Does passivization require a subject category?

Passives are usually described in terms of subjects: they shift subject status from the semantic agent of a clause to a patient or other argument. We might accordingly expect that languages without a robust subject category should lack passives. Here passives are examined in a language with at best only marginal evidence of a subject category, Central Pomo of California. The construction alters argument structure by eliminating the grammatical agent from the clause, but has none of the other syntactic effects expected of passives. It does not promote arguments to subject status. There is no shift in pronominal shape or case marking: patients remain coded as grammatical patients, and obliques as obliques. It is not exploited for purely formal syntactic purposes, such as ensuring coreferentiality of arguments in clause combining. It does, however, serve the same kinds of semantic, pragmatic, and discourse functions as passives in many other languages. It eliminates mention of generic, unknown, irrecoverable, unimportant, predictable, stereotypical, and non-topical agents. The density of passives in spontaneous Central Pomo speech indicates these functions are more than sufficient to justify its central place in the grammar.