Updated September 2026

Timeline of Neuroscience.

40 checkpoints, from a stain that showed one cell at a time to a petabyte map of a cubic millimetre. Scroll sideways.

Seeing the Cell

18731/40

1873

The black reaction

Camillo Golgi finds a stain that reveals a handful of neurons in their entirety.

Nervous tissue had defeated microscopists because staining it uniformly turned everything into an impenetrable felt. Golgi’s silver chromate method did the opposite: for reasons still not fully understood, it blackened perhaps one cell in a hundred, completely, including the finest processes. Suddenly a single neuron could be seen whole against a clear background. Golgi himself read the result as evidence for a continuous reticulum.

Golgi, On the structure of the grey matter of the brain

Seeing the Cell

18882/40

1888

The neuron doctrine

Santiago Ramon y Cajal uses Golgi’s own stain to argue that neurons are discrete cells.

Cajal improved the method, applied it to young tissue where myelination had not yet obscured the picture, and drew what he saw with extraordinary care. Everywhere he looked, processes approached each other and stopped. He concluded that the nervous system is built from discrete cells communicating across gaps, and that signals flow in one direction, from dendrite to axon. He and Golgi shared the 1906 Nobel and disagreed in their lectures.

Cajal, Nobel Lecture

Seeing the Cell

18973/40

1897

The synapse gets a name

Charles Sherrington names the junction between neurons and shows it behaves unlike an axon.

Working on spinal reflexes, Sherrington found that transmission across the junction was slower than conduction along a nerve, fatigued more readily, and could inhibit as well as excite. Those asymmetries implied a real functional boundary rather than a continuous wire. He proposed the term synapse, from the Greek for clasping together, and spent decades mapping reflex integration across it.

Sherrington, The Integrative Action of the Nervous System

Seeing the Cell

19094/40

1909

Cortex divided into areas

Korbinian Brodmann maps the cortex into 52 regions by cell architecture alone.

Brodmann sectioned and stained cortex from humans and other mammals and classified regions purely by the arrangement and density of cell layers, with no functional data at all. The resulting numbered map looked like taxonomy for its own sake. It turned out to predict function remarkably well: area 17 is primary visual cortex, area 4 is primary motor. The numbers are still used, more than a century later.

Brodmann, Vergleichende Lokalisationslehre der Grosshirnrinde

Chemical Signalling

19215/40

1921

Chemical transmission proved

Otto Loewi shows that nerves signal by releasing a substance, not only by electricity.

Loewi dreamed the experiment, woke, scribbled an illegible note, and repeated the dream the next night. He stimulated the vagus nerve of one frog heart in a bath, took the fluid, applied it to a second heart with no nerve attached, and the second heart slowed too. Something diffusible carried the message. He called it Vagusstoff; it was acetylcholine, and the result opened all of neuropharmacology.

Loewi, Nobel Lecture

Chemical Signalling

19296/40

1929

The first human EEG

Hans Berger records electrical rhythms from an intact human scalp.

Berger worked in near-secrecy for five years, partly because he was chasing a theory of telepathy he never confirmed. What he did find was a reproducible rhythm around ten cycles per second that appeared when a subject closed their eyes and vanished when they opened them. The alpha rhythm was the first demonstration that the living human brain produces measurable, state-dependent electrical activity.

Berger, Uber das Elektrenkephalogramm des Menschen

Chemical Signalling

19527/40

1952

The action potential, quantified

Hodgkin and Huxley model the nerve impulse as voltage-gated sodium and potassium currents.

Using the squid giant axon, wide enough to take a wire down its middle, and a voltage clamp that held membrane potential fixed while measuring current, they separated the ionic components of the spike. The four coupled differential equations they fitted reproduce the action potential quantitatively and still form the backbone of computational neuroscience. It remains one of the most successful quantitative models in all of biology.

Hodgkin and Huxley, A quantitative description of membrane current

Systems and Memory

19538/40

1953

Patient H.M.

A surgery for epilepsy removes the medial temporal lobes and, with them, the ability to form new memories.

Henry Molaison could hold a conversation, reason, and retain his childhood, but could not lay down new episodic memories. Crucially, he could still learn motor skills while denying he had ever practised them. That dissociation proved memory is not one system but several, anatomically separable. He was studied for more than fifty years, and his brain was sectioned and digitised after his death in 2008.

