Background: Much of philosophy since Kant has taken for granted that our basic experience of reality is structured by our cognitive apparatus, including notably our background conceptual frameworks. However, realists also want to say that when we succeed in knowing things through our mental apparatus, what we know is real in spite of its essential dependence on structures of cognition. There is for me an unresolved tension here.
In addition there is the problem of whether concepts themselves are discovered or engineered. If they are discovered does that commit us to extra Platonic entities? If they are engineered, what are the success criteria for a well engineered concept? If we take the pragmatist path of saying concepts are useful or the realist idea that concepts should "carve the world at its joints" we are back to trying to account for some sort of reality that either grounds usefulness or joint-carving, and this reality presumably also needs to be expressed via concepts, so I don't see how either pragmatism or joint-carving accounts can solve the problem of what makes concepts successful qua concepts.
Where I'm at: Recently I've been seriously considering the view that the thing we need to place at the foundation of both metaphysics and epistemology is pattern/meaning. Not meaning as in the meaning of a word; linguistic meaning is derived from the primary meaningfulness of that which the language is about. The view I'm considering is that we do not impose pattern/meaning on perception through our cognitive apparatus, but discover it directly in the world. When we perceive such a pattern/meaning, we might call the combination of the pattern and our perception of it a gestalt. Concepts are formed by noticing patterns in our gestalts (i.e., forming gestalts of gestalts). Once we've formed concepts they may operate to direct our attention selectively towards those patterns that fall under them. This creates the illusion that our concepts gives us the pattern, when in fact, we are only selectively noticing patterns which exist independently of the concepts that point us towards them.
We can consider Wittgenstein's famous example of the duck-rabbit. Suppose here there are two patterns, which correspond to two gestalts, the duck and the rabbit. If we approach the image with the conceptual lens of rabbit (suppose we've never seen a duck) then we'll only see the rabbit gestalt, and vice versa. But this doesn't mean the rabbit gestalt is constructed by the rabbit concept. In fact, the dependency flows in the other direction. We could only form a rabbit concept but first appreciating many rabbit gestalts.
There are two accounts in the academic literature I've come across that seem to be saying something similar, but both are outside the mainstream of analytic philosophy. The first is Merleau-Ponty's Phenomenology of Perception, where he argues for the primary reality of a realm of a cross-modal "significance," which I think can be reasonably interpreted along the lines I've set out above. The second is the mature view of the cognitive scientists Douglas Hofstadter as set out in his book Surfaces and Essences. Here, he accounts for cognition in terms of analogies, but in my reading, this view depends essentially on us having direct pre-conceptual perception of similarities/patterns in the world.
My question is whether any contemporary analytic philosophers defend (or attack) this sort of view, and if so what it is called, and what articles/books I should look to?