This article surveys empirical and theoretical work on Tense-Aspect-Mood (‘‘TAM’’) based split ergativity, and offers an account for how it arises. While these splits are typically assumed to represent a unified phenomenon, I demonstrate that non-ergative portions of split systems exhibit different patterns. I argue that these patterns reflect at least two different triggers of split ergativity: (i) non-perfective aspects are more likely to be built on complex auxiliary constructions, and (ii) imperfectivity is associated with demoted objects or lower transitivity. Both causes trigger the same result: in the ‘‘split’’ portions of the grammar the transitive subject is not marked with ergative case because it is not a transitive subject. This structural account of split ergativity allows us to avoid positing variable feature inventories on the same functional head (cf. Ura 2006), and also provides a straight-forward account of the so-called ‘‘counter-universal’’ splits (Gildea and de Castro Alves 2010), which cause problems for purely functionalist accounts (e.g. DeLancey 1981). Furthermore, it is shown that the factors which trigger these splits are not limited to ergative languages, but are present cross-linguistically—they are not visible in nominative-accusative systems because (by definition) there is no visible difference between transitive and intransitive subjects. The prevalence of splits in ergative systems is thus not taken to reflect any deep instability of ergativity.