Scoville and Milner, Loss of recent memory after bilateral hippocampal lesions

Systems and Memory

19599/40

1959

Feature detectors in visual cortex

Hubel and Wiesel find cortical cells that respond to oriented edges, in a hierarchy.

They were recording from cat visual cortex and getting almost nothing until a glass slide edge swept across the projector and a cell fired hard. The cells did not care about spots; they cared about bars at particular orientations. Simple cells fed complex cells fed hypercomplex cells, each pooling the last. The architecture they described is the direct ancestor of the convolutional network.

Hubel and Wiesel, Receptive fields of single neurones in the cat striate cortex

Systems and Memory

196210/40

1962

The split brain

Sperry and Gazzaniga show that severing the corpus callosum leaves two separately conscious hemispheres.

Patients whose hemispheres had been surgically disconnected to control epilepsy behaved normally in daily life. In the lab, an image shown only to the right hemisphere could be drawn by the left hand while the speaking left hemisphere denied seeing anything, then confabulated a reason for the drawing. The work made lateralisation concrete and raised questions about unified selfhood that are still live.

Sperry, Nobel Lecture

Molecules and Channels

197611/40

1976

Patch clamp resolves a single channel

Neher and Sakmann record the current through one ion channel opening and closing.

By pressing a fire-polished glass pipette against a cell membrane and pulling a gigaohm seal, they isolated a patch small enough to contain a single channel. The trace showed square steps: the channel is either open or shut, with no intermediate. It converted membrane biophysics from a statistical account of whole-cell currents into the direct observation of individual protein molecules working.

Neher and Sakmann, Single-channel currents recorded from acetylcholine receptor

Molecules and Channels

197112/40

1971

Computed tomography

Godfrey Hounsfield builds a scanner that reconstructs a cross-section of the living head.

X-rays had always flattened three dimensions into one shadow. Hounsfield rotated the source and detector around the head and used a computer to solve for the density at each point. The first clinical scan found a frontal lobe cyst. For the first time a physician could see inside a living skull without opening it, and neurology stopped being a discipline of inference from symptoms alone.

Hounsfield, Computerized transverse axial scanning

Molecules and Channels

197313/40

1973

Long-term potentiation

Bliss and Lomo show that a brief burst of stimulation strengthens a synapse for hours.

Hebb had proposed in 1949 that cells which fire together wire together, but it was a postulate with no mechanism. Bliss and Lomo, recording in rabbit hippocampus, delivered a high-frequency train and found the synaptic response stayed enlarged long afterwards. LTP gave learning a physical substrate that could be measured in slices, and it became the most studied phenomenon in cellular neuroscience.

Bliss and Lomo, Long-lasting potentiation of synaptic transmission

Molecules and Channels

197114/40

1971

Place cells

John O’Keefe finds hippocampal neurons that fire when an animal is in one particular location.

Recording from freely moving rats, O’Keefe noticed cells that were silent almost everywhere and fired vigorously in one spot in the enclosure. Move the rat, and a different cell took over. The population formed a map of space that persisted across visits. It was the first clear case of a cognitive variable, rather than a sensory or motor one, encoded in single-cell firing.

O’Keefe and Dostrovsky, The hippocampus as a spatial map

Imaging the Living Brain

197715/40

1977

Magnetic resonance imaging

Lauterbur and Mansfield turn nuclear magnetic resonance into a soft-tissue imaging method.

NMR could identify molecules in a test tube. Lauterbur added magnetic field gradients so that resonant frequency encoded position, and Mansfield worked out how to read the signal out fast enough to be practical. Unlike CT it used no ionising radiation and distinguished grey from white matter beautifully, which made it the default tool for structural neurology within a decade.

Lauterbur, Image formation by induced local interactions

Imaging the Living Brain

197916/40

1979

PET maps brain metabolism

Positron emission tomography images glucose use, tying activity to anatomy in living people.

A radiolabelled glucose analogue accumulates where cells are working hardest, and the positrons it emits annihilate into paired gamma rays that a ring of detectors can localise. For the first time a subject could perform a task and an experimenter could see which regions consumed more fuel. Cognitive neuroscience as an experimental discipline effectively begins here.

