Linguistically Expressible Concepts: a simple mentalese for human grammars

Here’s an old idea that I like: humans use pronounceable expressions to express concepts that exhibit logical relations. But if that’s right, what’s the format of the concepts expressed?
In the first part of the talk, I describe a possible Language of Thought (LoT) that is very simple, yet captures some important inferential patterns. From a formal perspective, this LoT is an elementary predicate calculus. Every predicate/concept is monadic; think of COW, BROWN, BROWN^COW, BLACK, HORSE, BLACK^HORSE, BROWN^HORSE, BLACK^COW, and so on. This LoT provides no form of concept-negation: you can’t use COW to build a concept that applies to whatever COW doesn’t apply to. Nonetheless, there is a way to reconstruct propositional logic. I’ll then show how to supplement this LoT in ways that increase expressive power dramatically, but leave the generative system unaffected. The net result is enough to reconstruct Aristotelian logic, while also capturing the inferential patterns that motivate Davidsonian event analyses— e.g., ‘A baker buttered a bun with a knife in a kitchen at dawn, so a baker buttered a bun’. I will be drawing on some ideas from my book Conjoining Meanings, a recent paper by Thomas Icard and Larry Moss (“A Simple Logic of Concepts”), and some work in progress with Icard.