Nervous system

The nervous system is the body's communication network of specialized cells. It receives sensory information, integrates it, and produces coordinated responses, from fast reflexes to voluntary movement, perception, and memory. In vertebrates it is divided into the central nervous system, made up of the brain and spinal cord, and the peripheral nervous system, which links the central system to the rest of the body. Its principal units are neurons, which send signals electrically along their membranes and chemically across synapses, and glial cells, which support, insulate, and protect them.

The nervous system matters because nearly every aspect of animal life depends on it. It controls movement, sensation, and cognition, and it also regulates internal conditions such as heart rate, breathing, digestion, and body temperature, working alongside the hormonal system to maintain homeostasis. Its disorders, including stroke, neurodegenerative disease, and psychiatric illness, are among the leading causes of disability worldwide, which makes the field central to medicine and neuroscience alike.

This entry covers the basic cell biology and electrical and chemical signaling of neurons, the anatomy of the central and peripheral divisions and their major regions, and the way the system develops, changes with experience, and evolved across animals. It then reviews the history of its study, the main research methods and imaging techniques, the principal neurological and psychiatric disorders, the relationship between biological neural networks and artificial neural networks, and the questions that remain open.

Definition and organizing principles

The nervous system is the body's main network for rapid signaling and control. It is built from neurons, which send electrical impulses along their axons and pass chemical messages at synapses, supported by glial cells. Its signals typically act within milliseconds and reach specific target cells. By contrast, the hormonal system releases chemical messengers into the bloodstream, where they reach many tissues more slowly and produce effects that can last from seconds to days. The two systems are tightly linked, and many hormone-producing glands are themselves controlled by neural circuits, but the difference in speed and specificity is the usual starting point for describing the nervous system.

The nervous system performs three linked functions. The first is sensing: specialized receptors detect stimuli from outside the body, such as light, sound, touch, and chemicals, and from inside it, such as blood pressure, blood glucose, muscle stretch, and tissue damage, and convert them into electrical signals. The second is integration: incoming signals are combined, compared with stored information, and weighed against the organism's current state. In vertebrates this occurs mainly in the brain and spinal cord, although simpler ganglia can process information locally. The third is response: the system drives effectors, chiefly skeletal muscle, smooth muscle, cardiac muscle, and glands, to produce movement or to change organ function and secretion. Repeated sensing, integration, and response loops help maintain homeostasis and make adaptive behavior possible.

Anatomically, the system is divided into the central nervous system (CNS), which comprises the brain and spinal cord, and the peripheral nervous system (PNS), which comprises the nerves, ganglia, and receptors lying outside them. Functionally, it is divided by the direction of information flow relative to the CNS. Afferent (sensory) pathways carry information from receptors toward the CNS, while efferent (motor) pathways carry commands away from it to muscles and glands. The efferent side is further divided into the somatic division, which controls skeletal muscle, and the autonomic division, which regulates the heart, smooth muscle, and glands largely without conscious control. These two schemes overlap rather than match: a single peripheral nerve usually contains both afferent and efferent fibers.

Nervous systems are found in most animals, although their complexity varies enormously. Sponges (Porifera) lack neurons and have no true nervous system. Cnidarians, including jellyfish and hydra, have diffuse nerve nets in which signals spread across a sheet of interconnected cells. Most bilaterian animals show centralization, with nerve cords and ganglia concentrated toward the anterior end of the body. Plants, fungi, and single-celled organisms do not have nervous systems. Neurons are specialized excitable cells that depend on multicellular organization and on dedicated transmission machinery. A single cell can sense and respond to its environment using its own membrane and internal chemistry, and plants show rapid electrical signaling, as in the Venus flytrap, but these responses do not involve neurons. Plants and fungi instead rely on chemical, hydraulic, and electrical signals that work without a specialized conducting tissue of this kind.

Neurons and glia

Neurons are the cells that generate and transmit the electrical and chemical signals on which the nervous system depends, but they share the tissue with glia, a diverse group of non-neuronal cells. In the human brain, non-neuronal cells are numerically comparable to neurons, although the ratio varies considerably from region to region.

Structure of a typical neuron

Most neurons have four recognizable parts. The cell body, or soma, holds the nucleus and most of the machinery for making proteins, and in many neurons it is only about 10–25 micrometres across. Dendrites are shorter, frequently branched extensions that act as the principal input surface, receiving chemical signals from other cells through receptors on their membranes and through the synapses that cover them. The axon is a single, slender process that leaves the soma at the axon hillock, where the neuron's membrane is most excitable, and carries action potentials away from the cell body. Axons vary enormously in length; the longest axons of human motor neurons extend more than a meter, from the base of the spinal cord to the toes. Near their targets, axons divide into axon terminals (presynaptic terminals), which are packed with mitochondria and release neurotransmitter into the synaptic cleft.

Classification by function

Neurons are commonly grouped by the role they play in circuits. Sensory neurons, also called afferent neurons, carry information from receptors in the skin, muscles, viscera and sense organs toward the central nervous system. Motor neurons, or efferent neurons, carry commands from the central nervous system to skeletal muscles and glands. Interneurons lie between these two groups, mostly within the brain and spinal cord, and integrate incoming signals before passing them on. Because the three groups are defined by function rather than form, a single anatomical type can serve more than one role.

Classification by morphology

Structurally, neurons are often sorted by the number of processes that emerge from the soma. Unipolar neurons have a single process from the cell body that then branches. In vertebrates, the sensory neurons that detect touch, pressure, pain and temperature are usually pseudounipolar: they develop from bipolar precursors whose two processes fuse close to the soma, so one stalk divides into a peripheral branch and a central branch. The true unipolar form is more characteristic of invertebrates, and the term is sometimes misapplied to vertebrate neurons. Bipolar neurons have two processes arising from opposite poles of the soma and are found in specialized sensory structures such as the retina, the olfactory epithelium and the cochlear and vestibular pathways. Multipolar neurons have one axon and many dendrites. They make up the majority of neurons in the central nervous system, and they include most motor neurons and interneurons. The two classification schemes overlap only partly: sensory neurons are typically pseudounipolar or bipolar rather than multipolar.

Glia in the central nervous system

Astrocytes are the most abundant macroglial cells in the brain and spinal cord. Their branching processes anchor neurons to blood vessels, and their endfeet contribute to the blood–brain barrier. They help regulate extracellular potassium and other ions, take up and recycle neurotransmitters such as glutamate, and take up glucose from the blood, converting part of it into glycogen for storage. Oligodendrocytes extend several flattened processes that wrap around segments of multiple axons, forming the myelin sheaths of central axons; a single oligodendrocyte can myelinate parts of several neurons. Microglia are small, mobile cells that act as the resident immune cells of the central nervous system. They are macrophage-like, removing dead cells, debris and invading microorganisms, and they change their shape and gene expression rapidly when tissue is damaged or infected. Ependymal cells, which line the ventricles and the central canal and contribute to cerebrospinal fluid circulation, are also usually counted among central glia.

Glia in the peripheral nervous system

Schwann cells myelinate peripheral axons. Unlike oligodendrocytes, each Schwann cell wraps a single internodal segment of a single axon, so one axon is myelinated by a chain of many Schwann cells. After nerve injury, Schwann cells clear myelin and debris and release growth-promoting factors that help damaged axons regenerate. Satellite cells surround the neuronal cell bodies inside sensory, sympathetic and parasympathetic ganglia. They help control the chemical environment around these neurons, supply them with nutrients and structural support, and respond to injury and inflammation. Satellite cells have been implicated in chronic pain, though the evidence for their role in specific pain conditions is still developing.

Shared functions and responses to injury

Across both divisions, glia provide structural support, supply nutrients and oxygen, insulate neurons from one another, and destroy pathogens or remove dead neurons. They also help maintain ion and neurotransmitter concentrations within narrow limits, which is essential for the reliable firing of action potentials. Myelin is the most consequential of these functions for signalling speed, because it allows impulses to jump between the gaps in the sheath known as nodes of Ranvier. Glia also respond to damage: microglia and astrocytes become activated after injury or infection, and the inflammatory signals they release can both limit tissue damage and, when prolonged, contribute to further injury. Glia were once regarded mainly as passive support cells, but research over recent decades has shown that they actively regulate synaptic function, metabolism and immune responses, which is why they feature prominently in studies of disease.

Resting potential and action potentials

Resting membrane potential

At rest, the inside of a typical neuron is about -70 mV relative to the outside. For neurons, the resting membrane potential is around -70 millivolts, indicating that the inside of the cell is 70 millivolts more negative relative to the outside. This voltage comes from an uneven spread of charged particles across the membrane. Sodium, chloride and calcium are kept at higher concentrations outside the cell, while potassium and various anions are kept inside, so the membrane holds two opposing gradients. For a typical neuron at rest, sodium, chloride, and calcium are concentrated outside the cell, whereas potassium and other anions are concentrated inside.

The membrane's permeability decides which gradient wins. Lipid bilayers block ions, so passage depends on channels, and at rest a population of non-gated "leak" channels stays open. The majority of these are potassium leak channels, allowing for regular exchange of potassium ions. To a much lesser extent, there are few chloride leak channels, and very few sodium leak channels. A commonly cited ratio is roughly 100 potassium to 45 chloride to 4 sodium channels. Calculations based on the Nernst equation show that potassium channels alone would drive the membrane to about -90 mV, near potassium's equilibrium potential; the chloride and sodium leaks pull that value back toward -70 mV. The resting voltage therefore lies between the equilibrium potentials of the individual ions, closer to potassium's.

Leak channels alone would gradually dissipate the gradients, so an active process is needed to maintain them. The sodium–potassium pump (Na⁺/K⁺-ATPase) uses energy from ATP hydrolysis to move ions against their gradients. For every molecule of ATP that is hydrolyzed, 3 Na ions are pushed out into the extracellular space, and 2 K+ ions are brought into the intracellular space. The pump does not set the resting voltage directly; its work keeps the ionic concentration differences, and therefore the leak-driven potential, stable over time.

Phases of the action potential

An action potential begins when a depolarizing input pushes the membrane to threshold, typically near -55 mV. At that point voltage-gated sodium channels open. The rapid influx of sodium results in a large depolarization called the rising phase. The voltage overshoots zero and peaks at roughly +30 mV before the falling phase begins.

Two processes then return the membrane toward rest. Sodium channels inactivate: they close through a separate gate and cannot reopen immediately. Approximately 1 msec after action potential initiation, the sodium channels inactivate. Meanwhile, voltage-gated potassium channels, which respond more slowly, open and allow potassium to leave the cell. This allows an efflux of potassium ions, which causes the repolarization of the falling phase. Because the potassium channels close slowly, the membrane briefly becomes more negative than rest, a phase called the undershoot or afterhyperpolarization. Finally, the pump and leak channels restore the resting state.

The event is all-or-none. Once threshold is crossed, the sodium and potassium channel sequence runs to completion with a stereotyped shape and size, regardless of how strongly the trigger exceeded threshold. Stimulus intensity is instead encoded in how many impulses are generated per second and in which neurons are activated.

Refractory periods

Two refractory phases limit firing. During the absolute refractory period, it is impossible to evoke another action potential in this period. This is because sodium channels are inactivated and cannot reopen regardless of input. The relative refractory period follows. Voltage gated potassium channels are slow to close, meaning that even after the sodium channels are ready to go again the membrane will require a greater depolarization to reach threshold, giving rise to the relative refractory period. A stronger stimulus can then trigger a spike. These phases set the maximum firing rate of a neuron and ensure that the impulse moves in one direction. The refractory period created by inactivated sodium channels prevents backward propagation, while closed sodium channels ahead allow continued forward movement of the action potential.

Propagation in unmyelinated and myelinated axons

In an unmyelinated axon, each patch of membrane depolarizes the next, so the action potential is regenerated continuously along the axon. Conduction is slow in this mode. Reported speeds for unmyelinated fibres are roughly 0.5 to 10 m/s.

Myelin changes this. Myelin is produced by glia (Schwann cells in the peripheral nervous system and oligodendrocytes in the central nervous system) and wraps the axon in insulating segments. In myelinated axons, voltage-gated Na+ channels are clustered at gaps in the myelin sheath called nodes of Ranvier. Current flows through the myelinated internodes and regenerates the action potential only at the nodes. The impulse therefore appears to hop from node to node, a process called saltatory conduction. Because the sheath reduces leakage and capacitance, the signal reaches the next node quickly, and the regenerative step happens only where sodium channels are concentrated.

