|From:||Marcus Smith <smithma@...>|
|Date:||Tuesday, January 2, 2001, 21:35|
I'm taking a new approach to conlanging with Igassik: I've invented a new
theory of grammar which I believe to be fairly natural but not found in
real languages. I'm developing the Igassik particular grammar based on what
is allowed by this "theory". I hope it will come out looking somewhat like
a real language, but still be pretty exotic. This is the morpho-phonemic
aspect of the grammar.
Each syllable must be able to fit into the following template CCVCC.
In roots and stems, only the V is obligatory, all C's are optional for a
A word may have up to two syllables in it: CCVCCCCVCC.
There are lots of rules about what kind of consonant may occur adjacent to
Vowels may never be adjacent.
Basically, the consonants follow the Sonority Heirarchy and a Place Heirarchy.
Sonority Heirarchy: the closer you get to the peak of the syllable, the
more sonorous the consonant must be. For that reason, each syllable must
conform to the following pattern.
Stop > Fric > Nas > Liquid > Glide > Vowel < Gl < Liq < Nas < Fric < Stop
This means that you can have onsets such as kl-, br-, tm-, fw-, and so
forth, but not the opposite order: *lk-, *rb-, *mt-, *wf-, etc. The reason
is that the legitimate examples follow the Sonority Heirarchy, with the
more sonorous material closer to the vowel, while the illicit ones do not.
Codas are the exact opposite. You can have -lk, -rb, -mt, etc but not *-kl,
*-br, *-tm, etc. Once again, this is because the more sonorous material
must be closer to the vowel.
The Place Heirarchy is also defined as you approach the peak of the
syllable. In this case, certain places of articulation must occur closer to
the peak than others. The relevant heirarchy for this is:
Labial > Velar > Alveolar > Dental > Oral < Den < Alv < Vel < Lab
By "Oral" I mean vowels and glottals which do not really have a place of
articulation. I also neglect Palatals, because in Igassik the only palatal
/j/ never occurs next to anything that would make the Heirarchy relevant.
That is, it never occurs adjacent to /w/, the other only glide.
The following are possible onsets: pt-, kt-, fs-, xT-, etc.
The following are possible codas: -tp, -tk, -sf, -Thx, etc.
Anything that is a permissible onset cannot be a coda and vice versa.
Where the Sonority Heirarchy and Place Heirarchy conflict, the Sonority
Heirarchy always wins. For example, -ft violates the Place Heirarchy
because a labial is closer to the peak than an alveolar, but it satisfies
the Sonority Heirarchy because the fricative is closer to the peak than the
stop. The opposite order *-tf would be invalid. Basically, this means that
the Place Heirarchy is only relevant in situations where both consonants
have the same manner of articulation.
Now for some fun word structure. I'll give a general outline, then give
some specific examples.
Each word is composed of three separate "Tiers": the skeletal, consonantal,
and vocalic. The skeletal tier defines the order of C's and V's for the
word: [CVC], [CCVCV], etc. The consonantal tier lists all the consonants in
the order they occur: [k,s,t,l], [b,g,n], etc. The vocalic tier lists are
the vowels in the order they occur: [ae, e], [o,u], etc. The order of
elements on a tier cannot be changed. So a word with the consonantal tier
[k,s,t,l] could never form a word like *katsol, since that would require
altering the order of consonants.
Words are formed by merging these three tiers together. Each consonant must
"attach" itself to a C in the skeletal tier and each vowel must attach
itself to a V. As I said above, the order of the tiers cannot be
rearranged. All the slots of the skeletal tier must be filled, and all the
elements of the consonantal and vocalic tiers must be attached. Thus, there
is only one possible order for each merger. Here are some examples.
throw - [CVCCVC];[k,s,t,l];[u,o] -> kustol
count - [CVCVC];[n,v,p];[a,o] -> navop
walk - [VCC];[N,k];[ae] -> aeNk
vomit - [CCVCCV];[k,Th,m,b];[a,u] -> kThambu
cook - [CVCV];[Dh,l];[a] -> Dhala
The last word Dhala 'cook' is interesting because the vocalic tier only has
a single segment, but the skeletal tier has two V's. Since ever V must be
filled, the /a/ is found in both V's. This make morphology much more
Affixes are just slightly different from what was described above. The main
differences are: 1) they may only have one syllable; 2) the skeletal teir
(which I write in lower case) also specifies whether it is a prefix or suffix.
