21. Brain Networks

What this lecture covers

The lecture shifts attention from individual cortical regions, the subject of most of the course, to the white matter connections between them and to sets of regions that act together as networks. It asks how connectivity can be measured in the living human brain and what resting-state correlations reveal about which regions function as a system. After watching, you can explain the main methods for studying brain connectivity, their limitations, and how researchers used them to show that language regions, multiple-demand regions, and theory-of-mind regions each form separate, internally correlated networks.

The lecture opens with a recap of the previous lecture on the temporoparietal junction (TPJ) and theory of mind, then argues that no single brain region works in isolation, so understanding what a region is connected to is essential to understanding what it does.

Key ideas

• White matter: myelinated axon bundles connecting distant brain regions, making up about 45 percent of the human brain, a higher proportion than in other studied species.
• Connectivity fingerprint: the distinctive pattern of a region's long-range connections, useful for identifying homologous regions across species and predicting a region's function.
• Diffusion imaging: an MRI method that measures the direction water diffuses in tissue, used to identify major fiber bundles like the arcuate fasciculus, based on the assumption that water diffuses more freely along axon bundles.
• Fractional anisotropy (FA): a measure of how strongly oriented water diffusion is in a given patch of tissue, commonly used to compare fiber tract quality between groups, though its biological meaning is not fully settled.
• Tractography: an algorithm that follows local diffusion orientations voxel by voxel to infer connections between gray matter regions; the lecture reports that in the lecturer's own testing it fails basic reality checks, such as correctly tracing the visual pathway from thalamus to cortex.
• Resting functional correlations: correlations between the time courses of brain activity in different regions while a subject is not performing any task, first reported by Biswal in 1995, used to identify sets of regions that act as a network.
• Default mode and multiple-demand networks: two systematic sets of regions found through resting correlations, one associated with unstructured "resting" cognition such as recalling memories or thinking about people, the other engaged across many different demanding cognitive tasks.

Walkthrough

Recap: theory of mind and the TPJ (32)

The lecture reviews the false-belief task, the selectivity of the temporoparietal junction for reasoning about others' thoughts (rather than their physical sensations or external properties), and evidence from autism and TMS studies that the TPJ carries information relevant to moral judgments about intentional versus accidental harm.

Why connectivity matters (793)

The lecture argues that individual cortical regions cannot be fully understood without knowing their inputs and outputs, since connectivity shapes what information a region receives and how it influences other regions. It lists reasons to study white matter: it is a large fraction of brain volume, connectivity patterns help define cortical areas and identify homologs across species, connectivity shapes development (citing rewired ferret studies), and disruptions in white matter are implicated in disorders such as dyslexia and autism.

Diffusion imaging and fiber tracts (1418)

The lecture explains diffusion imaging as a way to detect the direction of water diffusion, which is elevated along large, parallel fiber bundles. Major tracts such as the arcuate fasciculus, uncinate fasciculus, and inferior longitudinal fasciculus can be identified this way. A study from the Gabrieli lab is described in which lower fractional anisotropy in the arcuate fasciculus correlated with dyslexia in children, though the lecture stresses this is correlational.

Fractional anisotropy and the head-motion artifact (1703)

The lecture describes a study in which the standard finding that autism is associated with lower fractional anisotropy in long-range tracts disappeared once children were matched for head motion, and reappeared even in typically developing children when comparing their higher-motion scan session to their lower-motion session. This is presented as a warning that group differences in clinical neuroimaging studies can be artifacts of motion rather than real anatomical differences.

Tractography and its limits (2443)

Tractography follows local diffusion orientations to trace structural connections between gray matter regions. The lecture reports that in the lecturer's own collaborative testing, seeding in the lateral or medial geniculate nucleus incorrectly reaches multiple visual areas instead of only the expected target, illustrating known problems such as the crossing-fiber problem. The method is judged more reliable for producing approximate connectivity fingerprints than for confirming specific point-to-point connections.

Resting-state correlations and the default mode network (2938)

The lecture introduces Biswal's 1995 finding that motor regions in opposite hemispheres show correlated activity at rest, despite being physically distant and not engaged in any task. It describes the seed-based correlation method, and explains the default mode network as a set of regions more active during unstructured rest than during demanding tasks, associated with activities like recalling memories and thinking about other people, and later shown to be intercorrelated at rest.

Multiple-demand regions and network specificity (3145)

A second network, called the multiple-demand system, is activated by many unrelated demanding tasks (spatial working memory, arithmetic, perceptual judgments) and is linked by John Duncan's research to fluid intelligence, since damage to these regions reduces IQ regardless of location. The lecture closes with work by Edon Blank and colleagues showing that when language, multiple-demand, and theory-of-mind regions are identified individually in each subject and then measured at rest, each set is strongly correlated internally, the language and theory-of-mind systems show a modest positive correlation with each other, and neither is meaningfully correlated with the multiple-demand system, indicating these are distinct functional networks.

Before you watch

• Review the temporoparietal junction and theory-of-mind findings from the previous lecture on social cognition, since this lecture opens with a recap of them.
• Recall functional localizers and connectivity fingerprints from the earlier lecture on the fusiform face area, since the same individual-subject localization approach is used here.

Check your understanding

1. Why does the lecturer say diffusion tractography failed basic reality checks in his own data?
2. How did matching for head motion change the reported relationship between autism and fractional anisotropy?
3. What is the difference between how the default mode network was originally identified (task contrasts) and how it is identified now (resting correlations)?
4. What did Edon Blank's analysis show about correlations between the language system, the multiple-demand system, and the theory-of-mind system at rest?
5. Why does the lecture argue that individual brain regions cannot be fully understood without studying their connectivity?