This paper deals with the restrictive (limitative) marker kə̄n ‘only’ in Ulcha (Southern Tungusic). This
marker has nontrivial positional features: it can attach before inflectional suffixes (as a derivational affix) or after them (as an
enclitic). One might see the process of affix reordering described in Haspelmath (1993) as
“externalization of inflection”, when a former clitic becomes a derivational affix. However, there is evidence that the uses of
kə̄n after inflection are innovative as compared to those before inflection, not vice versa, and this direction of
diachronic development is very unexpected. In this paper, I propose an explanation for this nonstandard reordering pattern and show that in
fact it has the same motivation and the same mechanisms as previously reported types of affix reordering.