A Society of Networks: How We Learned to See and Steer the Brain's Hidden Systems

For most of the history of neuroscience, the brain was a map. If you wanted to understand a function, you found its place. Language lived here, in a patch of left frontal cortex that Paul Broca identified in 1861. Comprehension lived there, in Wernicke's region. By 1909, Korbinian Brodmann had divided the entire cortex into some fifty-odd numbered areas based on the microscopic architecture of their cells. This parcellation was so durable that neuroscientists still cite "Brodmann area 17" today. The implicit promise was seductive: with a fine enough map, you could pin every mental act to its address.

Even then, a different view held that function was distributed across the brain rather than filed in discrete locations, i.e., that you could damage tissue and lose a little of a capacity everywhere, rather than lose it entirely in one spot. The tension between localization and distribution would run, unresolved, through the rest of the century.

The Brain That Wouldn't Switch Off

Then, in the 1990s, the brain did something inconvenient... Researchers kept noticing a nuisance in their imaging data: a set of regions that reliably switched off whenever a subject concentrated on a demanding task, and switched back on the moment the mind was left to wander. It looked like noise. It was easy to ignore.

Marcus Raichle's group did the opposite. They asked what the brain was doing when it was doing "nothing", and in 2001 gave the answer a name: the default mode network [1]. The brain at rest, it turned out, is not idle. It is intensely, expensively active on its own terms, running an internal life of memory, imagination, and self-reference. We had been studying the brain by interrupting it, and had mistaken its resting voice for silence.

A second, quieter breakthrough had set the stage. In 1995, Bharat Biswal [2] noticed that spontaneous, slow fluctuations in widely separated regions rose and fell together, in lockstep, even when the subject lay still. If two regions fluctuate in sync, they are functionally connected, which means you could map the brain's networks from rest itself, with no task at all. This single insight turned network neuroscience from a curiosity into an industry.

Meet the Networks

The two decades that followed were a gold rush. Resting-state fMRI let researchers parcel the cortex into recurring, reproducible systems. By around 2011, Vinod Menon [3] had distilled a clinically powerful simplification, the triple network model, incorporating:

• The Default Mode Network (DMN): self-referential thought, memory, mind-wandering. When it goes awry, you see it in depression, rumination, and Alzheimer's.

• The Frontoparietal Network (FPN), or central executive: working memory, cognitive control, goal-directed attention.

• The Salience Network (SN): the switchboard that detects what matters and decides, moment to moment, whether the DMN or the FPN should hold the wheel.

But three is a tidy story, not the whole one. The widely used Yeo parcellation [4] describes seven canonical networks, adding the Dorsal Attention Network (voluntary, top-down focus), the Ventral Attention Network (stimulus-driven reorienting), the Visual Network, and the Somatomotor Network. Look closer, and these split further still, into limbic, language, cingulo-opercular, and other systems, to say nothing of the subcortical and cerebellar networks that surface methods reach only poorly.

Here is the first honest caveat worth putting in a blog post: the right number of networks is itself an open question. Depending on how finely you slice, you get anywhere from seven to seventeen or more. The networks are real, but their boundaries are partly a choice of resolution, like asking how many regions a country has.

fMRI vs. EEG: Mapping Brain Networks in Space and Time

Networks are invisible. To study them at all, you need to measure them, and here two technologies play complementary roles.

fMRI gives you space. It can localize a network to within millimeters and is how most of these systems were defined in the first place. But it is slow (it tracks blood flow, a sluggish proxy for neural activity), expensive, and immobile: a subject lying motionless in a multi-ton magnet.

EEG gives you time. Older than all of it, dating to the 1920s, EEG listens directly to the brain's electrical rhythms with millisecond precision. It is portable, affordable, and captures something fMRI cannot: the oscillations, the rhythmic frequencies at which networks hum and through which their nodes coordinate. Its weakness is spatial; the skull blurs the signal.

The punchline writes itself. fMRI tells you where; EEG tells you when and at what frequency. The modern synthesis describes the same systems in both languages at once, and EEG's portability is what makes network science thinkable outside the scanner room, in clinics, in homes, at scale.

From Measuring to Modulating: How tDCS, tACS, and tRNS Target Brain Networks

For all its elegance, everything above is correlational. We see that the DMN is overactive in depression; we cannot conclude that the overactivity causes the depression. To make causal claims and to intervene therapeutically, you need a way to gently push on a network and watch what happens.

That is what transcranial electrical stimulation (tES) offers: weak, non-invasive currents delivered through electrodes on the scalp. It comes in three main flavors, and the distinction matters for network targeting:

• tDCS (direct current) shifts a region's baseline excitability up or down. Think of it as adjusting a network's overall "tone."

