Connectomics: Exploring the Amazing Neuronal Pathways Inside Our Brains

Every good expedition needs a crew. Ours is named after the real parts of a real neuron — so by the time you know the characters, you already know the anatomy.

Here is the whole theory of this page in one paragraph. Everything you have ever thought, remembered, or felt ran along physical wires — real, touchable threads of living cells, thinner than a wisp of spider silk, connected to each other at trillions of tiny junctions. That wiring diagram is called a connectome. Nobody has ever seen yours. Nobody has ever seen a complete one for any mammal. And a worldwide community of scientists, engineers, and volunteers is — right now, this decade — building the machines and methods to change that.

Why is that hard? Because of a brutal mismatch of scales. The wires (axons) can be tenths of a micrometer thick — far below what a light microscope can separate — so the only camera that can trace every wire is an electron microscope, which uses beams of electrons instead of light. But an electron microscope sees only a tiny window at a time, and a brain is enormous by comparison. Mapping even a millimeter of brain at wire-level detail means turning tissue into trillions of pixels: the completed cubic millimeter of mouse visual cortex weighed in at roughly two petabytes of images — about two million gigabytes, from a speck the size of a grain of sand [17].

So connectomics works like a relay race. One piece of brain is the baton. It passes through seven legs — preserved, stained, sliced, imaged, aligned, reconstructed, checked — and each leg transforms it: from tissue, to plastic, to slices, to pictures, to a digital volume, to a wiring diagram, to knowledge. Drop the baton at any leg and everything downstream fails. Run every leg well and you get something science has never had before: the complete circuit diagram of a thinking machine.

The Neuronauts will run the relay with you. Here is the whole route in one picture — seven streams, one baton:

The relay in one map. Each leg below wears its stream color.

Then — like all the best fan theories — we'll shuffle the scenes into chronological order and show you the arc: where the field has been, where it is, and where it is going.

Your guide: Captain Cortex — every expedition starts with the case for going.

Before the crew suits up, a fair challenge: neuroscience already has brain maps. MRI scanners map whole human brains in an afternoon — the Human Connectome Project charted the highways connecting brain regions in over a thousand people [37]. And microscopists have made gorgeous portraits of individual neurons since Santiago Ramón y Cajal's ink drawings in 1900. So why mount a whole expedition?

Because between those two views lies a gap — and the gap is where thinking happens. MRI sees bundles of millions of wires as one blurry highway; a single-cell portrait shows you one tree and no forest. Neither can tell you which neuron talks to which. The circuit — the actual computing machinery, thousands of specific cells wired in specific patterns — sits squarely in the missing middle. Filling that gap is the founding argument of this field, and it comes with four reasons worth knowing, adapted from the "Why Connectomics?" lecture notes this program grew out of [35]:

The data mountain

One more thing the briefing must be honest about: the climb. Here is the classic back-of-envelope ladder of what wire-level imaging costs in data, animal by animal — order-of-magnitude estimates these lecture notes have used for years to explain why each rung demands new machines [35]. (Real modern datasets land in the same territory: the fly volume came to roughly a hundred terabytes, and a single cubic millimeter of cortex to 1.4–2 petabytes [10] [17] [18].)

A zettabyte is a billion terabytes — for one brain. That number is not a reason to turn back; it is the mission plan. It tells you exactly why the relay you're about to run had to be invented leg by leg, and why the field climbs the mountain in stages: worm, fly, cubic millimeter, mouse. Suit up.

Your guide: Glia, engineer — because this leg is pure chemistry.

The baton: a living piece of brain → a perfectly preserved block of tissue, hard as plastic, with every membrane coated in metal.

You cannot photograph something that keeps wiggling. So the first leg of the relay is to stop biological time — perfectly, everywhere, all at once.

Chemists do it with fixatives: molecules like glutaraldehyde that rush into the tissue and stitch its proteins together where they stand, like flash-freezing a dance mid-step. Then comes a problem you might not expect: to an electron beam, brain tissue is almost transparent. The solution is to stain the tissue with heavy metals — most famously osmium tetroxide — which cling to the fatty membranes that wrap every neuron. Membranes are exactly what you need to see, because a membrane is the boundary line between one cell and the next. The metal turns every boundary into dark ink [5].

Doing this for a speck of tissue is classic 1960s electron microscopy. Doing it uniformly through a large block — so the deepest neuron is stained as crisply as the shallowest — took real invention: en-bloc staining protocols worked out in the 2010s, including in Moritz Helmstaedter's lab, pushed reliable staining from slivers to whole-brain-scale blocks [6]. Finally the tissue is embedded in resin: sealed in hard plastic, like an insect in amber. Except this amber was poured on purpose, and the fly inside is a library.

Your guide: Dendra, listener — because this leg rewards patience and a steady hand.

The baton: one solid block → thousands of intact, ordered slices, each tens of nanometers thin.

An electron microscope can only see surfaces. To see inside the block, you must turn it into slices — and the slices have to be absurdly, almost unbelievably thin.