Phelps and Mazziotta, Positron emission tomography

Imaging the Living Brain

198017/40

1980

Memory reduced to molecules

Eric Kandel traces learning in Aplysia down to specific synaptic changes and gene expression.

Kandel chose a sea slug with enormous, individually identifiable neurons and a simple gill-withdrawal reflex. Habituation and sensitisation of that reflex mapped onto measurable changes at named synapses, and long-term storage required new protein synthesis via CREB. The reductionist bet paid off: the molecular machinery turned out to be broadly conserved, including in mammals.

Kandel, Nobel Lecture

Imaging the Living Brain

199218/40

1992

Functional MRI

The BOLD signal lets a standard MRI scanner image brain activity without any injection.

Ogawa had noticed that deoxygenated haemoglobin is paramagnetic and therefore changes the local magnetic environment. Because active tissue triggers an overshoot in blood flow, the ratio of oxygenated to deoxygenated blood shifts, and the scanner sees it. Non-invasive, repeatable and needing no tracer, fMRI made human cognitive neuroscience cheap enough to industrialise.

Ogawa et al., Intrinsic signal changes accompanying sensory stimulation

Imaging the Living Brain

199219/40

1992

Mirror neurons

Rizzolatti’s group finds premotor cells that fire both when acting and when watching the same act.

Recording in macaque premotor cortex, the team found neurons that discharged when the monkey grasped a peanut and also when it merely watched an experimenter grasp one. The finding was seized on as a mechanism for imitation, empathy and language, sometimes far beyond the evidence. The core observation is solid; the sweeping interpretations remain contested, which makes it a useful lesson in both directions.

di Pellegrino et al., Understanding motor events

Plasticity and Repair

199820/40

1998

Adult neurogenesis in humans

Eriksson and Gage find newly born neurons in the adult human hippocampus.

The textbook position for most of the twentieth century was that you are born with all the neurons you will ever have. Using tissue from cancer patients who had received BrdU, a marker that incorporates into dividing cells, the team found labelled neurons in the dentate gyrus. The scale and functional importance in humans are still debated, but the absolute prohibition fell.

Eriksson et al., Neurogenesis in the adult human hippocampus

Plasticity and Repair

199721/40

1997

Deep brain stimulation approved

An implanted electrode suppresses tremor, and a surgical lesion becomes an adjustable setting.

Benabid had observed during surgery that high-frequency stimulation of the thalamus silenced tremor as effectively as destroying the tissue, but reversibly. The implant turned an irreversible ablation into a parameter a clinician could tune. It now treats Parkinson’s, dystonia and essential tremor in hundreds of thousands of people, and is under trial for treatment-resistant depression and OCD.

Benabid et al., Chronic electrical stimulation of the ventralis intermedius

Plasticity and Repair

200222/40

2002

Brain-machine interfaces move a cursor

Monkeys learn to control a cursor directly from motor cortex activity, with no movement at all.

Nicolelis and Donoghue independently showed that a population of motor cortex neurons carries enough information about intended movement to drive an external device in real time, and that the animal adapts its firing to control the device better. The loop closed: neural activity out, feedback in, learning on both sides. Human trials followed within a few years.

Serruya et al., Instant neural control of a movement signal

Plasticity and Repair

200523/40

2005

Grid cells

The Mosers find entorhinal neurons that fire on a perfect triangular lattice across space.

Recording just upstream of the hippocampus in a large enclosure, they found cells whose firing fields tiled the environment in a hexagonal grid, at several nested scales. Nothing in the environment was hexagonal; the geometry was generated internally. Together with place cells it gave the brain something close to a coordinate system, and won the 2014 Nobel with O’Keefe.

Hafting et al., Microstructure of a spatial map in the entorhinal cortex

Control and Circuits

200524/40

2005

Optogenetics

A light-sensitive algal protein lets specific neurons be switched on with millisecond precision.

Boyden and Deisseroth expressed channelrhodopsin-2, an ion channel from green algae that opens in blue light, in mammalian neurons. Shine light, the neuron fires; stop, it stops. Combined with genetic targeting it made causal circuit experiments routine: not merely correlating activity with behaviour but driving a defined cell type and watching what the animal does.

