Mirror principle and split VoiceP

Abstract This paper demonstrates that the layered affixation of passive and causative morphemes in Mongolian and Japanese mirrors the Split VoiceP structure of passive and causative clauses, where the affixation of a voice suffix signals the introduction of an argument by a voice head spelled out by that suffix. This head-argument correlation, ensuring the morpho-syntax mirroring of voice suffixes, entails that the voice domain comprises separate projections that are split out of VoiceP. Clauses are built by introducing arguments as potential subjects (sbjs) through voice heads of the same substance, where a last-introduced sbj is promoted to the nominative position, with others, if any, remaining non-nominative within the voice domain. From this, it follows that Voice is a sbj-introducing, not merely argument-introducing, head, and that the interconnection of passives and causatives as voice constructions lies in the fact that both are derived over the Split VoiceP structure. Consequently, voice constructions can be accounted for in a unified way, without the need of postulating dedicated heads such as Passive and Cause, and the clause building mode is reducible to introduction of arguments as sbjs, which ultimately comes down to Free Merge.