Middle voice

Abstract

The middle differs from other argument changing operations in that it leads to different readings, in particular anticausative, reflexive, causative reflexive, and modal (facilitative). The readings depend on the semantic class of the base and the sortal properties of the highest argument. Middle verb forms that correspond to an active verb form generally display a reduced argument structure. Some readings also involve changes in the event structure. On the basis of data from Fula, I argue that the conceptual motivation of middle marking is the existence of actions that can be semantically modelled as involving an agent and a patient argument although in most cases, only one participant is involved in the event. Middle verb forms share the semantic representation of the corresponding active verb form. In contrast to active verb forms, middle verb forms do not project the agent argument into the argument structure. According to this analysis, the different middle readings are due to different interpretation strategies that can apply if the default initiator/controller of an action verb is not projected into the argument position which represents the actual initiator/controller of the event. Finally, the readings of the middle in Fula are compared with readings that can be observed in other languages and a set of parameters that accounts for the distribution is proposed.