A new perspective on grammatical voice
Masayoshi Shibatani
Rice University and Kobe University

Modern understandings of voice phenomena in terms of diathetic alternation (or change in the relationship between the subject and object of a verb) fail to align such a familiar voice opposition as active and passive with various other phenomena including the active-inactive/stative opposition, split case-marking phenomena (including split-ergativity), and the direct-inverse system. What is explored here is a direction toward a new conception of voice that permits a coherent and comprehensive treatment of various phenomena. Such an effort requires a better understanding of the conceptual parameters involved in the formation of voice categories. With this in mind, I offer the following as an initial characterization of what voice is all about.

(1) Voice is a grammatical category pertaining to the evolution of an action.

That is, voice is a category of opposition deriving from the grammaticalization of various aspects of conceptualization of an action. That a voice phenomenon such as the active-passive opposition revolves around the notion of action is well-known from the impossibility of expressing this voice opposition with an English expression like John resembles Bill (*Bill is resembled by John), but this simple observation has not been hitherto pursued and developed.

The characterization proposed above contains two basic parameters regarding the evolution of an action-its origin and development. Each of these consists of a number of sub-parameters. This paper will demonstrate how these parameters allow us to bring together and compare three phenomena that appear to be quite disparate, namely the active-passive opposition, the direct-inverse system, and the split case-marking phenomena. These phenomena all pertain to the sub-parameters of where the action in question originates and where it terminates.

It is further shown that the universals of voice constructions can be arrived at in terms of the correlation between event configuration (i.e., the nature of the origin and the terminal point of an action) and the form of the voice construction (i.e., the markedness relation between the relevant voice constructions).