Published speed figures vary with fibre diameter, species and measurement method. One source gives 150 m/s compared from 0.5 to 10 m/s for myelinated versus unmyelinated fibres, while another reports propagation that can exceed 10–50 m/s in humans. The speed advantage of myelin is consistent across these estimates, but the exact values should be read as approximate.

Synapses and neurotransmitters

Chemical synapses

Most signaling between neurons crosses chemical synapses. At these junctions the presynaptic terminal faces the membrane of a target cell across a gap of roughly 20 to 40 nanometres, the synaptic cleft. When an action potential reaches the terminal, voltage-gated calcium channels open. The resulting brief rise in calcium concentration near the release sites drives synaptic vesicles to fuse with the presynaptic membrane. Fusion depends on SNARE proteins, and synaptotagmin acts as the main calcium sensor. Each vesicle holds thousands of transmitter molecules, and release is probabilistic: a single spike typically empties only a fraction of the docked vesicles. The transmitter diffuses across the cleft in a fraction of a millisecond and binds receptors on the postsynaptic membrane. Ionotropic receptors are ligand-gated ion channels that produce a fast current, while metabotropic receptors couple to G proteins and act through second messengers over longer timescales. The signal ends when transmitter is taken back up by neurons or glia, broken down enzymatically (acetylcholinesterase at the neuromuscular junction is the classic case), or simply diffuses away.

Excitatory and inhibitory transmission

The effect of a transmitter depends mainly on the receptor it binds, not on the molecule itself. Glutamate, the main excitatory transmitter in the brain, acts on AMPA and kainate receptors, which are cation channels that depolarize the cell. NMDA receptors also admit calcium but are blocked by magnesium at resting potential until depolarization relieves the block, which makes them coincidence detectors central to long-term potentiation. GABA, the main inhibitory transmitter in the adult brain, acts on GABA-A receptors, chloride-permeable channels whose effect depends on the chloride electrochemical gradient. GABA-B receptors are metabotropic and act through potassium channels. Glycine provides fast inhibition in the spinal cord and brainstem.

A single excitatory postsynaptic potential is usually small, on the order of a fraction of a millivolt at the soma, so the neuron must integrate many inputs. Summation occurs in space, as hundreds to thousands of synapses are distributed over the dendrites, and in time, as closely spaced inputs overlap. Dendrites are not passive: NMDA-dependent and other dendritic spikes make the integration partly nonlinear. Inhibitory inputs can also act by shunting the membrane. The decision to fire is made at the axon initial segment, where the summed potential must reach the threshold for an action potential.

Neurotransmitter classes

Transmitters fall into several chemical families. Amino acids, including glutamate, GABA, glycine, and aspartate, carry most fast signaling. Monoamines, including dopamine, norepinephrine, serotonin, and histamine, are made from amino acid precursors by relatively few neurons whose axons project widely; they act mainly through G protein–coupled receptors, with serotonin 5-HT3 a notable exception that is ligand-gated. Acetylcholine is released at the neuromuscular junction and by autonomic preganglionic neurons, acting on nicotinic receptors (ion channels) and muscarinic receptors (G protein–coupled). Neuropeptides, such as substance P, oxytocin, somatostatin, and the endogenous opioids, are packaged in dense-core vesicles, released mainly during high-frequency firing, and act slowly and over longer distances through G protein–coupled receptors. Gases such as nitric oxide act as diffusible messengers that are not stored in vesicles.

The receptor families reflect these differences. Ligand-gated channels include the Cys-loop family (nicotinic, GABA-A, and glycine receptors) and ionotropic glutamate receptors. The large G protein–coupled receptor family includes most monoamine and peptide receptors. Early in the twentieth century, Dale's principle held that a neuron releases the same transmitter from all of its terminals; later work showed that many neurons co-release several transmitters.

Electrical synapses

Electrical synapses link cells through gap junctions, channels formed by connexin proteins in vertebrates and innexins in invertebrates. These channels let ions and small molecules pass directly from one cytoplasm to the next. Current flows in both directions, and transmission is nearly instantaneous, with no vesicle step and no receptor-binding delay. The trade-off is that electrical coupling offers little amplification, is less easily modulated, and cannot by itself produce the sign-changing, plasticity-prone signaling that chemical synapses provide.

Electrical synapses are most valuable where timing and synchrony matter. They help synchronize oscillating networks, such as the inferior olive, and they are central to rapid escape circuits, such as the Mauthner cell system in fish. In the retina they shape signals between neurons. Many circuits combine both types, so that a fast electrical link can be paired with a slower, modifiable chemical connection between the same cells.

Central and peripheral divisions

The most basic anatomical split divides the nervous system into the central nervous system (CNS), made up of the brain and spinal cord, and the peripheral nervous system (PNS), which comprises the remaining neural tissue. The boundary is drawn largely by bone. The brain sits inside the cranial cavity and the spinal cord runs through the vertebral canal, so structures within those cavities are conventionally central and structures outside them are peripheral. Anatomists acknowledge that this is an oversimplification, since some peripheral elements lie inside the skull or spine. A more precise criterion is glial. The CNS is myelinated by oligodendrocytes and the PNS by Schwann cells, and the transition between the two occurs at a point near where the nerve roots enter the spinal cord and brainstem. The optic nerve is a notable case: although it is called a nerve, it develops as an outgrowth of the brain and is composed of central tissue.

The CNS develops from the neural tube, which forms from the neural plate of the ectoderm during early embryogenesis. The rostral end of the tube expands into three primary vesicles, the forebrain, midbrain, and hindbrain, which later subdivide into the structures of the adult brain. The caudal portion of the tube becomes the spinal cord. The PNS arises mainly from neural crest cells and from ectodermal placodes, which together contribute the sensory and autonomic ganglia and the Schwann cells that wrap peripheral axons.

Cranial and spinal nerves

The PNS is organized around two sets of nerves. There are 12 pairs of cranial nerves, numbered I to XII from front to back according to where they attach to the brain. The olfactory (I) and optic (II) nerves arise from the cerebrum, while nerves III to XII emerge from the midbrain, pons, and medulla. Most cranial nerves serve the head and neck, although the vagus (X) extends to the thoracic and abdominal viscera. Individual cranial nerves may be sensory, motor, or mixed, and some contain sensory ganglia outside the brainstem.

Spinal nerves comprise 31 pairs: 8 cervical, 12 thoracic, 5 lumbar, 5 sacral, and 1 coccygeal. Each spinal nerve divides into a dorsal root, which carries sensory fibers whose cell bodies lie in a dorsal root ganglion outside the cord, and a ventral root, which carries motor fibers. Cervical nerves are named after the vertebra beneath them, while the remaining spinal nerves are named after the vertebra above. The ventral branches of most spinal nerves (all except T2 to T12) interlace to form plexuses: the cervical, brachial, lumbar, and sacral plexuses. These networks redistribute fibers so that each limb nerve contains contributions from several spinal segments. Thoracic nerves run more directly to the trunk as intercostal nerves.

Grey and white matter

Both the brain and spinal cord contain grey matter, which is rich in neuronal cell bodies, dendrites, glia, and unmyelinated fibers, and white matter, which consists mainly of myelinated axons. The two tissues are arranged in opposite patterns in the two organs. In the spinal cord, grey matter forms a butterfly-shaped core, with its anterior horns containing motor neurons, and white matter surrounds it, carrying ascending sensory and descending motor tracts. In the cerebrum and cerebellum the arrangement is reversed: the cerebral cortex is an outer layer of grey matter, while deeper white matter forms the bulk of the hemispheres. Deep grey nuclei such as the basal ganglia are embedded within this white matter. The brainstem shows an intermediate pattern, with grey nuclei scattered through white tracts. Functionally, grey matter is associated with processing information and white matter with transmitting signals between regions.

Anatomical terminology

Positions and directions in anatomy are described relative to the standard anatomical position: standing upright, facing forward, arms at the sides, and palms facing forward. Superior and inferior refer to up and down, anterior and posterior to front and back, and medial and lateral to the midline and the sides. Proximal and distal describe closeness to the trunk or to the point of attachment. In the body, ventral and dorsal are often used as synonyms for anterior and posterior, and this practice carries over to the spinal cord. Neuroanatomists frequently use rostral (toward the front of the brain) and caudal (toward the spinal end), because the embryonic brain bends and the dorsal and superior directions do not always coincide within it. Standard planes of section are the sagittal, which divides the body into left and right portions, the coronal, which divides it into front and back, and the axial or transverse, which divides it into upper and lower parts. Terms such as ipsilateral and contralateral describe whether a structure lies on the same side or the opposite side of the body as a reference point, a distinction that matters for the crossing of sensory and motor pathways.

Brain structure and regional functions

The cerebral cortex and its lobes

The cerebral cortex is a thin sheet of grey matter, roughly 2–4 mm thick, that covers the two cerebral hemispheres. In the neocortex, which makes up most of it, neurons are arranged in six layers, and the surface is folded into ridges (gyri) and grooves (sulci) that greatly increase its area within the skull. Each hemisphere is divided into four lobes: frontal, parietal, temporal, and occipital. The two hemispheres are linked mainly by the corpus callosum, a large bundle of white-matter fibres.

Primary sensory and motor areas are organized as topographic maps. The primary motor cortex lies in the precentral gyrus of the frontal lobe, and the primary somatosensory cortex lies just behind it in the postcentral gyrus of the parietal lobe. Both are arranged somatotopically, so that the body surface is represented in an orderly sequence, with the hands and face occupying disproportionately large areas. The primary visual cortex sits in the occipital lobe along the calcarine sulcus, and the primary auditory cortex lies on the superior surface of the temporal lobe. Higher-order regions surround these areas. In most people, language production is associated with Broca's area in the left inferior frontal lobe and language comprehension with Wernicke's area in the left posterior temporal lobe.

Subcortical structures

Beneath the cortex lie several nuclei that act as relays, filters, and regulators. The thalamus receives nearly all sensory input (olfaction is the main exception) and relays it to the appropriate cortical areas, while also participating in sleep, arousal, and attention. The basal ganglia, a group including the striatum, globus pallidus, and substantia nigra, receive extensive input from the cortex and are thought to help select which movements or actions are initiated and which are suppressed. Their dependence on dopamine from the substantia nigra is why the loss of these dopaminergic neurons produces the motor symptoms of Parkinson disease.

The limbic system is a traditional but loosely defined grouping of structures, including the amygdala, hippocampus, and cingulate cortex, that is strongly involved in emotion, motivation, and memory. The hippocampus is essential for forming new long-term declarative memories, as shown by the amnesia that followed bilateral medial temporal lobe removal in the patient studied by Scoville and Milner in 1957. The hypothalamus, a small structure below the thalamus, links neural activity with hormone release and regulates homeostasis of temperature, thirst, hunger, and circadian rhythms.

Brainstem and cerebellum

The brainstem, made up of the midbrain, pons, and medulla oblongata, contains the nuclei of most cranial nerves and provides the main route between the brain and spinal cord. It also houses circuits that are essential for life. The medulla contains the rhythm-generating networks for breathing, including the pre-Bötzinger complex, and centres that regulate heart rate and blood pressure. The reticular formation, which extends through the brainstem, helps control arousal and wakefulness, and severe damage to its ascending projections can cause coma. Because these circuits are so vital, injuries to the brainstem are often life-threatening even when they are small.

The cerebellum lies behind the brainstem and occupies only about a tenth of the brain's volume, yet it contains the large majority of the brain's neurons. It receives information about the state of the body and the intended movement, and it compares the two to adjust timing and force. Damage to the cerebellum typically produces ataxia, which is an uncoordinated gait and clumsy limb movement with preserved strength. Its role in motor learning is supported by the way error signals reach it through climbing fibres. Its contributions to cognition and emotion are an active area of study, but they are less firmly established than its role in movement.

Lateralization and large-scale networks

The two hemispheres are not functionally identical. Split-brain studies in the 1950s and 1960s, which were led by Roger Sperry and Michael Gazzaniga, showed that the left hemisphere in most people specializes in language, while the right is more involved in certain spatial and visuospatial tasks. Language dominance is strongly left-lateralized in most right-handed people, and it is more variable in left-handed people. Popular claims that individuals are "left-brained" or "right-brained" overstate these differences, since most mental activity engages both hemispheres.

