Typological research seeks to make valid generalisations about the diversity of human language. lt proceeds from observed linguistic traits, their distributions across geography and genealogy, and their relationships to other traits. Typological theories generate hypotheses, whose testing can prompt us to make more precise and systematic observations. Theories also propose explanations for observations. Corbett (2012), for example, generalises over extensive typological evidence to propose a theory of features, according to which, inflectional features are limited to only a handful of kinds. As in all sciences, theories in typology have apparent counter-exarnples. Rare linguistic traits, or rara (Plank 2000; Cysouw and Wohlgemuth 2010), often present such instances. If so, they challenge the theory, and warrant closer examination. Here I present an instance in which Corbett's theory prompted the re-examination of a rarum, namely the inflectional marking of complementised clauses in Kayardild. This led to a revised understanding of the phenomenon, and to support for the themy. Moreover, the re-examination presents the opportunity to compare three analyses of the same facts . The comparison highlights the methodological value for typology of attending to diachrony when evaluating rara. The argument has been made before that rara often arise through normal processes of historical change, only that because they require unlikely combinations of multiple changes, or rare preconditions, they are less likely to arise than common traits (Harris 2008, 201 O; Grossman 2016). Here I emphasise two corollaries of the nature and scientific value of rara. Because rara are particularly useful for hypothesis testing and attendant theoretical progress, and since our understanding of rara often requires reference to diachrony, it follows that methodologically, typology should ideally include as much about diachrony as it can in its evidentiary base; and epistemologically, it would be mistaken to characterise typology as an ahistorical discipline (e.g. Daniel 2011), since diachrony is paiticularly important for understanding some of the best evidence for testing typological theories. 'Tomorrow 1 will show you the man who gave me this boomerang. ' ft (3) Kayardild [E493.ex.12-13b] Jinaa bijarrb, [ngumbaa kurulutharranth] ? cina-a picarpa IJUIJ+pa-a ku-1.uluL+1Jara-inLa where-r dugong 2sG-XOBL-T kill-XPST-XDAT where dugong 2sG-COMP 1 kill-PST-COMP 1 'Where is the dugong which you killed?' (4) Kayardild [E493.ex.12-13a] Jinaa bijarrb, [nyingka kurulutharray]? cina-a picarpa JlÜJ+ka ku-tulu~+!Jara+ki-a where-r dugong 2sG-T kill-XPST-XERGLOC-T where dugong 2sG[ COMP2] kill-PST-COMP2 'Where is the dugong which you killed?'