Proto-Siye (PS) to Standard Siye (SS):
The Proto-Siye case system was tripartite.
Erg -ya, -na, -yam
Dative/Lative -tu (animate), -su (inanimate)
Instrumental -e (animate), -ki (inanimate)
As a tripartite language, the three main verb voices were active, passive, and antipassive. The active voice, as the default voice, was unmarked and transitive and its aspect could be perfective or imperfective. The passive voice deleted the ergative agent, promoted the accusative patient to the intransitive case, reduced the valency of the verb to intransitive and only had a perfective aspect, marked by -ne. The antipassive voice deleted the accusative patient, promoted the ergative agent to intransitive case, reduced the valency of the verb to intransitive and only had an imperfective aspect, marked by -me. The agent could be reintroduced to the passive by using the instrumental case -e or -ki, while the patient or goal could be reintroduced to the antipassive by using the dative/lative -tu or -su.
Proto-Siye developed a vowel dominance hierarchy, which affected the accusative -a and the instrumental -e. Words that ended in vowels stronger than /a/ deleted the accusative suffix; those stronger than /e/ eliminated the instrumental suffix. Now a patient in a transitive clause could have either -a, an accusative ending, or -#, identical to that of the intransitive case and therefore a new case – the absolutive. In the passive voice, then, the case of the subject was absolutive rather than intransitive. If the passive sentence, however, contained a (by default, animate) noun in the instrumental case which lacked overt case marking, it looked as though there were two absolutive nouns. Since the instrumental noun was animate and agentive, as well as in topic position, the Proto-Siye speakers re-analyzed the zero-marked instrumental as a zero-marked nominative. This analysis solved the double-absolute problem, but it also changed the valency of the verb back to transitive, and the marker -ne, which used to indicate a verb which had intransitive valency, passive voice, and perfective aspect, now indicated a verb which had transitive valency, active voice, and perfective aspect. Since the zero-marked form of the verb was already transitive and active, the marker -ne effectively distinguished only perfective aspect.
It did not take long before Proto-Siye speakers realized that, if -ne could mark the perfective aspect in such a construction, they could use -me in the same construction for the imperfective aspect. In this construction, -me lost its antipassive connotations and became limited to its aspectual meaning. This syntactical changed reduced the voices to two: active, with two aspectual markers, and antipassive, with one.
Then some Proto-Siye speakers, who were already using the indefinite pronoun 'mu' as a dummy agent in pseudo-passive constructions, began to use the antipassive with the perfective marker -ne on the analogy of the active voice. Since the less agent-like noun in the antipassive had an overt case marker -tu/-su, the shift to transitivity depended more on verbal semantics than syntactic shift. Usually only verbs of motion could increase valency, and even there it was erratic.Proto-Siye speakers began to use 'mu' as a dummy object with the former antipassive, and Standard Siye was born.
Thoughts? Corrections? References?