Mike S:
> Hello folks,
>
> I am still tinkering with self-segregating morphologies,
> and, inspired by the root-forming system of Morneau's Katanda,
> and the basic morphology of Dublex and Vorlin, came up
> with the following hybrid model.  I don't think I'll be
> using it myself, but I submit it for your curiosity.

I'll offer a few criticisms/comments.

> C ::= (one of the following)
>   p t c=[tS] k f s h=[S] x b d j=[dZ] g v z y=[Z] w=[G] m n q=[N] l r
>
> V ::= a | e | i | o | u
>
> Particle := C V V
>
> Primitive ::= C V C
>
> Semiroot ::= C V
>
> Root ::= [ Semiroot ] Primitive
>   where [] indicates zero or more occurances
>
> Consonants and vowels are pronounced like their IPA equivalents
> unless indicated otherwise.  They appear freely in the
> three basic morph types as shown except /q/ may not start
> a morpheme. In diphthongs, /i/ and /u/ become glides; vowel
> pairs such as /ae/ are rendered as two syllables with any
> glottal consonant.  An unwritten buffering schwa occurs
> between words that would otherwise yield a geminate or
> an overly difficult consonant cluster.

This gives 20 x 6 = 120 monosyllabic/3-segment particles and
20 x 19 = 380 disyllabic/4-segment particles.

> Note that there are 21 consonants, but one can't be used
> initially, so we have 20 permutations in these cases.
>
> Particles can be defined as prefixes, suffixes, or neither,
> depending on whatever syntax gets cooked up.  There are
> 20 x 5 x 5 = 500 of these in our morpheme space, giving
> us a fair bit of flexibility.
>
> Primitives are the basic content words of the language.
> There are 20 x 5 x 21 = 2100 of these available.

The combination of one or more CV semiroots following
by a CVV particle has not been exploited in this scheme,
which is wasteful. You could treat disyllabic CVV as
further inisolable primitives, and allow semiroots in
combination with monosyllabic particles to yield further
complex particles.

> Semiroots are formed by clipping a consonant off the end
> end of a primitive.  There are only 20 x 5 = 100 distinct
> semiroots, each directly corresponding with 21 primitives.
> When being used in a root, a semiroot may assume ANY of
> the 21 meanings associated with it;  furthermore, the
> semantic relationship between the semiroot and primitive
> is not precisely defined.  If several semiroots are attached
> to a primitive, there is nothing indicating the grouping
> precedence among the root components.  Perhaps the best
> way to think of the compositional system here is to think
> of acronyms with a hundred letters to choose from instead
> of 26.  Acronyms are not reversable on sight, but they
> are terse and easily memorizable.

A drawback with this is that there are no cranberry morph
extensions of primitives, since every semiroot has one of
a hundred possible meanings. If none of those 100 meanings
is semantically appropriate, then one will have to be
chosen completely or relatively arbitrarily, which will
tend to dilute and contaminate the effectiveness of the
system.

> However, each unique root that gets created WILL be assigned
> an EXACT dictionary definition.  Furthermore, the primitive,
> which is the "head" component of the root, will always
> indicate the precise semantical category of the root.
> Thus if we know that /ful/ means "bird", then we will know
> that /buful/ indicates some sort of bird, even if we have no
> clue what /bu/- means (It could mean /bul/='blue', /but/='boot',
> etc.).  In effect, the semiroot merely hints at the meaning
> and serves a memorization tool for new roots; the primitive
> defines a semantic space or range within which every
> corresponding composite root must reside.
>
> The productiveness of this system would depend very much
> on the ingenuity of the designer in organizing which primitives
> yield the same semiroot and which yield different ones.
> However, by starting out with 2100 primitives we already
> have a moderate sized vocabulary established; there are
> 210,000 two-syllable roots available and 21,000,000
> three-syllable roots.  So it seems that there is quite
> a bit of room to maneuver.
>
> There are a LOT of ways this system could be modified to suit
> personal taste.
>
> It should be noted that this system is nearly identical
> to Morneau's system of semiroots and classifiers with
> a few differences:
>
> - Katanda uses about 125 classifiers; this system uses
> up to 2100 primitives.
>
> - Katanda has specific root-starters which must precede any
> other semiroot component; most root-starters are not associated
> with a classifier. In this system, semiroots are uniform
> and directly derived from the classifiers.
>
> - In Katanda, word-level segregation is not accomplished
> by wordshape alone--one needs to be able to identify the
> rootstarters and classifers.  This system self-segregates
> by wordshape alone:  all roots are of form [CV]CVC.

The scheme is ingenious and simple, and I can see how it could
be argued to be an improvement on Katanda.

But as a practical scheme it seems both wasteful and unnecessarily
verbose. Wasteful, because some meanings whose frequency and
basicness warrant their being assigned to a primitive (e.g. say),
do not require hundreds of different derived roots for different
sorts of saying (unlike bird, tree, etc.). unnecessarily verbose,
because other -- perhaps less elegant -- schemes could be more
concise.

--And.