaying sequences explaining the minimal pair" [*]. In this way
diferences in French between minimal pair [vE~] 'wine' / [vE] 'I see'

[Richard inserts] 'I go', actually.

[Davius] do not
say us that necesarily it is a contrast between /E~/ and /E/ but
there are
diferent underlaying representations /vin/ - /vE/.

[Richard inserts] How do you preclude an underlying /vEz/ for <vais>
= [vE]?

[Davius] The alternace [-in-] /
[E] can bee seen por example in <divine/divin> confirmin that [E~] is
represented underlaying as /in/ not as /E~/ despite the existence of
minimal pair [vE~] / [vE]. Why do not use all manuals directly the
[*] of the phonemic principle?
--- End forwarded message ---
[Richard] Are there no problems with features that cannot be
projected back to a suitably modified version of Old French? I
believe the
motivation for analyses such as this is rules that work write better
writing than in speech. Analyses have been published that use a
representation very close to Latin!

A very relevant point is that there is evidence that people learn
lists of
irregular forms rather than carefully sequenced rules. Evidence as
to what
is actually underlying can pop up in grammatical errors. There are
parallels in Oceanic languages, which have final consonants that only
when affixes are added.