Noam Chomsky believed in a thing called "universal grammar," an underlying set of rules which all world languages' grammars follow. Furthermore, he believed that we have mental hardware specifically equipped for language, and more specifically that our brains are pre-programmed to understand certain grammatical features which all languages are observed to follow. He would claim that the reason that grammars are the way they are is because that space of possible grammars is defined neurologically.
Another prevailing opinion is that we just use general statistical pattern recognition circuitry to pick up language, among other things.
Both claims are coherent with the modern theory of language acquisition, it's just a question of what the specific acquisition circuitry is doing- is it language specific, or is it just general circuitry being used for that purpose? Chomsky believed the former. That said, I don't think either are quite right.
Notice that language is patterned in a very similar way that our perception interprets our environment. We do have specific neurology to facilitate such perception, which is why Chomsky's universal grammar seems true at first glance. But, on the other hand, acquisition is just statistical pattern recognition built on top of our system of perception. That is to say, grammar is confined as it is because that is the way an agent in an environment sees things, not because the acquisitional neurology is particularly restrictive.
We subconsciously segment our visual field into individual objects with purposes and associated behaviors. This inevitably yields describing things with a more or less closed set of words and not just a blurred swirled artistic gradation of meaning. Similarly, the baseline grammar of having subjects, objects, and verbs might well be a result of some fundamental way of seeing the world as actions caused by things which happen to other things. This mode of perception is easily taken for granted, but nothing is stopping you from being a hippie and parsing the world as just some asegmented blur with no distinct boundaries between things.
So, I think parts of both are correct insofar as they both describe an underlying structure: biology defines perception, but the method of statistical acquisition which lies above perception is fundamentally general beyond language. If you take the claim of Universal Grammar and replace "language" with "perceptive structure" then it becomes true: "we have mental hardware specifically equipped for perception, which naturally prescribes some patterns on language." That's why his theory seems true. Language is deeply connected to the way that we interpret things, so it's easy to not be able to see where the line between language and interpretation lies.