How thin? Around 30 to 40 nanometers — a few thousand times thinner than a human hair. The tool is an ultramicrotome fitted with a knife whose edge is a polished diamond, shaving the block the way a deli slicer shaves ham, if the deli slicer worked at the scale of individual viruses. And here is the unforgiving part: the slices must survive in order. A lost or crumpled slice is a torn-out page — every wire passing through it gets cut off from itself.

The field has invented two great answers. One is to skip handling slices entirely: serial block-face microscopy puts the knife inside the microscope, photographing the block's face and then shaving it away, over and over — an approach introduced by Winfried Denk and Heinz Horstmann in 2004 that helped launch modern connectomics [7]. The other is to catch every slice automatically on a moving strip of tape (the "tape-collecting ultramicrotome"), so thousands of sections wind onto reels like movie film — the approach behind both a landmark saturated reconstruction of mouse cortex and the human H01 dataset [8] [18]. For scale: the whole-fly-brain volume that became FlyWire was cut into 7,062 serial sections, each about 40 nanometers thin [10].

Your guide: Axon, scout — because this leg is all about speed.

The baton: thousands of physical slices → millions of overlapping electron-microscope photographs.

Now every slice must be photographed at a resolution of a few nanometers per pixel. Do the arithmetic and the problem announces itself: at this detail, a single cubic millimeter becomes about two petabytes of images [17]. With one ordinary microscope, you would grow old waiting.

So the field built faster cameras. The flagship is the multibeam scanning electron microscope, which splits the electron source into 61 beamlets that scan side by side — one instrument doing the work of a small fleet [9]. Machines like these ran for months to image the H01 human cortex sample [18]; camera arrays on transmission microscopes did the same for the fly brain, producing 21 million images [10].

The MICrONS project added a twist that makes this leg even more remarkable: before the mouse's cortex was sliced, researchers spent months recording the activity of roughly 75,000 of its neurons through a microscope while the mouse watched movies. Then the very same tissue went through the EM pipeline. Same neurons: first their activity, then their wiring [17]. It is the difference between having a road map, and having a road map plus months of traffic data for every street.

Your guide: Captain Cortex, mission lead — because this leg is about seeing the whole from the pieces.

The baton: millions of separate photos → one seamless, digital 3D brain volume you can fly through.

Here's the catch nobody warns you about: after slicing and imaging, what you own is not a 3D brain. It is millions of separate 2D photographs — and physics has quietly vandalized them. Slices stretch, fold, wrinkle, and rotate; each image is a slightly warped postcard of reality.

Alignment (scientists say "registration") is the leg where software rebuilds the loaf from the slices. Algorithms find matching features between neighboring images and gently warp each one — like smoothing crumpled paper — until every neuron's path continues perfectly from one slice to the next. The elegant standard approach models the whole stack like a mesh of springs that relaxes into its most consistent shape [11].

It sounds like housekeeping. It is destiny. A misalignment of a few dozen nanometers can disconnect an axon from itself, and everything downstream — the tracing, the counting, the science — inherits that mistake. When the fly brain's 21 million images snapped into a single navigable volume, something changed philosophically too: from that moment on, the brain was software. Anyone, anywhere, could fly through it [10].

Your guide: Syn, connector — because this is where wires and junctions get their names.

The baton: a gray 3D photograph → a labeled world, where every voxel knows which neuron it belongs to.

Zoom into the aligned volume and you'll see the truth about brains: gray, tangled spaghetti in every direction. Beautiful — and unreadable. Leg 5 is where each strand of spaghetti gets its own color: this voxel belongs to neuron #4,271; that one to neuron #88,904. Scientists call it segmentation.

Humans did this by hand for decades, and the arithmetic was crushing. The very first connectome — the worm C. elegans, 302 neurons — took White, Southgate, Thomson, and Brenner on the order of fifteen years, tracing photographic prints with colored pencils [3]. Even with modern software, densely reconstructing 950 retinal neurons consumed more than twenty thousand hours of human annotation [24]. At that pace, a cubic millimeter of cortex — two hundred times as many cells — was simply out of the question.

Then came deep learning. In 2018, Google and Max Planck researchers published flood-filling networks — neural networks that "pour" a color into one neuron and let it flow along the cell's shape, stopping at the membrane walls that the heavy-metal stain made visible back in Leg 1 [12]. (See how the relay works? A chemistry decision made months earlier is what the AI leans on now.) Machine segmentation is what made the modern flagship datasets possible at all: 139,255 neurons in a fly brain [14], 200,000 cells and 523 million synapses in a cubic millimeter of mouse cortex [17] — with the synapses, too, detected automatically.

Your guide: Glia again — because somebody has to fix what the machines get wrong. Repair is her whole brand.

The baton: a machine's best guess → a trustworthy map, checked by human eyes — maybe yours.

Here is the leg that makes this story different from every other big-science story: the public doesn't just fund this work. The public does this work.