More importantly, most cognitive functions do not depend on a single region. Functional imaging and lesion studies have shown that tasks such as attention, memory, and decision-making recruit distributed networks. Analyses of resting-state activity, such as the work of Yeo and colleagues in 2011, identified a set of large-scale networks, including the default mode network, the dorsal attention network, and the salience network, that interact with one another. As a result, the functional consequences of a lesion depend both on the damaged site and on the networks in which that site participates.

Spinal cord and reflex circuits

The spinal cord is the caudal part of the central nervous system. In adults it runs from the base of the skull to about the first lumbar vertebra. It relays information between the brain and the body, and it also contains circuits that can produce movement with little or no input from above. In cross-section, the grey matter forms a butterfly-shaped core. The dorsal horns receive sensory input, the ventral horns contain the motor neurons that drive skeletal muscle, and intermediate zones hold interneurons and autonomic neurons. The white matter surrounding this core consists of bundles of myelinated axons, grouped into columns, that carry signals up and down the cord. These bundles are the tracts discussed below.

Ascending and descending tracts

Ascending tracts carry sensory information toward the brain. Most follow a chain of three or four neurons. First-order neurons carry signals from receptors into the spinal cord through the dorsal root ganglia. Second-order neurons, located in the spinal cord or brainstem, relay the signal onward, and third-order neurons in the thalamus pass it to the cerebral cortex. Different pathways carry different modalities. The dorsal column system conveys fine touch, vibration and conscious proprioception, and it ascends on the same side until it reaches the medulla. The spinothalamic tracts carry pain, temperature and crude touch, and their second-order fibres cross to the opposite side shortly after entering the cord. The spinocerebellar tracts deliver unconscious proprioceptive information to the cerebellum.

Descending tracts carry motor commands. The corticospinal tract, which originates in the motor cortex, is the main route for voluntary and skilled movement of the limbs. Most of its fibres cross to the opposite side in the pyramids of the medulla and form the lateral corticospinal tract, while a smaller anterior tract crosses at spinal levels. Brainstem-origin tracts, including the reticulospinal, vestibulospinal, rubrospinal and tectospinal tracts, adjust posture, balance, muscle tone and the gain of reflexes. Because the descending tracts synapse on spinal interneurons and motor neurons, the spinal cord carries out the final stage of most voluntary movement.

Monosynaptic and polysynaptic reflexes

Reflexes are the simplest examples of spinal circuits. The stretch reflex is the most direct. When a muscle is lengthened, its stretch-sensitive receptors in the muscle spindles increase their firing. These afferent fibres enter the dorsal horn and synapse directly onto alpha motor neurons supplying the same muscle, which contracts to resist the stretch. The stretch reflex is monosynaptic and has the shortest latency of any spinal reflex. The knee-jerk reflex is the familiar example. At the same time, a branch of the afferent input activates inhibitory interneurons that relax the antagonist muscle, a process called reciprocal inhibition. Gamma motor neurons adjust how sensitive the spindles are.

The withdrawal reflex is polysynaptic. Nociceptive fibres of the Aδ type activate interneurons in the dorsal horn. Excitatory interneurons drive the flexors of the same limb, and inhibitory interneurons quiet the extensors, so the foot or hand is pulled away from a harmful stimulus. The crossed-extensor reflex usually accompanies it. Interneurons send signals across the midline so that the opposite limb extends and supports the body's weight during the withdrawal. Because these circuits contain one or more interneurons, they are slower than the stretch reflex but can produce more coordinated responses. Signals from higher centres can facilitate or suppress reflexes, but they cannot abolish them entirely.

Central pattern generators

A central pattern generator (CPG) is a network of spinal interneurons that produces rhythmic motor output. Such networks can generate alternating flexor and extensor activity for walking, and similar circuits underlie swimming and hopping in other animals. Classic experiments showed that the isolated spinal cord of some animals can generate rhythmic output without any input from the brain or from sensory feedback. In intact animals, however, descending commands from the brainstem and midbrain start locomotion and adjust its speed and pattern, and sensory feedback from the limbs helps set the timing of each step. In cats, hindlimb stepping can reappear after complete spinal transection when the animal is supported on a treadmill. After partial spinal injury, recovery of stepping depends in part on plastic changes within the spinal CPG shaped by training. Evidence for a human CPG is more indirect, and current research explores whether neuromodulation can reactivate these networks after injury.

Spinal cord injury

Injury to the spinal cord interrupts the tracts that link the brain to the body. Loss of function therefore appears below the level of damage. A lesion in the cervical region usually affects the arms and legs, while a lesion in the thoracic or lumbar region typically spares the arms. Complete injuries abolish voluntary movement and sensation below the lesion, and they also disrupt bladder, bowel, sexual and blood-pressure control, which depend on autonomic pathways. In the acute phase, many reflexes below the injury are depressed, a state known as spinal shock. Over time, some reflexes may become exaggerated because descending inhibition is lost. Partial injuries leave remnant pathways that can support partial recovery, and rehabilitation strategies aim to use the spinal circuits that remain. Central nervous system damage of this kind is distinct from injury to peripheral nerves, because the cord contains both the tracts and the reflex circuits that connect them.

Sensory input and motor output

Sensory input and motor output form the two ends of the nervous system's control loop. Receptors in the skin, muscles, joints, and sense organs convert physical and chemical events into electrical signals, and descending pathways turn central decisions into muscle contraction. Between these ends lie ascending relays in the central nervous system and feedback circuits that correct movement as it happens.

Sensory receptors and transduction

Receptors are usually classified by the kind of stimulus they detect. Mechanoreceptors respond to deformation and pressure, thermoreceptors to temperature, and photoreceptors to light. Chemoreceptors respond to chemical compounds, including the odorant molecules of smell, and nociceptors respond to damaging or potentially damaging stimuli that are felt as pain. Osmoreceptors track the solute concentration of body fluids, and baroreceptors in vessel walls register stretch, which lets the nervous system follow blood pressure. Some authors reduce this scheme to four broad modalities: chemical, mechanical, light, and temperature.

Transduction is the step in which a stimulus's energy becomes a change in membrane potential, usually by opening or closing ion channels in the cell membrane. The mechanism varies by modality. In mechanoreceptors, the channel proteins are themselves gated by force, so deforming the membrane directly alters ion flow. Skeletal muscles contain two types of proprioceptors, called muscle spindles and Golgi tendon organs, which transduce muscle length and muscle force, respectively, into electrical signals. Vertebrate photoreceptors run the other way: in the vertebrate photoreceptor, light activation of the visual pigment stimulates the hydrolysis of cGMP, which closes a cyclic nucleotide-gated channel and hyperpolarizes the cell. Mature cochlear hair cells, by contrast, lack voltage-gated sodium channels, so their transduction currents produce graded receptor potentials, of up to about 40 mV, rather than spikes. A receptor potential is a graded electrical signal that decreases in amplitude with distance along the membrane. In many somatosensory endings, a graded depolarization of this kind is the input that, once it reaches threshold, launches action potentials along the axon.

Ascending pathways and sensory maps

Most somatosensory neurons have their cell bodies in dorsal root ganglia, and their axons enter the spinal cord and ascend. The dorsal column–medial lemniscus pathway carries fine touch, vibration, and proprioception. Its fibres cross to the opposite side in the medulla, relay in the ventral posterolateral nucleus of the thalamus, and reach the primary somatosensory cortex of the postcentral gyrus. The spinothalamic tract carries pain and temperature, crossing soon after entering the cord. Vision follows a similar logic: retinal signals pass through the lateral geniculate nucleus to the primary visual cortex in the occipital lobe, and hearing passes through brainstem nuclei and the medial geniculate body to auditory cortex. Smell is the notable exception, because olfactory signals reach the cortex without a thalamic relay.

Each main relay preserves spatial order. The body surface is mapped onto the cerebral cortex in a somatotopic arrangement, and areas with dense receptor populations, such as the hands and lips, occupy disproportionately large territory in the sensory homunculus.

Motor pathways

Voluntary movement depends mainly on upper motor neurons in the primary motor cortex. The two descending pathways travelled by the axons of Betz cells are the corticobulbar tract and the corticospinal tract. The crossing of these fibres is anatomically important: the defining landmark of the medullary-spinal border is the pyramidal decussation, which is where most of the fibers in the corticospinal tract cross over to the opposite side of the brain. Within the spinal cord, the lateral branch serves limb muscles, while the anterior corticospinal tract supplies axial muscles, and its fibres do not decussate in the medulla. Corticobulbar fibres end on cranial motor nuclei and control the movement of muscles in the face, head and neck.

Upper motor neurons synapse on lower motor neurons in the brainstem and ventral horn of the spinal cord. These neurons project through the spinal nerves to connect to the muscles at neuromuscular junctions. The functional unit of this output is the motor unit. One motor neuron connects to multiple muscle fibers within a target muscle. The ratio varies with the task: the number of fibers innervated by a single motor neuron varies on the basis of the precision necessary for that muscle and the amount of force necessary. Muscles used for fine control, such as those moving the eye or fingers, therefore have small motor units, while large postural muscles have large ones. Brainstem pathways, basal ganglia, and the cerebellum also shape this output, a topic covered elsewhere.

Proprioceptive feedback

Motor output is continually adjusted by sensory feedback. Collectively referred to as the proprioceptive sense, this afferent information informs the central nervous system on the position of the body and limbs in space. Within each muscle, intrafusal fibres are proprioceptive sensor organs inside muscle spindles, specially modified muscle fibres that remain contractile and mainly sense muscle stretch and length. Spindle afferents are among the fastest sensory fibres in the body: each spindle is innervated by group Ia and group II afferent fibers, with a conduction velocity up to 120 m/sec. When a spindle is overstretched, its signal reaches the spinal cord, where it drives the stretch reflex (a reflex circuit covered in the spinal cord section). Gamma motoneurons regulate spindle sensitivity, playing a crucial role in proprioceptive feedback during muscle activity.

Golgi tendon organs, which lie at the junction between muscle and tendon, provide complementary information about force rather than length. Recent modelling work suggests that when combined with tendon organ feedback, muscle spindle feedback can distinguish self-generated from externally imposed forces. This combination of spindle and tendon signals is one reason the motor system can tell how much it is pulling against the world, not only where a limb is.

Autonomic and enteric control

Two branches and their wiring

The autonomic nervous system (ANS) belongs to the peripheral nervous system, but it works as a control system for the body's internal environment. It carries instructions between the central nervous system and organs such as the heart, lungs, smooth muscle, and exocrine and endocrine glands, and it does most of this without conscious effort. It controls the lungs, the heart, smooth muscle, and exocrine and endocrine glands. The autonomic nervous system controls these organs largely without conscious control; it can continuously monitor the conditions of these different systems and implement changes as needed.

Its two divisions differ in where they originate. The parasympathetic nervous system interacts via the cranial and sacral segments of the central nervous system, and the sympathetic nervous system arises from the T1–L2 spinal cord segments. Both use a two-neuron relay: a preganglionic neuron in the CNS synapses on a postganglionic neuron, which then acts on the target organ. The anatomy also differs in the position of the relay ganglia. Sympathetic ganglia lie close to the spinal cord, while parasympathetic ganglia sit within or near the visceral effector organs. Sympathetic fibers have short preganglionic and long postganglionic axons; parasympathetic fibers have long preganglionic and short postganglionic axons.

Effects on target organs

The divisions often push an organ in opposite directions, but the rule is not absolute. Many organs are controlled primarily by either the sympathetic or the parasympathetic division. Sometimes the 2 divisions have opposite effects on the same organ. Sympathetic activation dilates the pupils, raises heart rate and contractility, relaxes the airways, and reduces intestinal motility, while parasympathetic activation constricts the pupils, slows the heart, constricts the airways, and increases gut motility. What matters is the ratio between them, called autonomic tone, rather than one system simply switching the other off. Some targets, such as sweat glands, most blood vessels and the adrenal medulla, receive sympathetic input without an opposing parasympathetic input.

Central control and reflexes

Autonomic output is coordinated from above. A major output of the autonomic nervous system is the hypothalamus via the reticular formation. Brainstem nuclei in the medulla and pons also integrate signals for cardiovascular and respiratory control, and reflex loops, such as baroreceptor feedback from the carotid sinus and aortic arch, adjust heart rate and vascular tone within seconds to keep blood pressure stable. Clinical cases show how much depends on supraspinal control. After a spinal cord injury, supraspinal influence on the ANS is disrupted, leading to sympathetic blunting and parasympathetic dominance, which can cause cardiac dysrhythmias, systemic hypotension, and bowel and bladder dysfunction.

