Copredication and Complexity Revisited: A Reply to Löhr and Michel

Human language affords the ability to attribute semantically distinct concepts to a single nominal, a process now commonly termed 'copredication'. If we describe a lunch as being delayed but also filling , senses of distinct semantic categories (event, physical object) are simultaneously being accessed. Copredication is relevant to major debates in cognitive science, since it cuts to the core of how the lexicon is formatted, and how distinct lexico-semantic representations relate to each other. The apparent scope and limits of copredication licensing can be explored via acceptability judgment and processing experiments, exposing certain replicable and generalizable patterns that apply across lexical types, syntactic structures, and different languages (Murphy 2021a, 2021b). As such, laying out the psycholinguistic terrain in which to address this phenomenon is crucial - and accounts that lack a valid psycholinguistic and empirical basis should be highlighted as problematic if they are to be accommodated and refined. Löhr and Michel (2022) claim that copredication acceptability is determined by a "set of expectations that are influenced by higher-order priors associated with discourse context and world knowledge". I will show that their model encounters a number of obstacles, and ends up unintentionally supporting an alternative model in Murphy (2019, 2021a, 2021b, 2021c), which they attempt to critique.