No abstract

This paper addresses the question whether the type of passive that a language has is or is not related to apparently independent properties of the language, such as object marking, found in Bantu languages. The theory of object asymmetries (Baker, 1988a;Bresnan and Moshi, 1990) makes the fundamental claim that passive types are just one of the possible manifestations of an underlying asymmetry among objects. I argue that facts that might be taken as evidence against this claim turn out to have an analysis that is consistent with it and thus preserves important predictions made by the theory of object asymmetries. The analysis of these and other facts that reveal asymmetries among objects supports the assumption that the theory of object asymmetries relies heavily on a level of representation in which arguments are ranked by prominence, as determined by their thematic roles. The conclusion that an adequate theory must view the type of passive not as a property of the passive morpheme (contra Woolford, 1993), but as a manifestation of an underlying feature of the language, argues against the view that requires parameters to be morpheme-based: at least in some cases, parameters must be language-based. Passive structures with multiple internal arguments can be classified intotwo types: the alternating passive, which allows a choice of arguments as the passive subject, and the nonalternating passive, which does not allow a choice. An important tradition of research on multiple object constructions, in a variety of frameworks, from Perlmutter and Postal (1974, 1983), to Baker (1988a) and Bresnan and Moshi (1990), among others, tries to relate the type of passive that a language has to independent observable properties of the language, such as the behavior of object markers in Bantu languages. (Object markers, also known as 'object prefixes' and 'object clitics', are affixes that function as object pronouns and, in some languages, optionally also as object agreement markers (see Bresnan and Mchombo, 1987).) Woolford (1993), however, presents problems for this research program and proposes a theory in which the type of passive is independent of other observable properties of the language.The main goals of this paper are to show that the typology of passives * I am grateful to

A recent paper by Baker (1988b) claims that certain asymmetries in the syntax of applicatives in Chicheffa derive from an asymmetry in the assignment of the beneficiary and instrumental theta-roles. In this paper we adduce three types of evidence against this theta theoretic asymmetry. First, the facts of Wh-movement, which he presents as evidence for this asymmetry, on closer inspection, are shown not to follow from it. Second, we demonstrate that his claim that beneficiary applicatives cannot be formed from intransitive verbs is incorrect, And, third, we present the facts of locative applicatives, which behave in a manner opposite to what would be expected given the conception of Theta Theory assumed by Baker (1988b).

The observation that morphological structure mirrors the derivation of associated syntactic operations (the Mirror Principle of Baker 1985) needs to be built into linguistic theory. This paper argues against a syntactic explanation of the Mirror Principle, that is, against assuming that morphological structure is the product of a syntactic derivation. Instead, it assumes that all morpheme combinations take place in the lexicon and explains the Mirror Principle by including some syntactic information in the lexical entries of affixes.Baker (1985a) observed that natural languages are subject to the Mirror Principle: morphology and the syntax associated with that morphology must have parallel derivations. In other words, the order in which morphemes are combined in a word must match the order in which the syntactic processes associated with those morphemes take place.