The enteric nervous system

The enteric nervous system (ENS) is a dense network embedded in the gut wall, often described as a semi-independent "second brain." It is a network of more than 100 million nerve cells (neurons) in the wall of the gut, from the esophagus to the anus. Its neurons are arranged in ganglionated plexuses: submucosal ganglia and myenteric ganglia distributed between the longitudinal and circular muscles. With these neural components the enteric nervous system can elicit neural reflexes that control and coordinate motility, secretion, and blood flow. Local sensory neurons detect luminal contents and drive motor output through interneurons, so peristalsis and secretion can proceed without central input. The ENS can operate independently from the brain and spinal cord, meaning that you can still digest food even if you're unconscious.

Extrinsic input still modulates the gut. The ENS is affected by the parasympathetic nervous system, which promotes digestion, and the sympathetic nervous system, which inhibits digestion. The classification of the ENS is not settled. Previously, the ENS was considered part of the autonomic nervous system, but it is now generally treated separately, occasionally being referred to as the "second brain."

Stress, exercise and somatic coordination

The ANS does not act alone. The ANS maintains normal internal function and can work in concert with the somatic nervous system. During acute stress, sympathetic outflow is recruited quickly. Sympathetic preganglionic fibers that reach the adrenal medulla release acetylcholine onto nicotinic receptors on chromaffin cells, which in turn secrete norepinephrine and epinephrine mediating the fight or flight response. Sympathetic activity also increases the conversion of glycogen to glucose by the liver, while decreasing digestion to provide the energy for "fighting or fleeing." During exercise, the same pattern supports working muscle: heart rate and blood pressure rise to raise skeletal muscle blood flow, and parasympathetic influence on the heart is withdrawn, while gut blood flow and motility are reduced. These changes are coordinated with motor commands from the somatic system, so that cardiovascular and respiratory adjustments match the demands of movement.

Protective barriers and cerebrospinal fluid

Bone and meninges

The brain and spinal cord are enclosed in bone: the skull surrounds the brain and the vertebral column surrounds the spinal cord. Inside the bony case, three membranes called the meninges wrap the central nervous system. The outermost, the dura mater, is thick and fibrous and anchors the brain. It also contains the dural venous sinuses, which drain blood from the brain. The middle layer, the arachnoid mater, is thin and avascular, and its name refers to the fine strands of connective tissue that run from it to the pia. The innermost layer, the pia mater, adheres closely to the surface of the brain and spinal cord and carries blood vessels. The arachnoid and pia are often grouped as the leptomeninges, and the space between them, the subarachnoid space, holds cerebrospinal fluid. Together, bone and membranes limit mechanical injury, while the fluid cushions the soft tissue so that it does not strike the skull during movement.

Production, circulation and drainage of CSF

Cerebrospinal fluid (CSF) is a clear, low-protein fluid. Most of it is made by the choroid-plexus, a tuft of capillaries and specialized ependymal epithelial cells found in the walls of the lateral ventricles and the roofs of the third and fourth ventricles. CSF formation combines filtration from the blood with active secretion by epithelial cells, which is why its composition differs from plasma. Tight junctions between choroid plexus epithelial cells regulate how water and solutes pass from the capillaries into the ventricles.

From the lateral ventricles, the fluid flows through the interventricular foramina into the third ventricle, then through the cerebral aqueduct into the fourth ventricle. It leaves the fourth ventricle through its lateral and median apertures and enters the subarachnoid space around the brain and spinal cord. It also fills the central canal of the spinal cord. In adults, the total volume is roughly 125–150 mL, and about 500 mL is produced each day, so the whole volume is replaced several times daily.

Most CSF returns to the blood through arachnoid granulations, which are outpocketings of the arachnoid that project into the dural venous sinuses. Smaller routes include the choroid plexus itself and perivascular pathways. The classical account, in which fluid moves in a single "river-like" stream from the choroid plexus to the arachnoid granulations, has been questioned. Surgical inhibition of production did not reliably relieve the symptoms of hydrocephalus, and some computational work proposes that water moves through the brain parenchyma under Starling-type forces. The mechanisms therefore remain an active research question. Drainage is driven mainly by the pressure difference between the subarachnoid space and the venous sinuses, and reabsorption slows when CSF pressure falls to low values. When drainage fails, fluid accumulates and pressure rises, a condition known as hydrocephalus.

The blood–brain barrier

The blood–brain barrier (BBB) is a functional interface at the brain's capillaries rather than a membrane in the usual sense. Its core is the endothelium of cerebral capillaries, which lacks fenestrations and whose cells are sealed to one another by tight junctions. These junctions contain proteins such as claudin-5, occludin and ZO-1, and they form a diffusion barrier that excludes most blood-borne molecules. Small lipophilic molecules can cross passively, while glucose, amino acids and other nutrients are carried across by specific transporters. Efflux transporters actively pump many compounds back out of the endothelium.

Pericytes, embedded in the basement membrane, help control how many tight junctions form and influence the polarization of astrocytic end-feet. The role of astrocytes is debated. Some reviews describe them as essential for inducing and maintaining the endothelial barrier, while others note that they are not thought to provide the barrier themselves. Laser-ablation experiments found that removing astrocytic endfeet did not immediately cause leakage, yet mice lacking astrocyte gap-junction proteins show barrier dysfunction. The current view is that astrocytes support and tune the barrier rather than sealing it alone.

The choroid plexus forms a separate blood–CSF barrier, and the arachnoid also contributes to barrier function. The BBB and blood–CSF barrier therefore differ in location and cellular composition.

Consequences for drug delivery and infection

Because the BBB excludes most large or polar molecules, many drugs that act on the central nervous system penetrate it poorly, which is one reason CNS drug development has a low success rate. Strategies to address this include designing small lipophilic compounds, exploiting endogenous transporters, administering drugs directly into the CSF (intrathecal delivery), and transiently opening the barrier with techniques such as focused ultrasound.

The same selectivity limits the entry of pathogens and toxins from the blood. Infection and systemic inflammation, however, can alter tight junctions, glial function and transport, increasing permeability. Inflammatory mediators such as cytokines are one route by which peripheral infection affects the brain. Infections of the meninges, such as meningitis, involve the CSF-filled compartments directly, and barrier breakdown is also implicated in neurodegenerative and other neurological disorders.

Development from the neural tube

Neurulation

Vertebrate nervous system development starts with neurulation, the process that turns embryonic ectoderm into the neural tube, the structure that eventually becomes the brain and spinal cord. Cells of the axial mesoderm send inductive signals to the overlying ectoderm, and these signals switch on genes that drive the ectoderm to become a thick, multilayered neuroepithelium. The mechanism of the induction is complex and should be guided by inductive substances that are secreted by the cells of the axial mesoderm. The resulting neural plate forms first at the head end and lengthens backward as it recruits more neuroectodermal cells. In humans this happens during the third week, when the plate's edges lift into neural folds around a central groove.

Closure is a mechanical as well as a molecular event. The lateral sides of the plate rise and meet at the dorsal midline, where they fuse into a closed tube, and the open ends of the tube seal by the end of the fourth week of gestation. Neurulation occurs in two stages. In amphibians, birds, and mammals, primary neurulation occurs along most of the rostrocaudal axis, whereas secondary neurulation, which involves condensed tail-bud cells, occurs caudally in the lower sacral and coccygeal regions only. In mice, closure begins at a few sites and then proceeds in both directions in a process often described as zippering, which means the closure sites are not identical across species. Failures of this process cause neural tube defects, which are among the most common severe congenital malformations.

Neural crest

Cells at the edges of the neural folds that do not enter the tube become the neural crest. The neural crest is a transient multipotent cell population that originates in the dorsal neural tube. These cells then migrate extensively through the embryo. Neural crest cells generate neurons and glia of the peripheral nervous system, and melanocytes throughout the body. The cranial crest also contributes ectomesenchymal tissues, including cartilage, bone, and adipose tissue, which are normally considered of mesodermal origin. The range of fates is large: in humans, at least 47 cell types have been defined as neural crest derivatives.

Fate is shaped by position along the body axis. Schwann cells, melanocytes and sensory and sympathetic neurons can all derive from the truncal region of the neural tube, while some derivatives originate from specific regions of the crest, such as enteric ganglia from the vagal and sacral regions. Clonal labelling experiments in chick embryos showed that individual early crest cells can be multipotent: the clonal progeny of injected cells were found to reside in many or all of the locations to which neural crest cells migrate, including sensory and sympathetic ganglia, peripheral motor nerves and the skin, and phenotypic analysis revealed that at least some neural crest cells are multipotent in vivo. Later, these cells become segregated into different sublineages, which generate restricted subsets of differentiated derivatives.

Neurogenesis, migration and axon guidance

Once the tube forms, its wall consists of proliferating progenitor cells. Neurons are produced by division of these progenitors, and newborn neurons migrate from their birthplaces to their final positions, often guided by fibres of radial glia. Each neuron then extends an axon that must find a distant target. The growing tip of the axon, the growth cone, is the key sensing structure. More than a century ago Ramón y Cajal identified the growth cone at the tip of extending axons, and proposed that it senses extracellular chemical cues and produces force for axon guidance.

The molecular cues that steer growth cones fall into several families. Signaling mechanisms that act downstream of four prominent families of axon guidance cues—netrins, semaphorins, ephrins, and slits—have been extensively studied in both invertebrate and vertebrate model systems. Netrins can act as both attractants and repellents: attractive or repulsive responses to netrin depend on the receptors expressed on the growth cone, with one receptor type mediating attraction and a different receptor combination mediating repulsion. Semaphorins work over a shorter range, deflecting the axon's growth cone from certain regions and correcting its pathway. The classic idea that netrin acts as a diffusible gradient has been challenged, and some studies suggest that a more local, close-range effect is important. Because a limited set of cues must specify very complex wiring, guidance signals act on axonal growth cones in a combinatorial manner, and are often integrated in a non-additive fashion such that their combined effects are different from their individual actions.

Overproduction, pruning and activity-dependent refinement

Developing circuits are not built by precise placement alone. Far more neurons and synapses are generated than the mature nervous system retains, and many cells die by programmed cell death, a process long linked to competition for limited target-derived trophic support such as nerve growth factor. Synapse density in many human cortical regions peaks in early childhood and then falls substantially as connections are eliminated. Pruning removes weak or inappropriate synapses, and in some systems glial cells, including microglia, and complement proteins help tag and clear them. Neural activity is the principal guide to which connections survive: in the visual system, retinal inputs that fire together tend to remain, while those that fire out of step are weakened. In this way, the ongoing refinement of wiring links early genetic programmes to the experience-dependent adjustments discussed in the section on plasticity.

Plasticity, learning and memory

Synapses are not fixed. Their efficiency changes with patterns of activity, and this capacity, usually called synaptic plasticity, is considered one of the main cellular bases of learning and memory. The best-studied model is long-term potentiation (LTP), first described in the rabbit hippocampus in 1966. Its counterpart, long-term depression (LTD), weakens connections. Together they give a neuron's inputs a way to be strengthened or weakened over hours to years. The detailed mechanisms are most often worked out at glutamatergic synapses in the CA1 region of the hippocampus, where a brief burst of high-frequency stimulation produces a potentiation that can last for hours or longer.

Cellular models: LTP and LTD

In CA1, the NMDA-type glutamate receptor acts as a coincidence detector. Its activation can lead to a long-lasting modification in synaptic efficiency at glutamatergic synapses, and in this region NMDA receptors mediate the induction of LTP, while the expression of the potentiated state is thought to depend mainly on an increase in AMPA receptor-mediated transmission. Induction requires that presynaptic glutamate release coincides with strong postsynaptic depolarization, which removes the magnesium block of the NMDA channel and allows calcium to enter the dendritic spine. The calcium signal activates kinases such as CaMKII, which promote the insertion of additional AMPA receptors into the postsynaptic membrane and increase the size and strength of the spine. Experiments in the 1990s and later showed that the NMDA receptor's subunit composition affects how readily LTP is induced: NR2A- and NR2B-containing receptors contribute differently, and extrasynaptic NMDA receptors can also influence the threshold for potentiation. LTD, by contrast, follows low-frequency stimulation that produces a more modest calcium rise, which favours phosphatase activity and the removal of AMPA receptors from the synapse.