Even excellent AI makes two kinds of mistakes. A split breaks one neuron into pieces; a merge glues two neurons into a false hybrid. Each is a lie the map would tell forever if no one caught it. So reconstructions go through proofreading: people flying through the 3D data, comparing the colored reconstruction against the raw images, and fixing what the machines got wrong.

The field discovered something wonderful: this is a skill ordinary people can learn. The online game EyeWire turned retinal neurons into 3D puzzles, and its community — which grew to hundreds of thousands of registered players around the world — powered real published discoveries about how the eye detects motion [13]. Its descendant FlyWire went further: the finished fly connectome credits a community of proofreaders — scientists and trained citizen volunteers — and the gamers appear as co-authors on the Nature paper [14] [15]. Read that again: people who started by playing a browser game ended up with their names on one of the landmark neuroscience papers of the decade.

Your guide: Captain Cortex — because this is the question the whole relay was for.

The baton: a finished wiring diagram → answers. And better questions.

So you ran the relay and you're holding a connectome. What is it actually for? Fair question — and the field has receipts.

The proof-of-concept came from the eye. For decades, scientists argued about how the retina computes the direction of movement. Wiring maps of the retina settled it, revealing exactly which cells connect to which, with a spatial arrangement that produces direction-sensitivity — a decades-old debate closed by anatomy [19] [13] [24]. In the fly larva, a complete 3,016-neuron connectome exposed the architecture of a brain that learns — including circuit motifs resembling ones used in machine learning [21]. In mouse cortex, MICrONS delivered the double prize: because activity was recorded before the tissue was mapped (Leg 3, remember?), researchers could relate what neurons do to how they connect — uncovering a wiring principle by which inhibitory cells target their partners with unexpected specificity [17] [20]. And connectomes are becoming engines for theory: a network model of the fly's visual system, constrained only by the real wiring diagram plus deep learning, predicted how real neurons respond to motion [22].

One honest caveat, straight from the field's own elders: a connectome is a map of roads, not a recording of traffic. Connectivity alone doesn't tell you the signals, the chemistry, or the moods that flow over it [25]. That is not a weakness of the project; it is the reason connectomics is designed to be combined — with activity recording, with genetics, with behavior. The map is the foundation, not the finished house.

Here's the trick every great fan theory pulls at the end: take all the scenes you just watched and put them in chronological order. Suddenly you don't see seven separate techniques. You see one story arc, a century and a quarter long — slow, then fast, then unstoppable.

Start with the seven streams themselves. The brain-to-map relay runs 1→7 — but history did not build the legs in that order. Watch the streams cross when you sort them by the year each one became an industrial reality:

The reorder, drawn. Milestone years: slicing [7], discovery [19], alignment [11], proofreading [13], staining [6], imaging [9], segmentation [12].

Now zoom out from the streams to the whole century:

Notice the shape of that curve: an idea in 1900, a first map after 86 years, and then — once the machines and the AI and the crowds arrived — the leaps came almost yearly: 302 neurons in 1986, about 139,000 by 2024, half a billion synapses per cubic millimeter by 2025. Each jump took new chemistry, new machines, new algorithms, and more people — which is exactly why the next jump is a story about growing a field, not just growing a dataset.

The whole story in eight numbers

Scientists mark their predictions the way weather forecasters should: by confidence. So every claim below wears a badge. Established means it has already happened and is published. Emerging means the work is funded and underway, with results arriving. Aspirational means it is a goal the field openly aims for — a hope, clearly labeled as one. (Teachers: this three-badge discipline is the same one used across this site's public-engagement teaching kit.)

How the field grows

What the maps could buy us

One more reason to believe maps pay for themselves: we've run this experiment before. The Human Genome Project cost about $3.8 billion of federal investment; one industry-commissioned analysis estimated it had generated close to $800 billion in economic activity within two decades — an analogy, quoted with its provenance flagged, but the underlying lesson is solid: a completed map becomes infrastructure, the one-time cost every later discovery reuses for free. [43]

And here is the throughline, one last time. The worm took a handful of heroes fifteen years. The fly took hundreds of people, including gamers, four decades later. The mouse will take thousands — engineers and educators and proofreaders and students who haven't chosen a major yet. The map gets drawn exactly as fast as the field grows the people to draw it. That is not a footnote to this story. That is the plot.

You didn't just read a story — you picked up the field's actual vocabulary. Use it freely.

Want the full working vocabulary? The site keeps a complete connectomics dictionary.

Storytelling is the vehicle here; accuracy is the cargo. Every factual claim above traces to a numbered source below — peer-reviewed papers, official program announcements, or global health statistics. Counts like "139,255 neurons" describe a specific public data release, not a final truth about the animal; connectomes get corrected and improved over time. Where a claim is a hope rather than a result, it wears an Aspirational badge or is labeled a prospect in the text. Characters, metaphors, and jokes are ours; the numbers are the field's. If you find an error, tell us — this page holds itself to the same standard it teaches.