Re: Arabic and BACK TO Self-segregating morphology

On Mon, 19 Dec 2005 09:12:08 -0800, Gary Shannon <fiziwig@...> wrote:

>--- Wesley Parish <wes.parish@...> wrote:
>
><snip>
>
>> All Semitic languages to the best of my knowledge,
>> use a triconsonantal with
>> two vowels to express basic verbal ideas, adding
>> complexity and additional
>> consonants and vowels to express more complex ideas.
>
>My plan was for three consonants, each followed by a
>vowel or dipthong (possibly modified by "-n"), and an
>optional initial vowel or dipthong. Thus: "pkt" could
>be no shorter than "pokatu" nor longer than
>"ainpuinkiantiun" (although in practice none would
>really be THAT long). But in every case the vowels
>would be fully written out.
>
>There would be no inflections, and any change of vowel
>would mean a change in the meaning, not in case,
>tense, plurality, etc. Those would be marked by
>particles.
>
>The vowel changes would be uniform and and consistent.
>Thus if -a-i-a was the primary noun and a-a-in-a was
>the negative of the primary noun the with "nlj" we
>have "nalija" = knowledge and "analinja" = ignorance;
>with "wlt" we have "walita" = wealth and "awalinta" =
>poverty. Thus knowing the roots and patterns one can
>coin a new word or recognize a word not encountered
>before. So if we know that "sakisa" is success then
>even if we have never seen the word "asakinsa" we
>would know what it meant. And we would also know how
>to form "to succeed", "successful", successfully",
>"successful-person", "to fail", "unsuccessful",  etc.
>
>Also, since every base word is three consonants long
>when we encountered a four-consonant word we would
>know that a prefix or suffix had been added, and by
>the vowel patterns we would know which. And likewise,
>five-consonant and longer words would, by their vowel
>patterns be easily broken down into their roots and
>affixes according to vowel patterns, thus retaining
>the self-segregating property.
>
>--gary
>=========================================================================
Hi.

We had a discussion about how several subgroups of Afro-Asiatic languages,
including Semitic languages, handle this sort of thing, some months ago.

You could handle some of the problems by having more than one paradigm;
several of the natlangs in the group in question do that.

You know how Latin etc. has first declension, second declension, etc. and
first conjugation, second conjugation, etc.?

Since the paradigms you have in mind will include both nouns and verbs, and
perhaps other categories as well, you will not want to call
them "declensions" (which implies they are noun-ish) nor "conjugations"
(which implies they are verb-ish).

The "prototypical" root -- roots in the First Paradigm -- consists of three
consonants; C1 C2 C3 -- such that no two consecutive consonants are "too
similar".  (For conlang purposes, you could say that C2 can have neither
the same place-of-articulation nor the same manner-of-articulation as
either C1 or C3.)  They must also have the properties that; C1 is not the
same as the last consonant of any prefix; C3 is not the same as the first
consonant of any suffix; and none of C1 nor C2 nor C3 is a semi-vowel, nor
a "glottal" (like /?/) nor a "laryngeal" (like /h/).

The "prototypical" member of the First Paradigm for root C1 C2 C3 fits one
of the following four patterns;
a. (optional prefix) V1 C1 V2 C2 V3 C3 V4 C3 V5 (optional suffix)
b. (optional prefix) V1 C1 V2 C2 V3 C2 V4 C3 V5 (optional suffix)
c. (optional prefix) V1 C1 V2 C1 V3 C2 V4 C3 V5 (optional suffix)
d. (optional prefix) V1 C1 V2 C2 V3 C3 V4 (optional suffix)

Each of V1 through V5 stands for an _optional_ vowel, which takes its
value, if it has one, independently from all the rest.

There can be several secondary paradigms; they can be completely filled
out, but by different patterns.  A secondary paradigm will handle any root
C1 C2 C3 that fails _exactly_ one (1) requirement to be a root of the First
Paradigm.

1. C1 is the same as the last consonant of some prefix.
(A diffferent secondary paradigm for each possible such consonant).
If C1 is the same as the last consonant of some prefix(es), it may make it
difficult to divide that/those prefix(es) from the root.

2. C3 is the same as the first consonant of some suffix.
(A diffferent secondary paradigm for each possible such consonant).
If C3 is the same as the first consonant of some suffix(es), it may make it
difficult to divide that/those suffix(es) from the root.

3. C1 or C2 or C3 is a semi-vowel;  this would limit the distinguishability
of the different vowels that may be used before and after the semivowel.
There would be a different secondary paradigm for each combination of a
position (C1 or C2 or C3) and a semi-vowel.

4. C1 or C2 or C3 is a glottal or laryngeal consonant; these could
disappear between certain pairs of vowels.  There would be a different
secondary paradigm for each combination of a position (C1 or C2 or C3) and
a glottal or laryngeal.

5. C1 is the same as C2.  This would make patterns b. and c. hard to tell
apart.  Only one secondary paradigm is needed for this possibility.

6. C2 is the same as C3.  This would make patterns a. and b. hard to tell
apart.  Only one secondary paradigm would be needed for this possibility.

7. A root could be four consonants long; C1 C2 C3 C4.  Assuming it shows no
other variations from the "prototypical" root -- that is,
C1 != C2 != C3 != C4,
none is a semivowel nor a glottal nor a laryngeal,
and C1 is not the same as the last consonant of any prefix,
nor C4 the same as the first consonant of any suffix,
a single secondary paradigm could handle all of these.