Hebbian principles and coordinated change

Donald Hebb proposed in 1949 that when one cell repeatedly takes part in firing another, the connection between them is strengthened. In modern terms, this is a principle of correlated activity: synapses are modified according to the timing and coincidence of pre- and postsynaptic events. Spike-timing-dependent plasticity, in which the relative order of spikes determines whether a synapse is strengthened or weakened, is a more precise version of this idea. Hebbian rules are sometimes called associative because LTP has properties such as associativity, cooperativity and input specificity: a weak stimulus can be potentiated when it is paired with a strong one, and changes are confined to the active synapses rather than spreading to the whole cell. These features make LTP a plausible mechanism for storing information about co-occurring events, though it remains a model rather than a proven account of any particular memory.

Critical periods and experience-dependent wiring

Some circuits are especially sensitive to experience during restricted developmental windows. In the visual cortex, depriving one eye of input during early life, as shown by Hubel and Wiesel in the 1960s, shifts cortical responses toward the open eye, and the effect largely disappears if the deprivation starts in adulthood. Similar sensitive periods exist for the perception of speech sounds and for the acquisition of a first language, where exposure in early childhood usually yields more complete mastery than exposure later on. Critical periods are thought to reflect the maturation of inhibitory circuits and the stabilization of extracellular matrix, which together close the window for rewiring. Sensory experience therefore does more than tune existing circuits; it determines which connections persist, a theme that connects with the pruning described in the developmental sections of this entry.

Structural plasticity, neurogenesis and consolidation

Learning also changes the physical structure of neurons. Dendritic spines can form, enlarge and disappear over hours to days, and sustained training is associated with changes in spine number and morphology. Some of these structural changes are stabilized only when protein synthesis occurs during a critical interval after learning, which is why many models of memory consolidation depend on transcription and translation. Consolidation is also thought to involve a reorganization across systems: memories are initially dependent on the hippocampus and gradually become supported by distributed neocortical networks, a process that is thought to be aided by replay of activity during sleep. In adult mammals, new neurons are generated in limited regions, most clearly in the subgranular zone of the dentate gyrus and the subventricular zone lining the lateral ventricles, where they integrate into existing circuits. The functional importance of adult hippocampal neurogenesis in humans remains debated, and estimates of its extent vary between studies.

Neuroendocrine and neuroimmune interactions

The hypothalamic–pituitary–adrenal axis

Hormonal stress control is organized mainly around the hypothalamus, the pituitary gland and the adrenal cortex. The acute neuroendocrine response to adverse stress is characterized by the tripartite activation of three stress axes: the sympathetic nervous system, direct neural innervation of the adrenal cortex, and a cascade of hypothalamic hormonal messengers. The hormonal branch starts when a cluster of hypothalamic cells in the paraventricular nucleus secretes corticotropin-releasing hormone and vasopressin into the vessels that serve the pituitary (neurosecretory neurons in the paraventricular nucleus of the hypothalamus release both corticotropin-releasing hormone and arginine vasopressin into the portal circulation of the pituitary gland). These two factors act together on pituitary cells to release adrenocorticotropic hormone into the blood. That hormone then binds a melanocortin receptor in the adrenal cortex and triggers new synthesis of glucocorticoids: cortisol in humans and corticosterone in rodents (ACTH activates the melanocortin 2 receptor in the zona fasciculata of the adrenal cortex to initiate de novo synthesis and release of glucocorticoids).

Glucocorticoid output is not constant. It follows a daily cycle, with low levels during rest and a peak near waking, and this cycle is overlaid by near-hourly pulses (The activity of the HPA axis follows a distinctive circadian pattern with low glucocorticoid levels during the resting period, which increase to a peak around the time of wakening. Underlying this circadian rhythm is a highly dynamic ultradian rhythm of glucocorticoid release.). Pulse timing matters because pulsatile signals regulate transcription in the liver and brain (glucocorticoid pulsatility is important for dynamic transcriptional regulation of target genes in both the liver and the brain). The hormones also feed back on the system: because their receptors are distributed widely through the brain and peripheral tissues, they suppress further activation of the axis (glucocorticoids inhibit further HPA axis activation through intracellular receptors that are widely distributed throughout the brain and peripheral tissues). When this regulation drifts, the consequences can be significant. Chronic dysregulation can show up as an exaggerated cortisol response or a blunted one (Hyperactivity refers to an exaggerated response of the HPA axis, with heightened cortisol levels even during non-stressful periods. Blunted reactivity describes a diminished cortisol response to stressors.). Sustained high glucocorticoid exposure has also been linked to hippocampal changes that affect memory (prolonged stress with high glucocorticoid levels may lead to atrophy of the hippocampus).

Cytokines, sickness behavior and brain function

Immune signals reach the brain through several routes. Infection triggers the de novo production of cytokines by activated innate immune cells, and these molecules change behavior as well as physiology. The best-known pattern is sickness behavior, in which affected individuals become less active and less responsive, sleep more, eat less and withdraw socially (Inactivity, decreased responsiveness to external stimuli, sleepiness, decreased appetite, and social withdrawal are the usual signs of a microbial infection.). Interleukin-1 and tumor necrosis factor alpha are central to this response: their effects in the brain are thought to produce an organized motivational state (These changes are due to the effects in the brain of proinflammatory cytokines such as interleukin-1 (IL-1) and tumor necrosis factor alpha (TNFα) and represent the expression of a well-organized central motivational state.).

Several entry routes have been proposed. Cytokines may cross the blood–brain barrier by a saturable transport mechanism, act on circumventricular organs that lack the barrier, or activate vagal afferent neurons (possibilities include the direct entry of cytokine into the brain across the blood-brain barrier by a saturable transport mechanism, the interaction of cytokine with circumventricular organs such as the organum vasculosum of the lamina terminalis and area postrema, which lack the blood-brain barrier, and activation of afferent neurons of the vagus nerve). In response to immune challenge, peripheral cytokines and immune cells such as T cells can also pass the barrier through active transport (in response to immune challenge, peripheral cytokines and immune cells like T cells can pass through a protective blood–brain barrier using several active transport systems). Brain cells such as microglia also produce cytokines locally, which can lead to neuroinflammation (cells like microglia and brain mast cells secrete cytokines on the brain side of the BBB, a major consequence of which is neuroinflammation). Peripheral cytokines reach the brain partly through the HPA axis and partly through vagal input (cytokines of peripheral origin can modulate the CNS by acting on the hypothalamic–pituitary–adrenal (HPA) axis, or via afferent inputs of the vagus nerve). The loop also runs the other way. Endogenous glucocorticoids and vasopressin dampen the sickness-inducing effects of cytokines (The sickness inducing effects of cytokines are downregulated by a number of endogenous neuropeptides and hormones, including vasopressin and glucocorticoids.). When immune activation is prolonged, as in systemic infection, cancer or autoimmune disease, it can intensify sickness and contribute to depressive symptoms in vulnerable people (When activation of the peripheral immune system continues unabated, such as during systemic infections, cancer or autoimmune diseases, the ensuing immune signalling to the brain can lead to an exacerbation of sickness and the development of symptoms of depression in vulnerable individuals.).

Sensory neurons and immune cells at barrier tissues

Barrier tissues such as the skin, lungs and gut are where nerve endings and immune cells meet most densely. Sensory neurons there were once viewed as passive detectors, but evidence now supports a model in which they act as active immune sentinels that directly affect tissue immunity in the skin, lungs, and gastrointestinal tract. Nociceptors carry receptors for pathogen signals and for cytokines, so they can sense infection and tissue damage directly (Nociceptors and pruriceptors express various immune-sensing receptors, including Toll-like receptors, cytokine receptors, and alarmin sensors). The communication runs in both directions: immune cells alter neuronal excitability through cytokines and alarmins, while activated neurons release neuropeptides that reshape immune cell behavior (Immune cells also alter neuronal excitability through cytokines and alarmins, forming bidirectional circuits at barrier tissues.). The main neuropeptides include substance P, calcitonin gene-related peptide and vasoactive intestinal peptide, each acting through distinct receptor pathways (neurons secrete substance P, calcitonin gene-related peptide (CGRP), and vasoactive intestinal peptide (VIP), each engaging distinct receptor pathways and immune programs).

In the skin, substance P activates receptors that drive calcium signaling and mast cell degranulation (Substance P activates G protein–coupled receptors to induce Ca2+ signaling and mast cell degranulation.). Skin neuron subsets can push immune responses in opposite directions: SP+ peptidergic neurons and glutamate+ non-peptidergic neurons antagonistically regulate mast cell reactivity in the skin. In the lung, the airways are supplied mainly by vagal sensory neurons rather than by the dorsal root ganglia that serve most of the skin (recent tracer studies have demonstrated sparse innervation with an overwhelming majority of lung sensory neuron innervation provided by the vagal ganglia). Vagal neurons restrain allergic inflammation: CGRPβ suppresses pathogenic group 2 innate lymphoid cell (ILC2) responses and allergic lung inflammation. In the gut, TRPV1+ nociceptors integrate luminal and inflammatory stimuli to regulate mucosal immune homeostasis. These interactions are not simple relays; the outcome depends on context and timing, and inflammation may intensify or subside accordingly (Inflammation may intensify or subside, depending on context and timing. In this manner, neuroimmune communication is not a linear relay.).

Evolution across animals

Current reviews place the origin of nervous systems in cnidarians or in a closely related ancestor. Cnidarians have simple nervous systems and it was probably within this group or a closely-related ancestor that nervous systems first evolved. Their basic layout is a diffuse network of cells, which in places becomes denser: the basic plan is that of a nerve net which, at some locations, has condensed to form nerve plexuses, or circular or longitudinal nerve tracts. In the sea anemone Nematostella vectensis, neurons are more concentrated at the oral and aboral ends of the body axis. Cnidarian cells also tend to be multifunctional: at the ultrastructural level, many of them show features of sensory, motor, interneuron and neurosecretory cells together, which has led some researchers to propose that such cells resemble the ancestors of the specialized neurons seen in higher animals.

The nerve net is not a settled cnidarian-only feature. Ctenophores (comb jellies) are also listed in common zoology textbooks as 'nerve net' animals. However, electron microscopy of a ctenophore showed that its subepithelial neurons share a continuous membrane and form a syncytium, which suggests fundamental differences of nerve net architectures between ctenophores and cnidarians/bilaterians. Whether the nerve net is one ancestral design or arose more than once remains unresolved. At the genetic level, some developmental machinery is shared: in cnidarians, Wnt and bone morphogenetic protein (BMP) signalling, and Sox genes play a crucial part in the differentiation of neurons, and much of this network is conserved with bilaterians.

Centralization and cephalization in bilaterians

Bilaterian animals, with a left and right body plan, are the lineage in which a centralized nervous system with a head region developed. The majority view is that a CNS evolved just once – in the ancestral bilaterian. Even so, the picture is contested. Whether the ancestral bilaterian already had a central nervous system, or only a nerve net, is highly controversial, and the answer determines whether protostomes and deuterostomes inherited a shared CNS or built their own independently. A competing proposal holds that the chordate CNS arose from the ectodermal nerve net of a hemichordate-like ancestral deuterostome. Another hypothesis suggests that the bilaterian CNS may have grown out of the "ring ganglia" found in many cnidarian groups. Developmental studies add that the anterior–posterior organizing centers of the vertebrate brain have homologous gene networks in the ancestral deuterostome, although some of these centers were lost from invertebrate chordate nervous systems.

Mollusks show the same trend in a different form. In the simplest mollusks, ganglia are poorly developed, and only a simple nerve ring surrounds the esophagus with small cerebral ganglia on each side. Lengthwise nerve cords linked by transverse commissures form a ladder-like arrangement in many species. Centralization appears to have happened in parallel: one analysis of mollusks concluded that their nervous-system evolution involved at least four independent events.

Convergence in arthropods, cephalopods and vertebrates

Coleoid cephalopods (octopuses, squids and cuttlefish) are the clearest case of convergent complexity. They have the largest nervous systems among the invertebrates, and the octopus brain includes more than 30 differentiated lobes in Octopus vulgaris. Comparative anatomy describes cephalopod neural systems as resembling the cerebral cortex, thalamus, basal ganglia, cerebellum, hypothalamus and brainstem of vertebrates, though the degree of true equivalence is debated. Some researchers go further and argue that the octopus arm-nerve cord is functionally and structurally analogous to the vertebrate spinal cord, rather than a simple peripheral structure.

