Cross-linguistic empirical analysis of constraints on passive

Passive constructions are examined cross-linguistically (English, German and Chinese) in an exploration of factors that contribute to their relative acceptability. Passives are universally dispreferred to their active counterparts. We describe a methodology for empirical analysis of factors that modulate differences in acceptability between passive and active variants of the same propositional content. Properties of predicates and arguments are considered independently and in tandem. Properties include pragmatic qualities (e.g. whether a predicate would normally be perceived as negative), semantic features (e.g. whether a referent is animate or inanimate) and structural (e.g. whether a nominal is constituted by a sequence of words). Acceptability judgements from a general audience were elicited in response to contextualized sentences using a magnitude estimation technique. As passivization is clearly related to argument structure and linking of syntactic roles to semantic roles, and as it is ordinarily assumed that a hierarchy of thematic roles interacts with the acceptability of passive forms, items were selected with reference to these theoretically relevant factors. Independently of our analysis of data resulting from non-linguist judgements, we also evaluated trained linguist categorization of the sentences arguments using a representative set of theoretically motivated semantic role labels. In an experiment employing the methology, only the new/givenness of the subject to the discourse had a significant effect cross-linguistically. A consistent effect also emerged in connection with the affectedness of the object. Differentially, within each language a range of other factors emerged as significant.