[tentative explanation]
Proto-Tallic had a number of postpositions which were preserved in Talmit and further enlarged by compounding. The core case postpositions were
For the nominative and accusative, there were corresponding modified forms *wa and *nujə. They were originally what one could call an ergative-like aspect marking system. In Talmit this has evolved into a noun/state system. In Kymna, however the modified postpositions came to mark affectedness.
To demonstrate this, let us take the example
(1) *Pan-ja kojəba-nu (a)khagun
child=nom apple=acc eat
’The child eats an apple (habitually)’
The accusative marked by *nu conveyed a punctual event. Marking the object by *nujə instead gave durative
(1’) *Pan-ja kojəba-*nujə (a)khagun
child=nom apple=acc.dur eat
’The child is eating an apple (durative)’
Literally: ’The apple is experiencing a prolonged influence by the child’s eating’. If the predicate was an intransitive verb, the subject was marked by *ja for punctual aspect, by *wa for durative aspect.
Now let us take an abstract object, e.g.
(2) *Pan-ja khalis-nujə pwosun
child=nom happiness=acc.dur feel
’The child is happy (durative)’
Talmit took this pattern as basis, deleted the verb and interpreted *khalis as ’state of happiness’. The posposition *nujə > nójo thus came to indicate the state a subject marked by *ja:
(2) Pan-ja halís-nójo
child=nom.stat happiness=stat
’The child is in a state of happiness’
Real objects which can be manipulated then always became marked by nu. The intransitive pattern
(3) *Pan-wa (a)khagun
child=nom.dur eat
’The child is eating (durative)’
was replaced by
(3) Pan-ja aχagendé-nójo
child=nom.act eat-cmpd-action.state=stat
’The child is in a state of eating’
with a participial form of ’to eat’, and itself became habitual:
(3) Pan-wa aχágun
child=nom.act eat
’The child eats (habitually)’
Kymna, on the other hand, took the pattern with real objects as basis.
(1’) *Pan-ja kojəba-nujə (a)khagun > Pan kyvonas sohum (with deletion of *ja) was reinterpreted as ’The apple is affected by the child’s eating’.
Instead of denoting aspect, *nujə came to denote affectedness, while a new aspectual marking system on the verb arose. The second pattern was completely transformed into
(2) Pan salydzumōn
child happy-pred
’The child is happy’
with the predicative form of salydzuma ’happy’. The intransitive pattern (3) *Pan-wa (a)khagun ’The child is eating (durative)’ was reinterpeted as
(3) Panva sohum
child-nom.aff eat
’The child is affected by eating’.