That's a moot point, I think, as it can be argued easily that thought is a language itself, or requires one (whether it's a human language like english, or a language of symbols, or some other form of structured description of events and thoughts, even the structure just happens at the chemical level in the brain).
>Specific expressions of models are all equally good by the "correspondence" measure as well: anyone with even a grade-school math education should be able to recognize A=BCD and I=JK,K=QP as distinct in the literal sense but identical in some higher one. So, once again, it's obvious that we are looking at different ways of expressing the same thing; a thing which hasn't yet been proven not to exist, even though no way of expressing it can by itself be more than just "a different way."
It's obvious in these examples -- which, coincidentally are all from math (trigonometry, coordinate systems), that is the quintessential non-empirical/reality-based domain. It's easy to find isomorphisms like that in math, since that's inherent in their core purpose.
It's not obvious (or true) for the general case, talking about the outside world.
I do think that there's a mind-independent world that gives us basic referents to represent. But the world is not cut at the joints (it doesn't come self-divided into identifiable objects) and is not self-interpreting (it does not represent itself to us). Thales, the first philosopher, certainly had a concept of water - he thought that everything in the universe was made of water. But despite the fact that the referent is the same, his concept of water is markedly different from our own. There is not an unmediated correspondence between the concept and the referent; hence the historical variability and contingency.
This is debatable[0]. Presumably, people without language can still think, and even if you believe that they invent an internal language to think with you are placing language into a tool-role, where it can be created and destroyed in service of a higher goal (thought). However, that still leaves the problems with correspondence unexplained: but hopefully I can convince you that they are not too bad.
I'll sketch it for you:
Coordinate systems are all equally good by the measure of how well their predictions work. They're obviously constructions - but that doesn't itself imply that the models that use them are constructions, any more than the wrappers burgers come in makes their meat paper. (I mean, the models might be constructions too, but it's not the use of arbitrary coordinate systems that makes them so.)
Specific expressions of models are all equally good by the "correspondence" measure as well: anyone with even a grade-school math education should be able to recognize A=BCD and I=JK,K=QP as distinct in the literal sense but identical in some higher one. So, once again, it's obvious that we are looking at different ways of expressing the same thing; a thing which hasn't yet been proven not to exist, even though no way of expressing it can by itself be more than just "a different way."
So, one step above algebra and two steps above specific calculations, there's the concepts. From a scientific perspective unless you think the brain is supernatural it's really to be expected that ideas are no more than symbols. BUT: for the same reasons as the two cases above, it has yet to be shown that the concepts are not themselves dancing around a still higher truth, each (effectively true) idea differing only in the baggage of being human; that silly but fun phrase referring to how we like symbols but are presumably using them to mean something.
[0]https://plato.stanford.edu/entries/language-thought/#ObjLOT