• tACS (alternating current) delivers a rhythmic current at a chosen frequency, so it can entrain a network's natural oscillations and even strengthen or weaken the coupling between nodes by stimulating them in or out of phase. Because networks are defined partly by their rhythmic coordination, tACS is the most conceptually exciting tool for engaging them.

• tRNS (random noise) applies broadband noise that can enhance excitability and signal detection, exploiting a phenomenon called stochastic resonance.

Map these tools onto the networks, and the applications come into focus: nudging the DMN in depression and anxiety; supporting the FPN for working memory, attention, and cognitive enhancement; modulating the salience network in conditions where the brain mis-assigns importance to stimuli. The research is young, but the logic is clean, and stimulation does something measurement alone never can: it turns a correlation into a causal probe.

And this isn't distant, speculative research. Work is already out there showing how protocol-optimization methods (like Neuroelectrics' Stimweaver [5]) can determine how best to engage these networks, with encouraging results [6], [7].

The Frontier: Networks, Personalized

Here is where the threads converge, and where the most interesting open problems live.

Group-average networks may not map cleanly onto any individual brain. Your DMN is not quite shaped like mine. Networks also appear to be dynamic, reconfiguring moment to moment rather than sitting still as a fixed map. And reaching a specific, sometimes deep network through the scalp, without lighting up everything around it, remains genuinely hard.

Each of these problems points the same way: toward personalized, network-targeted neuromodulation. The vision is a closed loop: read an individual's brain with EEG, model how current will actually flow through their unique anatomy, choose which network, what frequency, and where to place the electrodes, then stimulate and measure the response, adjusting as you go.

This is precisely the frontier we work on at Neuroelectrics, where a single wearable system both records EEG and delivers tES, and where computational models of each person's head guide where the current goes. The ambition is not to stimulate a spot, but to engage a system: to treat the brain as the society of interacting networks that a century of science has revealed it to be.

What a Century of Brain Network Research Tells Us

If there is one through-line in this history, it is increasing humility. We began with confident maps of isolated spots. We discovered networks. Then we learned that those networks are dynamic, individual, and partly defined by the very methods we use to see them. Each step traded a little certainty for a lot of truth.

That humility is not a weakness in the science; it is the leading edge of it. The brain is not a map to be memorized but a system to be understood in motion, and increasingly, carefully, to be guided. We have spent a hundred years learning to listen. We are just beginning to learn how to respond.

References

[1] M.E. Raichle, A.M. MacLeod, A.Z. Snyder, W.J. Powers, D.A. Gusnard, & G.L. Shulman, A default mode of brain function, Proc. Natl. Acad. Sci. U.S.A. 98 (2) 676-682, https://doi.org/10.1073/pnas.98.2.676 (2001) 

[2] Biswal, B., Zerrin Yetkin, F., Haughton, V.M. and Hyde, J.S. (1995), Functional connectivity in the motor cortex of resting human brain using echo-planar mri. Magn. Reson. Med., 34: 537-541. https://doi.org/10.1002/mrm.1910340409

[3] Menon V. Large-scale brain networks and psychopathology: a unifying triple network model. Trends Cogn Sci. 2011 Oct;15(10):483-506. doi: 10.1016/j.tics.2011.08.003. Epub 2011 Sep 9. PMID: 21908230. 

[4] Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, Roffman JL, Smoller JW, Zöllei L, Polimeni JR, Fischl B, Liu H, Buckner RL. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol. 2011 Sep;106(3):1125-65. doi: 10.1152/jn.00338.2011. Epub 2011 Jun 8. PMID: 21653723; PMCID: PMC3174820. 

[5] Ruffini, G., Fox, M. D., Ripolles, O., Miranda, P. C., & Pascual-Leone, A. (2014). Optimization of multifocal transcranial current stimulation for weighted cortical pattern targeting from realistic modeling of electric fields. NeuroImage, 89(April), 216–225. https://doi.org/10.1016/j.neuroimage.2013.12.002

[6] Fischer, D. B., Fried, P. J., Ruffini, G., Ripolles, O., Salvador, R., Banus, J., Ketchabaw, W. T., Santarnecchi, E., Pascual-Leone, A., & Fox, M. D. (2017). Multifocal tDCS targeting the resting state motor network increases cortical excitability beyond traditional tDCS targeting unilateral motor cortex. Neuroimage, 157, 34–44. https://doi.org/10.1016/j.neuroimage.2017.05.060 

[7] Ruffini, G., Wendling, F., Sanchez-Todo, R., & Santarnecchi, E. (2018). Targeting brain networks with multichannel transcranial current stimulation (tCS). Current Opinion in Biomedical Engineering, 8, 70–77. https://doi.org/10.1016/j.cobme.2018.11.001