Linguistics Seminar Talk - John Collins and Tamara Dobler

Title: Polysemy and Roots: Wide vs. Narrow Fetching

Abstract: 

Polysemy and the copredication it allows for have been claimed to pose a severe challenge for traditional truth-conditional semantics (Chomsky 2000; Pietroski 2003, 2005, 2018; Collins, 2017, 2023). Myriad responses have been offered to this challenge. Our intent is not directly to engage in this debate over the fate of the truth-conditional viewpoint, but to investigate a related thread: if polysemy/copredication is a severe challenge to traditional semantic thinking, do the purveyors of the challenge offer a plausible way of accommodating polysemy?  Our focus will be on Pietroski’s (2018) model whereby lexical items are not assigned semantic values (worldly entities) but encode instructions to fetch and build concepts from memory addresses that potentially contain a number of concepts. We shall view this model as a blueprint for how to understand polysemy, and shall fill it in with a narrow individuation of fetch, which best captures Pietroski’s intention. By these lights, the ‘meaning’ of a lexical item is an instruction to take some concept or other from an address, without any concept being necessarily delivered; that is, fetch targets an address rather than whatever is at the address. Thus, the address itself becomes akin to a root concept on the model of a lexical root. A wide individuation of fetch, which we suspect is how Pietroski’s model is naturally viewed from a traditionalist optic, amounts to fetch always delivering a concept for a lexical item viewed in isolation of any syntactic structure. Wide fetching, as it were, does occur, in the sense that concepts at addresses are delivered, but which concept is fetched is a function of stages of syntactic construction in which the given lexical item sits. In this sense, the meaning of a lexical item cannot be an instance of wide fetching. Consideration of polysemy will bear out this conclusion. So, we think the polysemy/copredication challenge can be met by a non-truth-conditional model.