Countability distinctions and semantic variation

To what extent are countability distinctions subject to systematic semantic variation? Could there be a language with no countability distinctions—in particular, one where all nouns are count? I argue that the answer is no: even in a language where all NPs have the core morphosyntactic properties of English count NPs, such as combining with numerals directly and showing singular/plural contrasts, countability distinctions still emerge on close inspection. I divide these distinctions into those related to sums (cumulativity) and those related to parts (divisiveness, atomicity, and related notions). In the Sahaptian language Nez Perce, evidence can be found for both types of distinction, in spite of the absence of anything like a traditional mass–count division in noun morphosyntax. I propose an extension of the Nez Perce analysis to Yudja (Tupí), analyzed by Lima (The grammar of individuation and counting, 2014) as lacking any countability distinctions. More generally, I suggest that at least one countability distinction may be universal and that languages without any countability distinctions may be unlearnable.