It seems like there’s potentially a kind of subtle circularity in how the concept of countable (i.e. countably infinite) first-order languages is defined. On the one hand, one putative goal of first-order logic is to rigorously axiomatize the natural numbers. But on the other hand it seems like you kind of need the concept of natural number to already exist in order to define what a countable language is: arguably, recursive grammar definitions can mostly avoid direct mention of “number concepts”; but once you need to talk about arbitrary “arities” of relations and functions (like $n$-ary relation for “any $n$”) then “number concept” seems unavoidable.
Do logicians have an accepted way of resolving this?
- I could hypothesize this: Imagine you’ve started with a finite “mini-language” (finitely many symbols and sentences), used the “mini-language” to define natural numbers, and then leveraged the now-defined natural number concept for the full-blown definition of the countable first-order language concept.