And now for the post-lexical rules (and the phoneme inventory).
Important here are element theory (which is how GP represents the phonological make up of sounds), and projection government.
In GP, sounds are composed of one or more 'elements'. Elements are either there or not there (so no +/- whatever as in feature based analyses of phonology), and each element is supposed to be phonetically interpretable in isolation - for example, the element |U| in isolation is the vowel /u/, but in combination with other elements can produce other rounded vowels or labial consonants, while the element |ʔ| in isolation is a glottal stop, but combined with other elements produces various stop consonants.
In Kantaranyan, I'm using the set of elements proposed here
(in the 'segmental structure' chapter). I've chosen to distinguish the palatals and uvulars from coronals and velars respectively by the use of headed place elements (underlined below). Note that other papers propose different analyses of certain segments - especially in the case of the coronals.
Kantaranyan presently has the following phoneme inventory:
|ʔh.U|-/p/, |ʔh.I|-/t/, |ʔh.A|-/k/, |ʔh.A
|ʔ@.U|-/m/, |ʔ@.I|-/n/, |ʔ@|-/N/
|h@.U|-/β/, |h@.I|-/l/, |h@.I
|@.U|-/u/, |@.UA|-/o/, |@.A|-/a/, |@.IA|-/e/, |@.I|-/i/
/ʔ/, /N/ and /ə̯/ appear in codas only (although note that [ʔ] does appear in onset position as an epenthetic consonant before an underlyingly vowel-initial syllable), while /s/, /ʃ/ and /h/ are limited to onsets. Consonants from either cannot appear as the final consonant in a consonant-final verb root.
I mentioned in an earlier post the notion of 'complexity'. This is fairly simple - a segment with more elements is more complex than a segment with fewer elements. (Note, however, that the source I linked to in this post suggests a slightly more complex approach whereby different elements have slightly differing properties, but this is not directly relevant to the phonology of Kantaranyan, so I won't go into detail on that.)
I also mentioned 'projection government', which is how GP deals with non-local processes like vowel harmony. Essentially, constituents with something in common (such as nuclei) can be 'projected' to a special phonological tier where the process operates, so that the government processes can be said to be local within the projection. The directionality of projection government is set in a given language by a parameter. In Kantaranyan it operates left to right. (More details on projection government here
To prevent this post from getting too long, the actual post-lexical processes I've set up for Kantaranyan are in a separate document: http://dl.dropbox.com/u/41280522/Postlexical%20rules%20%28Kantaranyan%29.docx
(Note that the suppression of place elements in the coda is motivated by the need for final codas to be simple enough to be governed by the vowel in the nucleus, and that the 'consonant harmony' rule is based on projection government. Feet and stress are still determined as detailed in this post
The example words in the document are /qatu/ 'make', /esome/ 'cloud', /ʃaʃu/ 'fifteen', /kat/ 'take' and /taʔsa/ 'tail'.
Finally, to round this post out, here's the romanisation:
/p/ - <p>, /t/ - <t>, /k/ - <k>, /q/ - <q>, /ʔ/ - <h>
/m/ - <m>, /n/ - <n>, /N/ - <n>
/s/ - <s>, /ʃ/ - <x>, /h/ - <h>
/β/ - <v>, /l/ - <l>, /ʝ/ - <y>, /ʁ/ - <r>, /ə̯/ - <r>
/u/ - <u>, /o/ - <o>, /a/ - <a>, /e/ - <e>, /i/ - <i>
Note that epenthetic [ʔ] is not romanised.
Notes on final <r> and <n>:
-In final position in nouns, the graphemes <r> and <n> represent /ə̯/ and /N/ respectively.
-In final position in verbal stems, <r> represents /ʁ/ and <n> represents /n/.
Romanisation is based on phonemes rather than the surface form. The exception to this rule is the Romanisation of morpheme-internal codas, which are romanised according to the surface form as follows:
[p˺] - <p>, [t˺]/[c˺] - <t>, [k˺] - <k>, [q˺] - <q>
[m] - <m>, [n]/[ɲ]/[ŋ]/[ɴ] - <n>
[w] - <v>, [j] - <y>, [ɰ]/[ʁ̞] - <r>
So, that's the new phonology! Comments and feedback are welcome.