Abstract
<jats:p>Alignment can change from accusative to ergative and from ergative to accusative depending on the morphosyntactic environment producing the change. Accusative‐to‐ergative changes take place most typically in intransitive clause types where accusative case is not available, so a semantic object surfaces with nominative case. Such constructions can be reanalyzed as ergative if an obliquely marked subject is also present, and the event can be interpreted as transitive and dynamic. Change in the other direction begins most commonly in antipassive constructions, where subjects have nominative case rather than ergative. If the clause can be interpreted as fully transitive, then the construction can be reanalyzed as a nominative/accusative clause. Consequently, accusative‐to‐ergative and ergative‐to‐accusative are both natural types of syntactic change, provided that the requisite syntactic environments and motivations for reanalysis obtain. Alignment change nearly always results in a split between ergative and accusative alignment, since only one syntactic environment is targeted by the first change. The most common type of split is conditioned by tense or aspect, ergative alignment exhibited in the past or perfective. This is in part because past or passive participle constructions are a common origin for the change from accusative to ergative alignment. Such participles themselves are a common origin for markers of past tense or perfective aspect. Another common type of split is NP split ergativity, in which pronouns high in animacy, such as speech‐act participants, always exhibit accusative alignment. This is unsurprising given the special licensing needs of such arguments across alignment types.</jats:p>