PAST - [-vc];[t];[a] -> -at
FUT - [-vcc];[l,n];[i] -> -iln
CAUSE - [vc-];[s];[i] -> is-
When an affix is attached to a stem, the skeletan tiers merge.
walked [VCCvc];[N,k],[t];[ae],[a] -> aeNkaet [a -> ae because of vowel harmony]
will walk [VCCvcc];[N,k],[l,n];[ae],[i] -> aeNkiln
make walk [vcVCC];[N,k],[s];[ae],[i] -> isaenk
This looks like ordinary affixation, and it is. The fun comes when
disyllabic stems are considered. I said above that words may have a maximum
of two syllables. This doesn't change just because an affix has been added.
What happens in these cases is that the vowel of the affix aligns itself
with the closest vowel of the stem. Since only one vowel is permitted per
syllable, the stem and affix vowels merge into one slot. That is, a stem
CVCV plus an affix vc will merge to give CVCV/vc, where V/v refers to one
slot which receives vowels from the stem and the affix. The two vowels then
assimilate to each other; the stem vowel changes in height to match the
height of the affix vowel. ae + i = e; u + a = o; i + i = i; etc.
vomited [CCVCCV/vc];[k,Th,m,b],[t];[a,u],[a] -> kthambot
will vomit [CCVCCV/vcc];[k,Th,m,b],[l,n];[a,u],[i] -> kThambuln
A special case is words like Dhala which have two V's but only one segment
on the vocalic tier. Since one of the stem vowels will merge with the affix
vowel, there will still be two vowel slots to be occupied. Since the root
has a vowel in the vocalic tier and the affix does to, then there is no
need for the root vowel to occupy both V slots on the skeletal tier.
cooked [CVCV/vc];[Dh,l],[t];[a],[a] -> Dhalat
will cook [CVCV/vcc];[Dh,l],[l,n];[a],[i] -> Dhaliln
make cook [CV/vcCV];[Dh,l],[s];[a],[i] -> Dhisla
Note that the affix is an infix in the causative form even though it was a
prefix for aeNk 'walk'. This can also happen with suffixes.
counted [CVCV/vcC];[n,v,p],[t];[a,o],[a] -> navatp
will count [CVCV/vccC];[n,v,p],[l,n];[a,o],[i] -> navulp
make count [CV/vcCVC];[n,v,p],[s];[a,o],[i] -> nosvop
The future tense word [navulp] had a merged structure including -ccC, which
is too many consonants for a real word. Therefore, the segment of the affix
furthest away from the vowel is deleted. If both the root and the affix had
contained consonant clusters, then both of the affix consonants would have
Somethimes affixation can create sequences that are not permitted by the
language. When this happens, there are several solutions. One is deletion,
but that is only done when there is no other way to fix the problem, like
with navulp 'will count' above. The other repair strategy is metathesis. I
said above that the segments of a tier cannot be rearranged, but that does
not prevent them from changing orders in relation to segments on other tiers.
make vomit [CCV/vcCCV];[k,Th,m,b],[s];[a,u],[i] -> *kThosmbu
This word violaties the Sonority Heirarchy since /m/ is never the closest
segment to a peak. Therefore, the c of the affix merges into the root
skeleton at a different location.
[make vomit [CCv/vCcCV];[k,Th,m,b],[s];[a,u],[i] -> kThomsbu
Both skeletal tiers maintain their linear ordering through what looks like
metathesis on the surface. Metathesis can only occur when the end result
would permit the affix consonant to be in the same syllable as the affix vowel.
threw [CVCCV/vCc];[k,s,t,l],[t];[u,o],[a] -> kustalt
will throw [CVCCV/vccC];[k,s,t,l],[l,n];[u,o],[i] -> kustull -> kustul
make throw [CV/vcCCVC];[k,s,t,l],[s];[u,o],[i] -> kusstol -> kustol
If you made it this far, I'm very impressed. What do you think?
"When you lose a language, it's like
dropping a bomb on a museum."
-- Kenneth Hale