Vision is the best-known example. Camera-type eyes evolved independently, since an image-forming eye was absent in the last common ancestor of molluscs and chordates. The designs differ: in cephalopods, light reaches photoreceptors directly; by contrast, the vertebrate retina is 'inverted'. The cephalopod retina is also simpler, with only a single type of rhabdomeric photoreceptor oriented towards incoming light, and supporting cells but no interneurons. At the genomic level, the octopus shows massive expansions in two gene families previously thought to be uniquely enlarged in vertebrates: the protocadherins, which regulate neuronal development, and the C2H2 superfamily of zinc-finger transcription factors. Convergence also extends to function: cephalopods have converged with insects and vertebrates on the organization of circuitry for learning and memory. Even so, the solutions differ at the molecular level; octopus visual systems use dopamine receptors that function differently from those in vertebrates.

Shared genetic programs across species

Comparative genetics has shown that similar developmental genes are reused across distant lineages. The transcription factor Pax6 is the most studied example. Its mutations disrupt eye development in both mammals and insects, and genuine Pax 6 genes have now been isolated from a wide range of phyla, including mammals, fish, squid, nematodes and planarians. Because an almost universal use of Pax6 in eye formation is seen across most species, it has been taken as evidence for a single evolutionary origin of eyes, an idea sometimes linked to a single-celled pigmented photoreceptor found in the larvae of the jellyfish Tripedalia. Pax genes are also present in cnidarians, which are among the most basal animals with eyes.

These conclusions have limits. Shared sequence does not guarantee shared function: as one review puts it, sequence conservation of the protein-coding region does not necessarily imply that the function in eye morphogenesis is also conserved in evolution. Comparisons of this kind therefore depend on functional tests in each species, not on gene sequences alone.

Research history

The study of the nervous system began in antiquity. Galen of Pergamon, working mostly with animals in the second century CE, linked the brain and spinal cord to movement and sensation by cutting nerves and observing the resulting paralysis. For centuries afterward, anatomists refined this picture, and the debate over how nerve tissue is organized became a central question.

The first major theoretical dispute was between the reticular theory and the neuron doctrine. Camillo Golgi's silver staining method, introduced in 1873, showed individual nerve cells in unprecedented detail, yet Golgi himself argued that their processes fused into a continuous network. Golgi's "reticular" theory proposed that the nervous system could form a syncytium of connected nerve fibers, through which the nerve impulse could spread across a diffuse network. Santiago Ramón y Cajal used the same stain from 1887 onward and reached the opposite conclusion. In 1888 he reported that nerve cells in the bird brain were not continuous with one another. The term "neuron" was popularized by Wilhelm Waldeyer in 1891, and Cajal's position became the neuron doctrine, which held that the nervous system is made of discrete cells. Charles Sherrington later coined the word "synapse" for the junction between successive neurons. Golgi and Cajal shared the 1906 Nobel Prize in Physiology or Medicine, but the controversy between them continued. The doctrine was confirmed more decisively only after electron microscopy became available, since the neuron doctrine could not be formally proven until the electron microscope was applied to the nervous system. Golgi's concern with how a distributed system produces integrated activity remained relevant, and Cajal's last publication, in 1934, was a review of the evidence against the reticular view.

The next phase concerned how signals move along a single cell. Early twentieth-century work by Julius Bernstein proposed that the nerve impulse was a momentary breakdown of the membrane. Alan Hodgkin and Andrew Huxley challenged this view using the giant axon of the squid. The first recording of an action potential from a squid giant axon with intracellular electrodes was made in 1939. Their measurements showed that the potential could overshoot zero, which meant the membrane was not simply collapsing. Working with voltage-clamp methods, they devised a model in 1952 to quantify the permeability changes underlying the action potential. The model described how sodium and potassium currents, each governed by voltage- and time-dependent gating variables, produce the spike. Hodgkin and Huxley shared the 1963 Nobel Prize with John Eccles, who had worked on synaptic transmission, and the award recognized these contributions as the conceptual foundation for studying excitable cell signaling. Bernard Katz, who studied transmitter release at the neuromuscular junction, received a Nobel Prize for related work. For more on the ionic mechanism itself, see action potential and ion channels.

The molecular era began when the proteins that carry these currents could be isolated and sequenced. From that time, understanding of ion channels entered the molecular era. In 1982 the group led by Shosaku Numa cloned the nicotinic acetylcholine receptor channel from the electric ray Torpedo californica, and in 1984 it cloned the voltage-gated sodium channel from the electric eel Electrophorus electricus. The same group later isolated whole cDNA transcripts and expressed functional channels in Xenopus oocytes in 1986. Subsequent work revealed that the voltage-gated sodium channel gene family comprises nine homologous members. The same laboratory extended this approach to calcium channels. Cloning revealed that ligand-gated channels such as the acetylcholine receptor belong to a family that can open within milliseconds of transmitter binding; these channels undergo a fast conformational change on ligand binding to open an integral pore within a time frame of 1–100 ms. These molecular results connected the receptor and channel families to the synaptic and electrical physiology described earlier, and they became the basis for later work on synaptic proteins and the genetic causes of neurological disease.

Research methods and imaging

Methods for studying the nervous system range from recording the current through a single ion channel to mapping every synapse in a circuit. Each method answers a different question, and most published findings depend on combining several of them.

Electrophysiological recording

The patch clamp technique was developed by Erwin Neher and Bert Sakmann in the mid-1970s, and their first use was to measure ion channel currents at the neuromuscular junction in denervated frog muscle fibers. The method works by sealing the tip of a fine glass micropipette onto a small patch of membrane, which lets the current through that patch be read directly. Several recording configurations let investigators measure macroscopic currents of an entire cell or single-channel currents in a microscopic membrane patch. The work was later extended to slices of brain tissue and, in some studies, to living animals. Neher and Sakmann received the 1991 Nobel Prize in Physiology or Medicine for it for their work in developing the patch clamp technique.

Extracellular recording takes a different approach. An electrode placed outside a neuron, without breaking into the cell, detects the action potentials of one or several nearby cells. Because the electrode stays outside the cell, it is well suited to recording from awake animals over long periods. Multi-electrode arrays and tetrodes can record dozens to hundreds of neurons at once, and spike sorting algorithms then separate the signals from different cells.

Neuroimaging: fMRI and PET

Functional magnetic resonance imaging (fMRI) infers neural activity from blood flow rather than measuring it directly. From the BOLD signal, researchers infer neuronal activation in that brain region, and the BOLD signal reflects changes in oxygenated hemoglobin. Its temporal resolution is on the order of seconds and its spatial resolution is around a millimetre. BOLD-fMRI is a semi-quantitative index of neuronal function that cannot be compared across brain regions, subjects, or imaging sites. A 2025 study also reported that roughly 40% of voxels showing significant BOLD changes during tasks had oxygen metabolism moving in the opposite direction, which challenges the simple interpretation of the signal.

Positron emission tomography (PET) works through a radioactive tracer rather than magnetic fields. The most common tracer, FDG, is a glucose analogue, so FDG-PET is a fully quantitative index of neuronal activity that captures cerebral glucose uptake that is primarily localised to the synapses. Unlike fMRI, PET can map the distribution of particular molecules in the brain, including receptors and neurotransmitter-related targets, which makes it valuable for neurotransmitter and pharmacological studies. Its disadvantages include radiation exposure and the need for radioactive materials. Because PET has a good signal-to-noise ratio in a single scan, it is useful for rare events that cannot be repeated many times, whereas fMRI generally needs many repetitions to separate a weak signal from noise.

Genetic tools: optogenetics and calcium imaging

Optogenetics introduces light-sensitive proteins into defined cell types so that researchers can switch those cells on or off with light. The key protein, channelrhodopsin-2, was first identified in the green alga Chlamydomonas reinhardtii. In 2005, Karl Deisseroth, Edward Boyden and Feng Zhang showed that expressing it in mouse hippocampal neurons made the cells fire action potentials when exposed to blue light. Because the response takes place within milliseconds, the technique can test whether specific spike patterns in specific cells cause a particular behavior. Deisseroth, Peter Hegemann and Georg Nagel shared the Nobel Prize for this work "for their discoveries concerning light-gated ion channels and optogenetics".

Calcium imaging takes the opposite approach by reporting activity rather than controlling it. Genetically encoded calcium indicators fluoresce when intracellular calcium concentration rises, and calcium enters the cell during spiking. Imaging these indicators, often with two-photon microscopy, therefore allows many neurons to be monitored at once in living tissue, though the signal is slower and less precise in time than electrical recording.

Connectomics

Connectomics aims to map the wiring of neural circuits at the level of individual synapses. The approach relies on serial electron microscopy, in which thin slices of tissue are imaged and reconstructed into three-dimensional maps of cells and their connections. Projects have produced complete connectomes for small animals such as the fruit fly larva and adult fly, and a cubic millimetre of mouse visual cortex has been reconstructed. A connectome records structure, not activity, and it generally does not reveal neurotransmitter identity or the strength of each connection. Combining connectomes with functional imaging and optogenetic experiments is therefore the main strategy for understanding how wiring produces behavior, a question that remains open.

Neurological and psychiatric disorders

Stroke

Stroke is a sudden loss of blood supply to part of the brain. It has two main forms with different mechanisms. Ischemic stroke accounts for about 83 percent of all cases of stroke. In these cases a blocked artery starves neurons of oxygen and glucose, and the affected tissue begins to die within minutes. Hemorrhagic stroke makes up roughly 17 percent of cases and results from a weakened vessel that ruptures and bleeds into the surrounding brain. The blood then compresses nearby tissue. Hemorrhages are classed as intracerebral hemorrhage and subarachnoid hemorrhage. Typical triggers include chronic high blood pressure, aneurysms, and arteriovenous malformations, which are a cluster of abnormally formed blood vessels. The two forms also interact: an ischemic stroke can transform into a hemorrhagic stroke as a result of reperfusion.

The proportions of each type have been debated. Epidemiological studies from the 1970s and 1980s reported the percentage of ischemic stroke as 73% to 86%, with hemorrhagic stroke as only 8% to 18%, and the rest were often unclassified because CT scanning or autopsy was not performed. A single-institution registry of 757 patients later found a much larger share of hemorrhages, 41.9% hemorrhagic and 58.1% ischemic. The authors suggested that wider CT availability and greater use of antiplatelet drugs and warfarin might explain the difference.

Alzheimer and Parkinson disease

Neurodegenerative diseases are defined largely by the accumulation of abnormal proteins. Amyloid beta (Aβ)-rich extracellular protein deposits and neuronal inclusions rich in aggregated tau are the neuropathological hallmarks of Alzheimer's disease. In Alzheimer disease, amyloid plaques form outside cells while tau accumulates inside neurons in a hyperphosphorylated form. Parkinson disease is characterized by inclusion bodies, termed Lewy bodies (LBs) and Lewy neurites (LNs), consisting of damaged organelles, lipids and aggregated alpha-synuclein (aSyn). The loss of dopaminergic neurons in the midbrain produces the characteristic motor symptoms, while cognitive decline can appear later.

These pathologies frequently overlap. In approximately 50% of AD cases, aSyn pathology in the brain is common at autopsy and vice versa, in around 50% of PD cases, tau pathology is common. In Lewy body disorders, Alzheimer-type changes are also common, and the combination is associated with a shorter interval between onset of motor symptoms and development of dementia and a shorter life span. A 2025 cohort study that used cerebrospinal fluid assays and PET imaging examined 284 Aβ-positive and 308 Aβ-negative subjects and reported that alpha-synuclein positivity was linked to faster amyloid-related tau accumulation. Such observations are correlational and do not establish mechanism, but they have shaped interest in combined therapeutic targets. Imaging of amyloid and tau with positron emission tomography and fluid biomarkers now supports diagnosis in research settings.

Demyelinating diseases and peripheral neuropathies

Multiple sclerosis is a chronic immune-mediated disorder of the central nervous system characterized by demyelination, axonal loss, and neuroinflammation, culminating in progressive neurological disability. The inflammatory process involves a breakdown of the blood–brain barrier: proinflammatory cytokines and matrix metalloproteinases degrade the BBB, allowing autoreactive lymphocytes to attack myelin sheaths. Within the lesions, chronic microglial activation leads to the release of proinflammatory cytokines and reactive oxygen and nitrogen species, which contribute to demyelination, tissue damage and neurodegeneration. Antibodies and complement activation may also contribute. The disease usually begins in young adults; globally, it affects about 2 million people, most of them in their 20 s and 40 s. Its cause remains unknown, and the mechanisms by which lesions form are still debated. Current disease-modifying drugs act mainly on the immune system, and some work by impeding infiltration of activated T cells into the CNS.