Boyden et al., Millisecond-timescale genetically targeted optical control

Control and Circuits

200725/40

2007

Brainbow

Random combinations of fluorescent proteins give neighbouring neurons distinguishable colours.

The bottleneck in tracing circuits is that adjacent processes look identical. Lichtman’s lab engineered cells to express a stochastic mix of three fluorophores, producing around a hundred distinguishable hues. Individual axons could then be followed through dense tissue by colour alone. The images are genuinely beautiful and, more to the point, made dense reconstruction tractable.

Livet et al., Transgenic strategies for combinatorial expression

Control and Circuits

201326/40

2013

CLARITY makes tissue transparent

Replacing lipids with a hydrogel renders an intact brain optically clear while keeping its proteins.

Light scatters at every lipid membrane, which is why thick tissue is opaque and why anatomy meant slicing. Chung and Deisseroth cross-linked proteins and nucleic acids into a hydrogel mesh, then washed the lipids out electrophoretically. The result is a whole brain you can image through and stain repeatedly, without ever cutting it into sections and reassembling them.

Chung et al., Structural and molecular interrogation of intact biological systems

Control and Circuits

201327/40

2013

National brain projects

The US BRAIN Initiative and the EU Human Brain Project fund neuroscience at physics scale.

Both were announced within months and took opposite bets: BRAIN funded tool development broadly, the Human Brain Project centred on large-scale simulation. BRAIN produced cell atlases, new recording technologies and imaging methods. The HBP drew sustained criticism for its premise and was substantially restructured. The contrast is a useful case study in what big science can and cannot buy.

BRAIN Initiative

Connectomes

201828/40

2018

A full fly brain imaged at synapse resolution

Electron microscopy captures every neuron and synapse in an adult Drosophila brain.

The dataset is around a hundred teravoxels, assembled from tens of thousands of ultrathin sections imaged and aligned automatically. Tracing it by hand would have taken lifetimes; machine segmentation plus distributed proofreading made it feasible. It is the first complete synaptic wiring diagram of a brain capable of learning, flight and courtship.

Zheng et al., A complete electron microscopy volume of the adult Drosophila brain

Connectomes

201729/40

2017

Neuropixels probes

A single silicon shank records hundreds of neurons simultaneously across the depth of the brain.

Previous electrode arrays sampled tens of cells. Neuropixels puts around a thousand recording sites on a shank thinner than a human hair, fabricated with standard semiconductor processes so it can be produced at scale and cheaply. Large-scale population recording stopped being a specialist capability and became something an ordinary lab could do, which changed what questions were askable.

Jun et al., Fully integrated silicon probes for high-density recording

Connectomes

202130/40

2021

Speech decoded from cortex

An implant lets a man who cannot speak produce sentences by attempting to say them.

Chang’s group implanted an electrode array over sensorimotor cortex in a participant with anarthria after a brainstem stroke, and trained models to classify attempted words from the neural signal. The system produced sentences from a fifty-word vocabulary at useful rates. Later work extended the vocabulary substantially and added a synthesised voice and facial avatar.

Moses et al., Neuroprosthesis for decoding speech

Atlases and Interfaces

202331/40

2023

A human brain cell atlas

Single-nucleus sequencing catalogues more than three thousand cell types across the human brain.

Cell types had been defined by shape, location and firing pattern, which meant boundaries were arguments. Sequencing the RNA of individual nuclei at scale replaced that with clustering over expression, and produced a far finer taxonomy than anyone expected, with sharp regional differences. It gives every subsequent circuit study a common parts list to refer to.

Siletti et al., Transcriptomic diversity of cell types across the adult human brain

Atlases and Interfaces

202432/40

2024

A cubic millimetre of human cortex

Harvard and Google reconstruct every cell and synapse in a one-millimetre sample of human brain.

The fragment contains around 57,000 cells and 150 million synapses, and the dataset runs to about 1.4 petabytes. Machine segmentation did the tracing that humans could not. Among the findings were axons forming dozens of synapses onto a single target, a motif nobody had predicted. It is roughly a millionth of a human brain, which is a fair measure of how far a complete map remains.

