In this paper, we argue for a unified analysis of Positive Polarity Items (PPIs) formed with some, such as something, someone, some NP, and nominal minimizers, such as a damn thing, a red cent, a drop. Our analysis capitalizes on Strawson’s (1974) observation that the use of some NP involves inferences about the availability of a more precise identification of NP. We show that opposite polarity properties of some-PPIs and nominal minimizers are derivable from pragmatic requirements on the granularity of the context structure where they can be used.