Peripheral neuropathies involve damage to nerves outside the brain and spinal cord, affecting the axons or the myelin made by Schwann cells. Common causes include diabetes, which is the most frequent, as well as certain chemotherapy drugs, inherited disorders such as Charcot–Marie–Tooth disease, and autoimmune conditions such as Guillain–Barré syndrome. Symptoms typically begin in the feet and hands and include numbness, tingling, pain, and weakness.

Psychiatric disorders

Depression, schizophrenia, and anxiety disorders are diagnosed from clinical criteria rather than from any single lesion or laboratory test, and they are studied through a mixture of neurobiology, genetics, and clinical research. Early theories linked depression to deficits in monoamines such as serotonin and noelepinephrine. Drugs that raise these transmitter levels act within hours, yet their antidepressant effects take weeks, which suggests that downstream changes in neurotransmitter signalling and neuroplasticity are also important.

Schizophrenia research has centred on dopamine and glutamate. Antipsychotic drugs that block dopamine D2 receptors reduce positive symptoms such as hallucinations and delusions, and the hypothesis that reduced NMDA receptor function contributes to the disorder was supported by the psychotomimetic effects of NMDA antagonists. Twin studies indicate high heritability, and genome-wide association studies have identified many common variants of small effect, so the disorder is polygenic. Anxiety disorders involve circuits linking the amygdala with prefrontal regions, and the GABA system is the main target of anxiolytic drugs such as benzodiazepines. Across all three conditions, the link between molecular findings and clinical diagnosis remains incomplete.

Relationship to artificial neural networks

Early formal models

The idea that nervous tissue could be described in the language of logic came before digital computers. In 1943 the neuroscientist Warren McCulloch and the logician Walter Pitts published "A Logical Calculus of the Ideas Immanent in Nervous Activity." The paper introduced a highly simplified unit, now called the McCulloch–Pitts neuron, which sums weighted inputs and fires when the sum reaches a threshold. Excitatory and inhibitory connections were represented by positive and negative weights. The authors showed that networks of these units could realize logical operations such as AND, OR and NOT, and that such networks could in principle carry out any logical function. The model ignored synaptic timing, learning and the nonlinear behaviour of real dendrites, but it made the case that neural activity could be treated as computation. Its influence reached later work on computing, including von Neumann's 1956 treatment of reliable computation built from unreliable elements.

Perceptrons and connectionism

The next step was to make the weights learnable. Frank Rosenblatt's perceptron (1958) adjusted its connection weights whenever its output was wrong, which made it the first widely studied network that learned from examples. Its limits became a central debate after Marvin Minsky and Seymour Papert (1969) showed that a single layer of such units cannot compute functions such as exclusive-OR. Interest revived in the 1980s with connectionist models, in which knowledge is stored in the pattern of connection strengths rather than in explicit symbols. The key technical advance was backpropagation for training multi-layer networks, popularized in 1986 by David Rumelhart, Geoffrey Hinton and Ronald Williams. In this framework, learning means changing connection weights, a principle loosely related to the Hebbian idea that co-active neurons strengthen their synapses, though backpropagation uses a global error signal rather than local co-activity.

Points of resemblance

Biological and artificial networks share several organizing ideas: many simple units, graded connection strengths, distributed representations, and computation that emerges from connectivity rather than from any single element. Hierarchical processing also has parallels. Kunihiko Fukushima's neocognitron (1980), which drew on Hubel and Wiesel's studies of visual cortex, anticipated the layered, locally connected architectures that later became convolutional networks. Deep learning extends this by stacking many trainable layers and fitting them to very large datasets. Some systems trained this way develop internal representations that resemble neural activity in certain sensory areas, which is one reason researchers use them as models of perception. Such similarities are functional rather than literal, however, and the term "neuron" for an artificial unit is largely a metaphor.

Differences in learning rules

The most serious disagreement concerns how learning happens. Backpropagation requires each weight to be updated using error signals computed from downstream layers, and it assumes that the backward pass uses the same weights as the forward pass. Biological wiring does not appear to supply such symmetric feedback paths, a difficulty known as the weight transport problem. Backpropagation also separates a forward inference phase from a backward learning phase, whereas neural circuits in the brain operate continuously. Researchers have proposed alternatives that use local signals or random fixed feedback, such as feedback alignment, predictive coding and sign-symmetry rules, but none has yet matched backpropagation's performance at scale. Learning in the brain is also data-efficient in ways that current systems are not: people and animals often generalize from few examples, while many deep networks require very large labeled datasets.

Energy use and architecture

Energy is a second major difference. The adult human brain runs on about 20 watts and contains roughly 86 to 89 billion neurons and on the order of 100 trillion synapses. A 2020-era comparison in a physics review estimated that a brain's energy budget could power it for years, whereas the same energy would allow only about 80 hours of training for the language model BERT on 64 GPUs. Large artificial systems draw megawatts during training and inference. The architectures also differ: most deep learning hardware separates memory from processing and runs on synchronous, dense numerical operations, while neurons are massively parallel and communicate mainly through discrete spikes whose timing carries information. Spiking neural networks and neuromorphic hardware are attempts to close this gap, and they remain an active area of research.

Limits of the analogy

Brain–network comparisons are useful but limited. Real neurons include dendritic computation, diverse cell types, neuromodulation, and interactions with glial cells and blood supply, none of which appear in standard artificial units. Brains also grow and reorganize throughout life, whereas most artificial networks have fixed architectures chosen by engineers. Successes in machine learning therefore do not by themselves show how the brain computes, and neuroscientific findings do not automatically transfer to engineering. Each field uses the other as a source of hypotheses, with the understanding that resemblance in function does not imply identity in mechanism.

Open questions

The neural basis of consciousness

Neuroscientists can describe how individual neurons and circuits respond to stimuli, but there is still no accepted account of why some neural activity is accompanied by subjective experience and some is not. Researchers have proposed candidate neural correlates of conscious perception, often involving large-scale activity across cortical networks rather than one region. Such proposals, however, mainly show which activity tends to co-occur with awareness. They do not explain why any physical process should be felt at all. This gap between physical description and subjective experience is why the problem is studied jointly in neuroscience, cognitive-science and philosophy-of-mind, and why its experimental tests remain contested.

The neural code

Neurons communicate largely through sequences of action potentials, all of which have the same amplitude and shape, so the information must lie in the pattern of those events. Candidate schemes include the firing rate of single cells, the precise timing of spikes relative to one another or to ongoing oscillations, the coordinated activity of populations of neurons, and the strength and arrangement of synaptic connections. These schemes are not mutually exclusive, and the brain may use different codes in different areas or for different tasks. Distinguishing among them requires recording from many neurons at once and relating their activity to behavior, which is technically demanding and is complicated by the variability of responses even to repeated identical stimuli.

Regeneration in the central nervous system

Axons in the adult mammalian central nervous system (CNS) regrow poorly after injury, which is a principal reason that spinal cord injury, traumatic brain injury and stroke often leave lasting deficits. The problem was recognized early: in the early twentieth century, Ramón y Cajal argued that adult CNS neurons could not regenerate. Modern work attributes the failure to several factors. Mature neurons have a diminished intrinsic capacity to switch on growth programs. Myelin contains inhibitory proteins, such as Nogo and myelin-associated glycoprotein, that cause growth cones to collapse. After injury, reactive astrocytes contribute to a glial scar that acts as both a physical and chemical barrier. Aging worsens several of these factors. By contrast, adult peripheral nerves can regrow over long distances and often recover function. Some exceptions have been reported, including serotonergic neurons of the rostral raphe nuclei, which can regenerate axons into the forebrain without experimental intervention, and transcriptomic evidence that injured corticospinal neurons in mice partially revert to an embryonic-like state. Whether these observations can be translated into effective treatments for humans remains unresolved.

Connectomes and individual variation

Connectomics aims to map the neurons and synapses of a nervous system at electron-microscope resolution. In 2024, the FlyWire consortium published a synapse-level wiring diagram of the entire adult female fruit fly brain, containing roughly 140,000 neurons and more than 50 million synaptic connections. In the same period, the MICrONS project reconstructed about one cubic millimetre of mouse visual cortex. It combined electron microscopy with calcium imaging of tens of thousands of neurons and detected more than 500 million synapses, although the fully reconstructed volume still represents only a small fraction of a mammalian brain. These datasets raise two open questions. The first is how static anatomical maps relate to dynamic function, since a connectome captures structure at one moment while living circuits change with activity, state and learning. The second concerns variation between individuals. Even within a single species, people differ in the detailed wiring of their brains, and it is still unclear how much of this variation reflects experience and how much reflects genetics. Answering this requires large, comparable datasets from many individuals, which are far more difficult to obtain than the single-animal maps produced so far.

