My problem was and remains the quasi-exclusion of syntax in the theory. I do believe (and that’s in fact the very only belief I hold) in deep-structure and in generative grammar.
This is not to say that associative learning is wrong but that it is secondary in the core of learning. Even though associations can be extremely complex, as when the brain compares firing patterns for example, I think it is reductive, and therefore dangerous, to eschew grammar and non causal meaning from the picture.
I think this question is also relevant to programming, and I would even imagine system integrity, which is why I upvote this post.