The voice system of Amharic

Abstract

This dissertation examines the voice system of Amharic. I have studied the voice features including the anticausative and a pair of causative functional items focusing on their selection, interpretation and syntactic projection.

The voice items display some interesting properties that made them worth an in­depth investigation. First, the single anticausative morpheme is associated with multiple constructions such as the passive, reciprocal, reflexive, middle and the like. This raises the foundational theoretical problem on the relation between form and meaning.

Furthermore, the causative items display quite striking contrasts with the anticausative in selection. The direct causative item filters verb classes. The indirect causative does a similar filtering, but mostly on the types of external arguments. The anticausative, on the other hand, imposes little selection restriction on the verb and argument types.

How the selection patterns are correlated with the interpretation and with the syntactic position of the voice items is the main concern of this study.

In this dissertation, I have proposed that all the crucial properties of the voice items, including the observed distinctions in selection, projection and interpretation, can be explained by understanding the voice items two types–those that belong to both the syntactic and semantic domains, and those belong to the syntax only.

The natural consequence of this understanding is that the syntax specific features could impose selection only within the syntactic component. They freely combine with all semantic classes of verbs. The interface features, on the other hand, impose selection both on the syntactic derivation as well as the semantic composition.

The causative items in Amharic are interface features—they merge with their semantic interpretations, as their syntactic categories. Because of this, they impose selection restrictions on the items they merge with not only in their syntactic properties but also on their semantic attributes.

The anticausative, on the other hand, is proposed to be a purely syntactic feature. That means, its distribution is fully determined by its syntactic category. Furthermore, I will argue that multiple decausative constructions associated with the anticausative morpheme such as the passive, reciprocal, reflexive, and middle, etc., are not coded into the morpheme itself. The morpheme comes semantically underspecified. The decausative constructions get associated with the morpheme only later in the derivation/composition due to syntax internal contextual factors. This is to mean that the passive, the reflexive, the reciprocal, etc., constructions are grammatically irrelevant in this language. They are simply allosemes that get assigned later in the semantic component.