8/ intLANG was preferentially recruited by the Nonword > Fixation contrast and aLANG by Sentence > Nonword — a robust functional double dissociation carried by a full crossover interaction.

7/ Third, the two networks are functionally dissociable. intLANG responds robustly during rhyme judgments and nonword reading that emphasize phonology, whereas aLANG is preferentially recruited during meaning-based sentence processing.

6/ Second, the segregation extends beyond the cerebral cortex. Both networks can be fully recapitulated by functional connectivity from adjacent cerebellar regions, indicating they are segregated, brain-wide networks.

5/ Three lines of evidence establish the two networks as separate. First, intrinsic connectivity among the component regions displays block-modular features: strong connectivity within each network, weaker connectivity between networks.

4/ The two networks are anatomically adjacent with juxtaposed component regions in both anterior and posterior cortex, making their separation challenging. This may be why intLANG has been difficult to resolve as a distinct distributed network in past studies.

2/ intLANG, an intermediate language network, is anchored to the precentral speech areas and the Sylvian parietal-temporal area.

I am thrilled to share that I’ll be joining the University of Notre Dame (@notredame.bsky.social) as an Assistant Professor of Psychology this July!☘️🧠 Please reach out if you're interested in joining my lab! More details to follow soon.

To demonstrate this, we obtained both networks and network-level task response in a new participant during revision of this work, using only NBACK task data from a single ∼1 h session. There was a strong preferential response in FPN-A as compared with other networks. (10/13)

We further showed that between-individual differences in task responses can be obtained from network estimates derived from only task data, without acquiring separate resting-state data. (8/13)

By pooling extensive resting-state and task data, we were able to triple the amount of data available for analysis within each individual, enabling precise mapping of five higher-order association networks within the thalamus. (6/13)

We then quantitatively demonstrated that pooling resting-state data with motor task data stabilizes the similarity of correlation matrices between test and retest datasets. This suggests that we can pool all resting-state and task data to increase statistical power. (5/13)

Furthermore, networks estimated solely from task data predicted functional specializations across multiple higher-order cognitive domains in independent task datasets just as well as traditional resting-state network estimates did. (4/13)

Direct comparisons of network estimates from both datasets reveal a convergent functional architecture of the brain. While the fine-grained spatial details of these networks varied across individuals, they were largely preserved within each individual. (3/13)

Using only task data, we derived a 15-network multi-session hierarchical Bayesian model (MS-HBM) estimate, and the results were remarkably similar to those derived from traditional resting-state data. (2/13)

Our new paper is out now in Neuron! 🎉 With @vaibhavtripathi.bsky.social @maxwellelliott.bsky.social Joanna Ladopoulou, Wendy Sun, Mark Eldaief, and Randy Buckner Paper link: www.sciencedirect.com/science/arti...

Additionally, network estimates from task-regressed data predict functional response properties in independent contrasts similar to parallel analyses using traditional resting-state fixation data. (5/6)

We demonstrated that networks can be estimated robustly within individuals using solely task-regressed data. The idiosyncratic spatial details varied between individuals but were largely preserved within each individual across datasets under independent acquisition conditions. (4/6)

Our findings indicate that functional correlation matrices derived from task data are highly similar to those derived from traditional resting-state acquisitions. The largest factor affecting similarity between correlation matrices was the amount of data. (3/6)