Dogon adjective-numeral inversion

Abstract In several Dogon languages, the ordering of simple N-Adj-Num DPs is strict. Adjectives and numerals also differ from each other in morphology and tonosyntax. When an inversion licensor (usually a reference restrictor) is addèd, N-Adj-Num is optionally inverted to N-Num-Adj. Pronominal possessors are also subject to reordering in some of the languages. The order of two adjectives in N-Adj1-Adj2 is free, so a syntactically conditioned stem-class merger between numerals, adjectives, and pronominal possessors would account for optional inverted orderings. However, morphological and tonosyntactic processes follow inversion and are sensitive to stem-class categories, showing that inversion cannot be postsyntactic. This suggests that the relevant word-classes belong to a supercategory medial modifier that is activated in linearization by an inversion licensor, but is disregarded by morphology and tonosyntax. Typological comparisons, e.g. to Slavic numerals, are superficial, so the approach here is micro-typological, focusing on variations within a broadly shared family-specific system.