As far as “inate structures” of language two areas seem to be key:
Broca’s and Wernicke’s Areas process objects and grammar in separate areas.
Broca’s area seems to be specialized for motor production, and the learned patterns correspond somewhat to parsing and producing grammar.
Wernicke’s area seems to be specialized for high-level object representation. I see this as being of somewhat the same complexity as the Cortical IO retina construct. More details here.
These two areas are connected by a major association fiber tract
Considering the relation between grammar and word store should give some flavor of understanding of the kind of information that can be carried by an association fiber, in this case, the Arcuate fasciculus.
By the way - the first link mentions "Mirror neurons.” The relation between maps and object store is relevant here: Maybe it’s time to ditch Mirror Neurons? The referenced Friedemann Pulvermüller paper show that the idea of a central semantic object store is hopelessly idealized, and that in fact, the representation is grounded and distributed over all of the sensory processing streams. Referring to a concept from SQL database lore, Wernicke’s area is the key or index to the distributed semantic concept representation.