-----

There might also be "tertiary paradigms".  These would be minor paradigms,
like the secondary paradigms, except, they would either not be completely
filled out, or, completing them might require "suppletion" -- where one
root supplies many of the entries, and a different root, with no
phonological hint that it is related, supplies the rest of the entries.

8. You would need "tertiary paradigms" to handle cases where the root is
only two consonants long (provided they were distinct, neither was a
semivowel nor a glottal nor a laryngeal, the first one didn't match the
final consonant of any prefix, and the last one didn't match the initial
consonant of any suffix.)

You would also need "tertiary paradigms" to handle cases where the root
violates _exactly_ two (2) of the requirements to be in the First Paradigm;

for example (I can make no guarantee this list is exhaustive);

C1 is the same as the last consonant of some prefix, and C1 is a semivowel
or glottal or laryngeal;
or,
C1 is the same as the last consonant of some prefix, and C1 is the same as
C2;
or,
C1 is the same as C2, and C1 and C2 are some semivowel or glottal or
laryngeal;
or,
C1 is the same as C2, and C2 is the same as C3;
or,
C2 is the same as C3, and C2 and C3 are some semivowel or glottal or
laryngeal;
or,
C2 is the same as C3, and C3 is the same as the first consonant of some
suffix;
or,
C3 is a semivowel or glottal or laryngeal, and C3 is the same as the first
consonant of some suffix;
or,
C1 is the same as the last consonant of some prefix, and C2 is the same as
C3;
or,
C1 is the same as C2, and C3 is the same as the first consonant of some
suffix;
or,
C1 is the same as the last consonant of some prefix, and C3 is the same as
the first consonant of some suffix.

----

Any root that failed three or more of the requirements to be in the First
Paradigm, would just have to be Irregular.

----

The "prototypical" prefixes and suffixes would have a single consonant
("servile consonants"); some could have two consonants.
"Prototypical" prefixes and suffixes would look like (V1) C (V2);
prefixes and suffixes could also look like (V1) C1 (V2) C2 (V3).

Here, the (Vn)'s are _optional_ vowels, varying in value independently of
each other.

-----

The list of prefixes and suffixes would have to be "small" and "memorized".
(not necessarily consciously memorized, and not necessarily "small" in the
opinion of an L2 learner.)

One easy thing to do might be to take a certain minority set of consonants
("servile consonants") and declare that prototypical prefixes all had one
of those as their final consonant; take a (possibly-different) minority set
of "servile consonants" and declare that prototypical suffixes all had one
of those as their initial consonant.
Keep exceptional affixes extremely few.
Then don't begin many roots with consonants from the first set, nor end
many roots with consonants from the second set.

This will make it easier to split a word into prefix-stem-suffix.

----

You could declare the two-consonant roots to be very ancient, and also very
commonly used; that would help explain why they're shorter, and stay in the
language that way in spite of any regularizing impulses.

Four- and five- consonant roots could be lexicalized compounds, one or both
components of which is a two-consonant root; or, lexicalized members of
older paradigms, (a three-consonant root with a one- or two- consonant
prefix or suffix, or both a one-consonant prefix and a one-consonant
suffix), which have "broken away" from their original paradigms (as "went"
broke off from the "wend" paradigm to join the "go" paradigm) to take on an
independent life of their own.

-----

If all of the vowels (V1 through V5 etc.) in the above patterns are  short
monophthongs, then, diphthongs and long vowels will occur only at component
boundaries (prefix/stem or stem/suffix), or, (and possibly triphthongs
could occur this upcoming way too), where a consonant is a semivowel.

----

Example:

Suppose there are 4 manners of articulation and 4 places of articulation,
yielding 32 distinct consonants -- 16 voiced and 16 mute.  Pick one place
and one manner of articulation and declare any consonant at that place, or
any consonant in that manner, "servile"; that gives 14 servile consonants
and 18 others.

Now there are 18 choices for C1 and 18 choices for C3.  C2 can not be in
either the place or the manner of C1, nor in either the place nor the
manner of C3; if C1 and C3 are neither at the same place, nor of the same
manner, as each other, that still leaves two possible places, and two
possible manners, for C2; so there are at least 8 choices for C2 (there
could be 12 or 16 if C1 and C3 were similar, depending on how similar they
were).

So there are at least 18 * 8 * 18 = 2592 roots.

Now imagine there are just two vowels; and imagine all prefixes and
suffixes have a single consonant.

There are, then, 3 * 14 * 3 = 126 possible affixes. (Each optional vowel
can either be not there, or have either of two values.)

Let's assume 63 are prefixes and 63 are suffixes.

Then each of patterns a, b, and c yields
64 * 3^5 * 64 = 995,328 different words for each root.
(no prefix or one of 63 prefixes) *
(five places for optional vowels) *
(no suffix or one of 63 suffixes).

That's more (since I left out pattern d) than
4,096 * 243 * 3 = 2,985,984 words per root;
which comes more than to 7,739,670,528 words in just the First Paradigm.

(I'm personally willing to bet you won't get around to using them all.)

All this for only
2592 + 729 + 126 = 3447 definitions.
2592 definitions at one per root;
729 = 3 * 243 definitions for the patterns a, b, and c;
126 definitions at one per affix.

----

Tom H.C. in MI