Thread by @wmatchin

For 10 years, I have focused on the neurobiology of syntax. Finally, I feel that we have some solid answers in our new paper (with @GregoryHickok) just accepted for publication in Cerebral Cortex 1/13 docs.wixstatic.com/ugd/13f374_d0b…

Main problem with dominant Broca’s area view is aphasia. Broca’s aphasia patients, patients with damage to Broca’s area retain substantial syntactic knowledge, unlike patients with posterior temporal damage, who have the biggest sentence and syntactic comprehension deficits. 2/13

Localization of syntax to posterior temporal lobe faces obstacles, too. First, this region plays a key role in lexical access, which may appear to conflict with a syntactic role. Conveniently, the linguistics literature is in strong consensus: syntax is highly lexicalized. 3/13

Syntactic objects are actually lexical items in that they are elements of syntactic computation. For these reasons, we actually expect overlap in localization between syntax and the lexicon. 4/13

Also, agrammatic production is clearly caused by damage to frontal structures. Syntactic effects in neuroimaging studies are found in both inferior frontal and posterior temporal areas. So there really are two “syntax areas”. How do we make sense of this? 5/13

By decomposing syntax. Perhaps the most basic syntactic distinction is linear order (precedence) vs. hierarchical structure (dominance). These two dimensions relate to the demands of production and semantic interpretation, respectively. 6/13

Speech production relies on dorsal stream areas, semantic processing relies on ventral stream areas. Linear morpho-syntactic relations are driven by demands of production, and hierarchical syntactic structure is driven by demands of semantic interpretation. 7/13

By this logic we expect linear morpho-syntax to be processed in Broca’s area and hierarchical (and non-linear) syntax to be processed in the posterior temporal lobe. 8/13

Evidence comes from a little-studied syndrome in aphasia: paragrammatism. Subject-verb agreement errors in paragrammatis, e.g. “the birds was a color” (Yagata et al., 2017) follow from hierarchical deficit, as agreement must be computed over structure, not order. 9/13

We are now doing lesion-symptom mapping of agrammatism and paragrammatism. We show a clear double dissociation: agrammatism results from inferior and middle frontal damage, paragrammatism results from posterior temporal-parietal damage (come to SNL 2019 to see). 10/13

Our model fits neatly into dual stream architecture of speech. Perceptual targets (hierarchical structures) necessary for comprehension linked with higher-order gestures (morphological sequences) necessary for production, forming working memory loop for revision/prediction. 11/13

This paves the way for studying the evolution of syntax, which appears to rely on the same type of neural architecture as phonological processing. 12/13

This work has been a long time coming. We are looking forward to empirically testing this model in future work. Any feedback is most welcome! 13/13

Also, special thanks to @smwilsonau for incredibly useful feedback and overall encouragement.

The paper is now up on @PsyArXiv psyarxiv.com/6394f

With @StarkBrielle, @aphasia_lab, Alexandra Basilakos, and Dirk den Ouden.