In his (1921) book, Language, Sapir made the famous observation, "All grammars leak" (38). By this he meant that within the systematic paradigms, rules and routinized patterns of any grammar, we always find a few irregularities and surprises. The same can be said for linguistic typologies. Typological theories are critical tools for linguists, for exploring differences and similarities among languages, for learning about the cognitive factors and social practices that make languages the way they are, and for making predictions about other properties of languages that are members of a certain type. So what do we do when a typology leaks? This paper follows the spirit of such work as Aske (1989) on path types and Mithun and Chafe (1999) on grammatical relations types to understand the grammatical and functional motivations of language-internal typological diversity: that is, why and how a single language uses patterns and constructions of more than one type.