Shapson-Coe et al., A petavoxel fragment of human cerebral cortex

Atlases and Interfaces

202433/40

2024

The fly connectome, annotated

FlyWire publishes a proofread wiring diagram of 140,000 neurons with cell types assigned.

Raw segmentation is not the same as a usable map. The FlyWire consortium crowdsourced proofreading and annotation until every neuron had a type and every connection a weight. Researchers immediately began simulating circuits from it and predicting behaviour, with some predictions confirmed experimentally. It is the first connectome complete and clean enough to compute with.

Dorkenwald et al., Neuronal wiring diagram of an adult brain

Atlases and Interfaces

201934/40

2019

Brain organoids develop rhythms

Stem-cell-derived tissue self-organises into cortex-like structure and generates oscillatory activity.

Grown from human pluripotent cells, organoids form layered structures with multiple cell types and, after months, produce coordinated electrical oscillations resembling those of a preterm infant. They are a genuine model system for human-specific development and disease, and they also force questions about moral status that the field has been slow to formalise.

Trujillo et al., Complex oscillatory waves emerging from cortical organoids

Atlases and Interfaces

202135/40

2021

Psychedelics re-enter clinical research

Controlled trials of psilocybin for depression report effects comparable to standard antidepressants.

After half a century in which the compounds were effectively unstudiable, randomised trials resumed under tight protocols. Results have been promising and genuinely hard to blind, since participants know which arm they are in. The mechanistic story involves a transient increase in cortical plasticity, which connects a clinical result back to the synaptic literature.

Carhart-Harris et al., Trial of psilocybin versus escitalopram for depression

Atlases and Interfaces

201536/40

2015

The gut-brain axis

Gut microbes are shown to influence host neurodevelopment and behaviour through defined pathways.

Germ-free mice raised without any microbiome show altered anxiety behaviour, blood-brain barrier permeability and microglial maturation, and colonising them shifts those measures. The effects run through vagal signalling, immune mediators and microbial metabolites. Much popular coverage has outrun the evidence, but the causal core is well replicated in animals.

Sharon et al., Specialized metabolites from the microbiome

Theory

199937/40

1999

Predictive coding

Rao and Ballard propose that cortex sends predictions down and only prediction errors up.

Feedback connections in cortex vastly outnumber feedforward ones, which is difficult to explain if perception is bottom-up feature extraction. Predictive coding inverts it: higher areas predict lower-level activity, and only the residual travels upward. The model explained extra-classical receptive field effects that had resisted explanation, and it has become the dominant theoretical frame for cortical function.

Rao and Ballard, Predictive coding in the visual cortex

Theory

200138/40

2001

The default mode network

Raichle identifies a set of regions that are more active at rest than during focused tasks.

Imaging studies routinely subtracted a rest condition as a baseline. Raichle asked what the baseline itself was doing and found a consistent, highly connected network that deactivates during external tasks and engages during recall, planning and mind-wandering. It reframed the brain as intrinsically active rather than reactive, and the network is disrupted in several psychiatric conditions.

Raichle et al., A default mode of brain function

Theory

199739/40

1997

Dopamine encodes reward prediction error

Schultz shows midbrain dopamine neurons signal the difference between expected and received reward.

The cells fire to an unexpected reward, stop firing to a fully predicted one, and dip below baseline when an expected reward fails to arrive. That is precisely the error term in temporal difference learning, a reinforcement learning algorithm developed independently in computer science. It is the cleanest known correspondence between a machine learning quantity and a measured neural signal.

Schultz et al., A neural substrate of prediction and reward

Theory

202640/40

2026

Scale meets the hard part

Recording and mapping have industrialised; explanation has not kept pace.

It is now routine to record thousands of neurons, map millimetres of tissue at synaptic resolution and catalogue thousands of cell types. What has not arrived is a theory that turns those measurements into understanding of how a brain produces behaviour. The gap between description and explanation is the field’s central problem, and it is not obviously a data problem any more.

Allen Institute for Brain Science

End of the line

40 checkpoints, and the hard part is still ahead.

Almost every entry here is a new way of looking. The recurring lesson is that the instrument comes first and the understanding follows, sometimes by decades — and that the understanding is the part that is currently missing.

Back to the site Read the essays

1873Scroll or drag