Morphosyntactic convergence in Kadu agreement

Kadu is a small family of languages spoken in the Nuba Mountains of Sudan. Table 1 compares cognate noun forms and their associated verbal subject agreement prefixes in two languages of this family, Krongo and Katcha. Krongo has the agreement prefixes n-, m- and Ø-, which may be used for either singular or plural agreement, and k-, which is used only for plural agreement. Katcha has the same agreement prefixes as Krongo, except that instead of opposing n- and k-, Katcha uses only k-.1 (References for forms in the tables are given in the Appendix.)At first glance, this juxtaposition might seem unremarkable: splits and mergers of inflectional categories are hardly a rarity. In particular, the number of agreement classes or genders in a language may grow or shrink over time (DiGarbo & Miestamo 2019). But in trying to reconstruct how this came about, it becomes clear that it could not have followed any of the known pathways of phonological, morphological or semantic change. I argue that it was of a type of change that has not previously been described. Specifically, that Katcha (and other Kadu languages) underwent a merger of classes triggered by a morphosyntactic generalization. In the original system, noun forms marked with productive singular affixes were, unsurprisingly, assigned to a different agreement class from those marked with productive plural affixes. The innovation was to combine these into a single class, so that all productive number affixes, whether singular or plural, take the same agreement.The paper is structured as follows. §2 provides background on the language family, on noun morphology, and on the agreement system. §3 presents the evidence that there were originally two distinct agreement classes, still preserved in Krongo but conflated in at least some of the other Kadu languages. §4 proposes an explanation for this diachronic change, arguing that it was triggered by joining all productively inflected singular and plural noun forms into a single agreement class. §5 concludes, arguing that a such a diachronic morphosyntactic convergence, motivated by sheer morphological markedness, has not been observed before. Superficially similar systems of ‘inverse number’ are shown to involve a different set of principles