Skip to main content

Liina Pylkkänen

Tuesday 22 January at 4pm, Colin Bannard will give a colloquium in the UCSD Linguistics Department (co-sponsored by CRL), in AP&M 4301.

Building Meanings: The Computations of the Composing Brain


Although the combinatory potential of language is in many ways its defining characteristic, our understanding of the neurobiology of composition is still grossly generic: research on the brain bases of syntax and semantics implicates a general network of “sentence processing regions” but the computational details of this system have not been uncovered. For language production, not even a general network has yet been delineated. Consequently, the following two questions are among the most pressing for current cognitive neuroscience research on language:

(i) What is the division of labor among the various brain regions that respond to the presence of complex syntax and semantics in comprehension? What are the computational details of this network?

(ii) How does the brain accomplish the construction of complex structure and meaning in production? How do these processes relate to parallel computations in comprehension? 
In our research using magnetoencephalography (MEG), we have systematically varied the properties of composition to investigate the computational roles and spatiotemporal dynamics of the various brain regions participating in the construction of complex meaning. The combinatory network as implicated by our research comprises at least of an early (~200-300ms), computationally specialized contribution of the left anterior temporal lobe (LATL) followed by later and more general functions in the ventromedial prefrontal cortex (vmPFC) and the angular gyrus (AG). The same regions appear to operate during production but in reverse order. In sum, contrary to hypotheses that treat natural language composition as monolithic and localized to a single region, the picture emerging from our work suggests that composition is achieved by a network of regions which vary in their computational specificity and domain generality.