Linguistic meaning requires stable concepts

Shea’s (2024) identifies concepts with temporary working memory (WM) labels: “two-ended” devices connected to stored information at one end and composing with other labels at the other. I argue that this framework cannot yet account for linguistic understanding. First, Shea’s theory requires stable categorical mechanisms to track category identity over time, although he is officially neutral as to their nature. Second, even granting such mechanisms, neither they nor his temporary WM labels can provide the needed compositional vehicles. The empirical evidence I review converges on the need for stable, freely compositional vehicles that I call “G-concepts.” On the view I develop, G-concepts – stable symbols individuated by their compositional role – compose into thoughts and bear word meaning, while “S-concepts” – stable representations that underwrite categorization – orient the mind toward the world. The Hybrid view preserves Shea’s interface insight while providing the compositional resources needed for linguistic cognition.