Syntax refers to the rules governing how words combine into sentences. Humans acquire and use complex grammatical structures allowing finite means to create infinite expressions. Language comprehension and production require parsing and generating sentences according to grammatical principles, suggesting dedicated cognitive mechanisms for syntactic processing.
From your study of language networks, you know that language processing is distributed across a left-lateralized network including Broca's area (inferior frontal gyrus) and Wernicke's area (posterior temporal gyrus), connected via the arcuate fasciculus. These regions do not perform equivalent jobs: Wernicke's area is central to accessing word meaning (semantics), while Broca's area is particularly implicated in processing the hierarchical structure of sentences — syntax. Understanding syntax means understanding how words combine into phrases and sentences in rule-governed ways, and why you can effortlessly generate and comprehend sentences you have never heard before.
The core theoretical claim about syntax is productivity: a finite set of grammatical rules can generate an unlimited number of sentences. You can take any sentence and embed it inside another ("She knows that he believes that they think that..."), and the result, while cumbersome, remains grammatical. This property, called recursion, is present in all known human languages and is absent from natural animal communication systems. Noam Chomsky argued that this productivity implies syntax is not learned by imitation or statistical pattern extraction from the input, but rests on an innate language faculty — Universal Grammar — that specifies the abstract principles all human grammars share and the parameters along which they vary. The child does not learn syntax so much as set parameters within a pre-structured grammatical space.
The opposing view holds that syntax can be acquired from the statistical regularities in the linguistic input, without positing innate grammatical knowledge. The brain is extraordinarily good at detecting distributional patterns — which words co-occur, in which orders, with which frequency — and infants show sensitivity to these patterns before their first birthday. Connectionist models demonstrate that neural network architectures trained on naturalistic language can acquire generalizations that look grammatical without being explicitly programmed with grammatical rules. The debate between these positions (nativism vs. statistical learning) remains unresolved, though most contemporary accounts are interactionist: some domain-relevant predispositions constrain what statistical patterns are attended to and how they are generalized.
During comprehension, the language system constructs a hierarchical parse tree — a representation of how words group into phrases and how phrases relate to each other — incrementally as each word arrives. This parsing is remarkably fast and largely automatic: the brain begins constructing syntactic structure within ~100-150 milliseconds of each word onset, before the word's meaning is fully accessed. Syntactic violations produce a distinctive EEG response (the P600 component) around 600 milliseconds after the violation, suggesting that syntactic analysis is ongoing and that violations are flagged and repaired. Understanding this sequence — from the word-level access you learned about in language networks to the phrase-level and sentence-level combinatorics covered here — gives you the foundation for studying sentence comprehension and speech production planning in more advanced courses.