Ad hoc concepts and the relevance heuristics

The idea that interpreting a lexeme typically involves a context-dependent process of meaning construction has in recent years become common ground in linguistic theory. This view is very explicit in relevance theory (Sperber and Wilson 1995), which posits that speakers systematically infer ad hoc concepts (Carston 2002). Such an approach raises theoretical issues, though. First, it directly poses a challenge for the exact nature of (and difference between) concepts and ad hoc concepts (Carston 2002, 249). In addition, as Wilson (2011, 2016) and Carston (2013, 2016) point out, this view also uncovers the following paradox: if speakers are assumed to follow a path of least effort (relevance heuristics), why should they so systematically infer ad hoc concepts rather than test the encoded concept first? The aim of this paper is to reflect on this theoretical puzzle. It will first be argued that the hypotheses formulated both by Wilson and by Carston seem rather post hoc and fail to fully resolve the apparent paradox. Attention will then be given to the assumed nature of (ad hoc) concepts to show that the problem can be resolved when an alternative (non-atomic) view of concepts in terms of meaning potential is adopted.