Philip wrote:
One of my reactions was that I'd probably say "des Zwenns" rather than
"des Zwennes" -- but that I wouldn't consider the latter wrong: it'd
merely sound formal or maybe stilted or archaic.

This kind of thing happens all the time because of the patterns
that exist in the language, and when a speaker encounters a
novel form, they can fit it into whatever pattern seems right.
Some forms can fit into more than one pattern, though.  So, a
simplified set of patterns would look like this:

das X          <->    des Xes
NP              <->    NP
"Z (nom.)" <->     "Z (gen.)"

das X          <->    des Xs
NP              <->    NP
"Z (nom.)" <->     "Z (gen.)"

That's a simplified way of capturing both patterns.  When a
speaker has to decide which pattern to choose, though, several
factors come into play.  For example, if there a bunch of words
in German that are "das Cenn" taht fit into (1), then (1) will
probably be the one most speakers go with.  If a fair number
fit into each one--and, in particular, if there are some that can
go with both--then the speaker is left to figure out what the
difference is between the two patterns.  The generalization you
seem to have picked up on is that (1) is a bit more archaic than
(2), making (2) the more natural choice.  Crucially, (1) isn't
ungrammatical, because presumably there are words that do
fit into that pattern, and you may see (or use) them in print,

The same kind of things happens all the time with new verbs
in English, and in cases where the two are clearly not phonologically
related.  For example, I've noticed with several people that the
verb "to wing" (i.e., to do something that requires preparation
without any preparation, and on the spot) doesn't fit comfortably
into the past tense.  The two relevant patterns are as follows:

X                 <->    Xd
V                 <->    V
"Z (pres.)"  <->     "Z (past)"

X(1syl., V=[i])   <->    X(V=[V])
V                        <->    V
"Z (pres.)"          <->     "Z (past)"

Excuse the crudity of the rule, but no one ever formalized how
to do this type of thing using Bochner's framework.  What (4)
says is that you have a monosyllabic word whose vowel is [i]
(or [I], for those who speak a non-SoCal dialect) in the present
tense and [V] in the past tense.  Despite the spelling and the
phonetic realization of what phonologically is a short vowel in
"wing", "win/won" is an example.  Present/past participle examples
also can fit into this pattern, with "sing/sang/sung", "sink/sank/sunk",
"ring/rang/rung".  That pattern would look something like this:

X(1syl., V=[i])   <->    X(V=[e])   <->    X(V=[V])
V                        <->    V                 <->    V
"Z (pres.)"          <->     "Z (past)"    <->     "Z (past part.)"

(Again, your vowel in the second step may be [&].)

When it comes to deciding where "wing" fits, several factors
come into play.  First, "winged" is always going to work, but I
think speakers are uncomfortable with this one because there
are so many words that end in "ing" (or "inC[=velar]") that are
irregular: swing, ring, cling, bring, fling, sing, sink...  (I almost
wrote "clink"--that's another iffy one.)  That leaves two possible
patterns: wing/wung/wung and wing/wang/wung.  First, you
can pick out an even more general pattern that captures part of
(5) and (4):

X(1syl., V=[i])   <->    X(V=[V])
V                        <->    V
"Z (pres.)"          <->     "Z (some type of past tense form)"

So even though a past participle isn't the simple past, it's a form
that is used with the past tense in some way.  That might make
the pattern in (4) stronger than the one in (5).

Finally, if you get more specific, you can come up with a pattern
like this:

X(1syl., V=[*wiN])   <->    X(V=[*wVN])
V                                <->    V
"Z (pres.)"                  <->     "Z (past)"

That is, there is an extant pattern where a word that at least
ends in "-wing" becomes "-wung" in the past tense--and that
one is "swing".  Since it's such a common word, it very well may
be the model that "wing/wung" is based on.

So, back to conlanging.  How does this translate?  An easy way
is with novel forms or borrowings.  If you have all your patterns
listed, you can model how a speaker will think of that word.  If
a speaker encounters a new word that can fit whatever the
regular pattern is for borrowings as well as an irregular pattern,
you can see these types of effects popping up.  A couple quick
natlang examples:

kitako "base" <-> vitako "bases"

Borrowing from Arabic:
kitaab "book" <-> kutub "books"

Swahili form:
kitabu "book" <-> vitabu "books"

Zism "body" <-> ?aZsaam "bodies"

Borrowing from English:
film <-> films

Arabic form:
film "film" <-> ?aflaam "films"

In one of my languages, Gweydr, there are a bunch of different
plural patterns--enough that I still haven't managed to finish
listing them on this page:

As I may have mentioned once before (somewhere), none of
my language have concultures, but they do have conconcultures
(i.e., I pretend, for linguistic purposes, that they have a conculture,
but I never sat down and "created" one, or wrote it down, and
I don't ever plan to).  Conconculturally, Gweydr is spoken in a
place that's next to (or down the hill from) where an unrelated
language called Sidaan is spoken:

Because they're in close proximity, Gweydr borrows words from
Sidaan, and vice versa.  One such word is /naq/, "pot" (in Sidaan
it just means "pot"; in Gweydr, it's a specific type of pot--the ones
that Sidaan speakers make which differ slightly from the usual
pots Gweydr speakers have).  The Sidaan form is fairly simple
and regular:

naq "pot" <-> naqa "pots"

When borrowed into Gweydr, the /q/ > [k], and the vowel
specifically becomes a [-ATR] low back vowel.  The ordinary
plural pattern would be this:

nak "pot" <-> nakis "pots"

But two things kind of scream out for an irregular pattern.  First,
it's a monosyllabic word with a vowel of [a] (which, for simplicity's
sake, is [-high, +back, -ATR]), and it ends in [k].  The "usual" plural
marker is /-ks/, and many monosyllabic words whose vowel is
back have a fronted version.  So while the technically correct version
of the plural is shown in (11), what you often get (and I'll show
some other cases to fully illustrate the pattern) is this:

Gweydr form:
nak "pot" <-> n&k "pots"
kInak "pot (acc.)" <-> kIn&k "pots (acc.)"
t&n&k "pot (inst.)" <-> t&n&ks "pots (inst.)"
nEn&k "pot (com.)" <-> nEn&ks "pots (com.)"

The pattern in (11) is declension I, and the pattern in (12) is a
variant of declension IB.  It's characterized by a fronted vowel
in the plural for the grammatical cases, and a fronted vowel
in the singular plus a fronted vowel and a plural suffix in the

Anyway, by listing all the patterns, you can see where novel
forms can fit in, and also how words can change over time.
(E.g., if irregular words gradually become regular, then there
will be fewer and fewer tokens in a given pattern, which can
make the pattern weaker, which will then affect how novel
words are fit into the lexicon, etc.)  It also doesn't necessarily
need to be the case that one form is a phonological reduction
of an earlier form, but it can be, of course.

"A male love inevivi i'ala'i oku i ue pokulu'ume o heki a."
"No eternal reward will forgive us now for wasting the dawn."

-Jim Morrison