This paper addresses the scope behavior of bare nouns (BNs) in Kaingang (Jê, Brazil) and its relevance to the study of semantic variation. I show that Kaingang BNs exhibit variable scope in relation to other clause-mate operators, such as negation, intensional verbs, if-clauses, universal quantifiers, and frequency adverbials. Adopting a type-shifting framework (Chierchia in Topics in the syntax and semantics of infinitives and gerunds. PhD diss., University of Massachusetts, 1984; Events and grammar, pp. 53–103, 1998a; Natural Language Semantics 6: 339–405, 1998b; Partee in Studies in discourse representation theory and the theory of generalized quantifiers, pp. 115–144, 1986; Krifka in Proceedings of SALT 13, pp. 180–203, 2003; Dayal in Linguistics and Philosophy, 27: 393–450, 2004, among others), I analyze Kaingang indefinite BNs as predicate NPs mapped into arguments via a choice function (CF) that gets existentially closed at any level in the clause, à la English indefinites under Reinhart’s (Linguistics and Philosophy, 20(4): 335–397, 1997; Interface strategies: Optimal and costly computations, 2006) analysis. I also argue that the difference in scope properties across languages between variable-scope and narrowest-scope BNs is linked to the quantized/cumulative opposition of the predicates they express: the former are quantized predicates, while the latter are cumulative. I then claim that variable-scope and narrowest-scope BNs are created by two distinct type-shifters: the former by CF and the later by Krifka’s ∃ operator (Krifka in Proceedings of SALT 13, pp. 180–203, 2003). In this way, I propose that both operators are sensitive to the quantized/cumulative opposition and model this as conditioned by a binary semantic feature, namely [±quantizedness], which checks the (non-)quantized properties of the predicates that undergo type-shifting. The CF type-shifter is [+quant], and thus defined only for quantized predicates, while Krifka’s ∃ type-shifter is [−quant], selecting only non-quantized ones. This analysis is shown to correctly predict the variation in scope properties of BNs across languages.