Research
1. Organization of the human language system
How many distinct networks support the human language system, how are they arranged, and what does each one contribute? In our recent study, we used precision functional mapping across three independent cohorts (22 individuals, 216 MRI sessions) to dissociate two nested left-lateralized perisylvian networks: an intermediate language network anchored to precentral speech areas and the Sylvian parietal-temporal area, and a surrounding association language network extending into higher-order prefrontal and temporal cortex. The two are functionally distinct, the first recruited by phonology and the second by meaning-based sentence processing. Both are fully recapitulated by functional connectivity from adjacent cerebellar regions, which indicates they are segregated brain-wide networks.The next step is to ask whether the same nested arrangement appears in subcortical structures including thalamus and striatum, and what that would say about how the language system is assembled.
2. Brain-wide association circuits
I discovered a recurring spatial motif in the human association cortex that I call a supra-areal association megacluster (SAAM). Each SAAM contains two adjacent domain-flexible control networks (FPN-A/B) that are consistently positioned next to three domain-specialized networks (LANG, DN-B, and DN-A). Functionally, the control networks flexibly respond to working memory load, while the other three networks dissociate across language, social inference, and episodic processing domains. These SAAMs are reproducible across individuals and extend beyond the cerebral cortex: the same five networks also form SAAMs in the cerebellum, basal ganglia, and thalamus, indicating a brain-wide architectural principle. We now ask whether the functional dissociations found in the cererbal cortex hold in those subcortical structures as well. Characterizing the organization of subcortical structures poses specific challenges due to their small size and idiosyncratic anatomical variability among individuals. To address these gaps, we conduct intensive sampling of individual participants across repeated sessions, using scan sequences optimized for deep subcortical regions and a wide range of task manipulations. Our goal is to produce next-generation precision maps of brain-wide circuits and investigate their functional selectivity.
- Organization of the human cerebral cortex estimated within individuals: networks, global topography, and function
- Within-individual precision mapping of brain networks exclusively using task data
- Within-individual organization of the human cognitive cerebellum
- Ventral striatum is preferentially correlated with the salience network
- Human striatal association megaclusters
3. Anatomical grounding of network estimates
The circuits supporting human higher-order cognition are old. Distributed association zones linked by long-range reciprocal projections appear in the Old World monkeys and also in the common marmoset, a New World monkey that diverged from our lineage roughly 45 million years ago. Comparing precision network estimates in humans against tracer injections in the marmoset, we found the same ordered sequence in both, from a local motor network to an apex transmodal network in frontopolar and rostral temporal cortex. All six distributed regions of the human default network can be predicted from marmoset projections. What is conserved is not only which zones are connected, but the arrangement in which the networks sit. Association cortex is disproportionately expanded in our species, so a motif compressed into the marmoset frontal lobe spreads across large territories in ours, plausibly what let one ancestral circuit fractionate into multiple specialized ones. And because connections can be traced in the nonhuman primates and not in humans, every point of agreement doubles as a check that our indirect measurements track neuroanatomy.
4. Precision targets for neuromodulation
Neuropsychiatric disorders are increasingly understood as disorders of large-scale distributed brain networks. Precision functional mapping preserves the idiosyncratic anatomy of each person, enabling reliable estimates of networks in the native volume of an individual's brain — the space in which stimulation is actually planned and delivered. We combine tools from individualized preprocessing, precision network mapping, and electric field (E-field) modeling to guide clinical neurostimulation decisions, including transcranial magnetic stimulation (TMS). We aim to characterize network-level effects of TMS and to make target selection in therapeutic neuromodulation more precise.
- Precision network modeling of TMS across individuals suggests therapeutic targets and potential for improvement
- Increased aperiodic exponents track depressive symptom severity in refractory epilepsy
5. Developmental origins of the network arrangement
Where does network arrangement in adults come from? The same five association networks sit side by side, in the same order, in zone after zone of association cortex (Du et al. 2024). Our hypothesis is that the arrangement is the residue of development: association zones that begin broadly connected and undifferentiated, then fractionate and specialize through activity-dependent competition during the early years of life. If this account is right, fractionation and specialization happen in infancy and toddlerhood - a window we are eager to get at but will not be scanning at Notre Dame for at least two years. For now, we can only reach it through collaboration. If you work with infants or toddlers and precision functional mapping interests you, please get in touch!