Both individuals and predicates can be referred to in different ways which carry different senses or connotations. Despite this being discussed since at least Frege, it poses a deep problem to standard extensional semantics. For example, as discussed by Jennifer Saul, “Clark Kent went into the phone booth and Superman came out” simply means something different from “Superman went into the phone booth and Clark Kent came out”.
I introduce a novel way of modelling these kinds of semantic phenomena using Sampling Propensity (Icard, 2016). The core idea is that the basic atoms of semantic calculus are generated from a set of potential candidates via a generative cognitive procedure. In other words, when one thinks of Clark Kent, they directly think of someone wearing glasses or being a mild-mannered journalist, whereas Superman draws to mind blue-and-red leotards and heroics. This sampling procedure is also at play with common nouns in generic sentences like “lions have manes” or “mosquitoes carry malaria”. Crucially, it can distinguish co-extensional nouns like “drink” and “beverage”, which occasionally yield different truth values in the same kinds of generic sentences.
The account includes a fully formalised compositional system in which individual concepts and category concepts are modeled as an extension linked with a sampling propensity and where some propositions are evaluated by continually sampling exemplars from a concept. The sampling approach also links competence and performance where finite sampling yields performance and sampling repeatedly converges on competence. This approach also has ramifications for quantification, particularly the generic "flavour" of non-partitive "all".