Rara and theory testing in typology: The natural evolution of non‐canonical agreement

In typology, rara provide valuable tests for theoretical hypotheses. Here I consider the rarum of PERSON inflection in Kayardild, which has only two surface contrasts but is found across all words in complementized subordinate clauses. I introduce a general schema for reasoning about the diachronic emergence of rara, and reconstruct the evolution of Kayardild subordinate PERSON agreement, from an earlier state in which a main‐clause inverse system was coupled to a system of complementizing CASE agreement. Serendipitously, the same synchronic facts have been analysed twice earlier without the benefit of the full diachronic backstory, and so present a retrospective case study in what diachrony offers for the analysis of rara, structures which by definition are difficult to contextualize using synchronic typology alone. I argue that since rara are so valuable for the testing of typological theories, and since diachrony may offer the only source of convincing explanation for them, it follows that typological science will need to refer to diachrony for the successful development of theory. It cannot rely on synchrony alone .