References

1. NEURON STRUCTURE AND CLASSIFICATIONcontent.byui.edu
2. Anatomy of Nerves - Classification - Neurones - TeachMeAnatomyteachmeanatomy.info
3. Basic Neural Units of the Brain: Neurons, Synapses and Action Potentialarxiv.org
4. The Neuron — BrainFacts.orgbrainfacts.org
5. Glial cells: Types and functionmedicalnewstoday.com
6. Glial Cells: Types, Function & Role in the Brainintuitionlabs.ai
7. Types of glia - Queensland Brain Institute - University of Queenslandqbi.uq.edu.au
8. 35.3: Neurons and Glial Cells - Glia - Biology LibreTextsbio.libretexts.org
9. Gliaen.wikipedia.org
10. Unipolar Neuron - an overviewsciencedirect.com
11. Multipolar Neurons Are Motor and Interneurons, Not Sensory - ScienceInsightsscienceinsights.org
12. Video: Nervous Tissue: Neuron Typesjove.com
13. Types of neuronskenhub.com
14. Resting Membrane Potentials - Free Sketchy MCAT Lessonsketchy.com
15. The Membrane at Restopentext.uoregon.edu
16. 1.2 Membrane Potential: Its components and maintenanceecampusontario.pressbooks.pub
17. Action Potentialsopentext.uoregon.edu
18. Action Potentialsopenbooks.lib.msu.edu
19. Refractory Periods - Action Potential experimentsilearn.med.monash.edu.au
20. Action Potentialsopenbooks.lib.msu.edu
21. Saltatory Conduction: Jumping to New Conclusions: Current Biologycell.com
22. Saltatory conductionen.wikipedia.org
23. Myelin and saltatory conductionarxiv.org
24. Refractory period (physiology)en.wikipedia.org
25. Nodes of Ranvier: structure and functionkenhub.com
26. Saltatory Conductionib.bioninja.com.au
27. 12.1 Structure and Function of the Nervous Systemopen.oregonstate.education
28. Lab 8: Nervous system Iiastate.pressbooks.pub
29. Overview of the Nervous System (Section 2, Chapter 1) Neuroscience Online: An Electronic Textbook for the Neurosciencesnba.uth.tmc.edu
30. Major, Cranial & Spinal Nervesstudy.com
31. Peripheral nervous system: Anatomy, divisions, functionskenhub.com
32. Nerves - Brain, Spinal Cord, and Nerve Disorders - MSD Manual Consumer Versionmsdmanuals.com
33. Cranial Nerve - an overviewsciencedirect.com
34. Cranial Nerves Anatomyteachmeanatomy.info
35. Are there 31 pairs of cranial nerves and 12 pairs of class 11 biology CBSEvedantu.com
36. Peripheral Nervous System - an overviewsciencedirect.com
37. Grey matter of the spinal cordkenhub.com
38. Gray Matter - an overviewsciencedirect.com
39. Gray vs White Matter: Brain Structure & Functionstechnologynetworks.com
40. 6.7: Gray vs. White Matter- Headquarters and Highways in the CNS - Medicine LibreTextsmed.libretexts.org
41. Anatomy of the ascending spinal cord pathwaysosmosis.org
42. Spinal cord: Ascending and descending tractskenhub.com
43. Spinal Reflexesopentext.uoregon.edu
44. Withdrawal Reflex - an overviewsciencedirect.com
45. Reflexesmhcc.pressbooks.pub
46. Reflexes - Free Sketchy MCAT Lessonsketchy.com
47. Central pattern generation of locomotion: a review of the evidence - PubMedpubmed.ncbi.nlm.nih.gov
48. Prominent Role of the Spinal Central Pattern Generator in the Recovery of Locomotion after Partial Spinal Cord Injuries - PMCpmc.ncbi.nlm.nih.gov
49. Spinal cord pattern generators for locomotion - ScienceDirectsciencedirect.com
50. Central Pattern Generators in Spinal Cord Injury: Mechanisms, Modulation, and Therapeutic Strategies for Motor Recovery - PMCpmc.ncbi.nlm.nih.gov
51. Sensory Transduction: A Common Blue Printgavinpublishers.com
52. Sensory Systems (Chapter 41) Flashcardsquizlet.com
53. Sensory transductionuw.pressbooks.pub
54. Receptor Potential - an overviewsciencedirect.com
55. Motor Responsespressbooks-dev.oer.hawaii.edu
56. 14.3 Motor Responses - Anatomy and Physiology 2eopenstax.org
57. Pyramidal tracts: Corticospinal and corticonuclear tractskenhub.com
58. 13.5: Somatic Motor Responses - Medicine LibreTextsmed.libretexts.org
59. Molecular correlates of muscle spindle and Golgi tendon organ afferentsncbi.nlm.nih.gov
60. Muscle spindles and Golgi tendon organsderangedphysiology.com
61. (PDF) Muscle spindles, Golgi tendon organs, and the neural control of skeletal muscleacademia.edu
62. Computing muscle mechanical state variables from combined proprioceptive sensory feedbackncbi.nlm.nih.gov
63. 13.1 Sensory Receptorsopen.oregonstate.education
64. Sensory Receptorsuw.pressbooks.pub
65. Sensory receptorsiu.pressbooks.pub
66. Sensory Receptor - an overviewsciencedirect.com
67. Muscle spindles and golgi tendon organs: Videoosmosis.org
68. Corticospinal Tract and Other Motor Pathwaysbrainaacn.org
69. Unit IV Motor Systemspocketdentistry.com
70. 17.19: Autonomic Nervous System - Biology LibreTextsbio.libretexts.org
71. Autonomic Dysfunction and Management after Spinal Cord Injury: A Narrative Reviewncbi.nlm.nih.gov
72. Autonomic nervous system - Knowledge @ AMBOSSamboss.com
73. Overview of the Autonomic Nervous System - Brain, Spinal Cord, and Nerve Disorders - Merck Manual Consumer Versionmerckmanuals.com
74. Sympathetic Vs Parasympathetic Nervous Systemsimplypsychology.org
75. Autonomic Nervous System: Sympathetic, Parasympathetic, & Entericaccessmedicine.mhmedical.com
76. Enteric nervous system: Video, Causes, Symptomsosmosis.org
77. Enteric Nervous System - an overviewsciencedirect.com
78. Autonomic modulation using peripheral nerve field stimulationimage-ppubs.uspto.gov
79. The enteric nervous system in gastrointestinal disease etiologyncbi.nlm.nih.gov
80. Cerebrospinal Fluid Secretion by the Choroid Plexusjournals.physiology.org
81. A computational model of cerebrospinal fluid production and reabsorption driven by Starling forcesncbi.nlm.nih.gov
82. Cerebrospinal fluid flow: Anatomy and functionskenhub.com
83. Cerebrospinal Fluidjove.com
84. Formation, circulation and function of cerebrospinal fluidderangedphysiology.com
85. Cerebrospinal Fluid Flow - an overviewsciencedirect.com
86. Pia materen.wikipedia.org
87. Meninges: What They Are & Functionmy.clevelandclinic.org
88. Meninges: Dura, arachnoid, pia, meningeal spaceskenhub.com
89. Anatomy, Back, Spinal Meninges - StatPearls - NCBI Bookshelfncbi.nlm.nih.gov
90. The blood–brain barrier: an overview: Structure, regulation, and clinical implications - ScienceDirectsciencedirect.com
91. Tight junctions at the blood brain barrier: physiological architecture and disease-associated dysregulationfluidsbarrierscns.biomedcentral.com
92. Astrocytic endfeet re-cover blood vessels after removal by laser ablationnature.com
93. The blood–brain barrier in systemic infection and inflammationnature.com
94. Blood Brain Barrier - an overviewsciencedirect.com
95. Cx43 carboxyl terminal domain determines AQP4 and Cx30 endfoot organization and blood brain barrier permeabilityncbi.nlm.nih.gov
96. Formation of the neural tubeembryology.ch
97. Hingepoint emergence in mammalian spinal neurulationbiorxiv.org
98. Frontiersfrontiersin.org
99. Specification and formation of the neural crest: Perspectives on lineage segregation - PMCpmc.ncbi.nlm.nih.gov
100. Neurogenesis From Neural Crest Cells: Molecular Mechanisms in the Formation of Cranial Nerves and Gangliancbi.nlm.nih.gov
101. Mammalian multipotent neural stem cellsimage-ppubs.uspto.gov
102. Gradient-reading and mechano-effector machinery for netrin-1-induced axon guidancencbi.nlm.nih.gov
103. Signaling from Axon Guidance Receptors Greg J. Bashaw1 and Ru¨ diger Klein2med.upenn.edu
104. Flavin monooxygenases regulate Caenorhabditis elegans axon guidance and growth cone protrusion with UNC-6/Netrin signaling and Rac GTPasesncbi.nlm.nih.gov
105. Graphic Processing Unit Simulation of Axon Growth and Guidance through Cue Diffusion on Massively Parallel Processorsarxiv.org
106. Ephrin-A5 potentiates netrin-1 axon guidance by enhancing Neogenin availabilityncbi.nlm.nih.gov
107. Molecular signaling directing neural plate border formationijdb.ehu.eus
108. Neurulation - an overviewsciencedirect.com
109. Neural Tube - an overviewsciencedirect.com
110. 27.4B: Neurulation - Medicine LibreTextsmed.libretexts.org
111. RHO-1 and the Rho GEF RHGF-1 interact with UNC-6/Netrin signaling to regulate growth cone protrusion and microtubule organization in Caenorhabditis elegansncbi.nlm.nih.gov
112. Synergistic integration of Netrin and ephrin axon guidance signals by spinal motor neuronsncbi.nlm.nih.gov
113. Long-term potentiation of NMDA receptor-mediated synaptic transmission in the hippocampusnature.com
114. Extrasynaptic NMDA receptor dependent long-term potentiation of hippocampal CA1 pyramidal neuronsncbi.nlm.nih.gov
115. Long-Term Potentiation in the CA1 Hippocampus Induced by NR2A Subunit-Containing NMDA Glutamate Receptors Is Mediated by Ras-GRF2/Erk Map Kinase Signalingncbi.nlm.nih.gov
116. Activation of the hypothalamic-pituitary-adrenal stress axis induces cellular oxidative stressncbi.nlm.nih.gov
117. Characterizing Dynamic Interactions between Ultradian Glucocorticoid Rhythmicity and Acute Stress Using the Phase Response Curvencbi.nlm.nih.gov
118. The role of the hypothalamic-pituitary-adrenal axis in neuroendocrine responses to stress - PubMedpubmed.ncbi.nlm.nih.gov
119. Stress, hypothalamic-pituitary-adrenal axis, hypothalamic-pituitary-gonadal axis, and aggressionlink.springer.com
120. Neuroimmune Interactions: From the Brain to the Immune System and Vice Versajournals.physiology.org
121. Sickness Behavior: A Neuroimmune-Based Response to Infectious Diseaselink.springer.com
122. The mechanisms of immune-to-brain communication in inflammation as a drug target - PubMedpubmed.ncbi.nlm.nih.gov
123. Neuroimmune Signaling: Cytokines and the Central Nervous Systemlink.springer.com
124. Cytokine Actions on Behaviorlink.springer.com
125. From inflammation to sickness and depression: when the immune system subjugates the brainnature.com
126. Sensory Neuroimmunology: Bidirectional Neuro-Immune Circuits Governing Pain, Itch, Inflammation, and Host Defense at Barrier Surfaces - PMCpmc.ncbi.nlm.nih.gov
127. Sensory Neuroimmunology: Bidirectional Neuro-Immune Circuits Governing Pain, Itch, Inflammation, and Host Defense at Barrier Surfacesdoi.org
128. Sensory neurons: An integrated component of innate immunity: Immunitycell.com
129. Sensory neurons regulate stimulus-dependent humoral immunitybiorxiv.org
130. Sensory neurons promote immune homeostasis in the lung - ScienceDirectsciencedirect.com
131. The nervous systems of Cnidarianslink.springer.com
132. Neural nets - ScienceDirectsciencedirect.com
133. Syncytial nerve net in a ctenophore adds insights on the evolution of nervous systemsncbi.nlm.nih.gov
134. Evolution of eumetazoan nervous systems: insights from cnidariansroyalsocietypublishing.org
135. Evolution of bilaterian central nervous systems: a single origin?ncbi.nlm.nih.gov
136. Cephalopod Encephalizationlakeforest.edu
137. The octopus genome and the evolution of cephalopod neural and morphological noveltiesnature.com
138. (PDF) Cephalopod Brains: An Overview of Current Knowledge to Facilitate Comparison With Vertebratesresearchgate.net
139. A single-cell atlas of the bobtail squid visual and nervous system highlights molecular principles of convergent evolutionncbi.nlm.nih.gov
140. A single-cell atlas of bobtail squid visual and nervous system highlights molecular principles of convergent evolutionbiorxiv.org
141. Phylogenetic plasticity in the evolution of molluscan neural circuitssciencedirect.com
142. Octopus Brains Are Solving Problems the Same Way Vertebrates Domarinebio.org
143. Pax genes in eye development and evolution - ScienceDirectsciencedirect.com
144. Pax 6: mastering eye morphogenesis and eye evolution: Trends in Geneticscell.com
145. Full article: Eye evolution and the origin of Darwin's eye prototypetandfonline.com
146. Cnidarians and the evolutionary origin of the nervous systemresearchgate.net
147. Sequential actions of β-catenin and Bmp pattern the oral nerve net in Nematostella vectensisncbi.nlm.nih.gov
148. Cephalopods and Vertebrate Brainsfrontiersin.org
149. Genetic analysis of medaka fish illuminates conserved and divergent roles of Pax6 in vertebrate eye developmentncbi.nlm.nih.gov
150. Reticular Theory - an overviewsciencedirect.com
151. Neuron doctrineen.wikipedia.org
152. Golgi and Cajal: The neuron doctrine and the 100th anniversary of the 1906 Nobel Prize - ScienceDirectsciencedirect.com
153. Hodgkin and Huxley's modelling of the action potential - The Physiological Societyphysoc.org
154. Hodgkin and the action potential 1935-1952researchgate.net
155. Evolution of our understanding of the sodium channel fast inactivation: From Hodgkin and Huxley to the structural erarupress.org
156. The crucial decade that ion channels were proven to existlink.springer.com
157. Sodium channels - John N. Wood, Federico Iseppon, 2018doi.org
158. Molecular Neurobiology and Genetics: Investigation of Neural Function and Dysfunction: Neuroncell.com
159. Neurology through history: The advent of the neuron doctrinemedlink.com
160. Reticular theoryen.wikipedia.org
161. A L A N L L O Y D H O D G K I N The ionic basis of nervous conductionnobelprize.org
162. A short history of voltage-gated calcium channels - PMCpmc.ncbi.nlm.nih.gov
163. Patch Clamp Technique - an overviewsciencedirect.com
164. Electrical Activity of Individual Neurons: Patch-Clamp Techniqueslink.springer.com
165. System for electrophysiological measurementsimage-ppubs.uspto.gov
166. 4.5: Tools of Cognitive Neurosciencerotel.pressbooks.pub
167. Simultaneous BOLD-fMRI and constant infusion FDG-PET data of the resting human brainnature.com
168. Nobel Prize for Optogeneticsthe-scientist.com
169. Planar patch clamp electrodesimage-ppubs.uspto.gov
170. BOLD signal changes can oppose oxygen metabolism across the human cortexnature.com
171. Comparison of [15O] H2O Positron Emission Tomography and Functional Magnetic Resonance Imaging in Activation Studies - PMCpmc.ncbi.nlm.nih.gov
172. Light activatednature.com
173. Method for controlling brain cells with light wins Nobelscience.org
174. The History of Optogenetics Revisedthe-scientist.com
175. Combination therapy for ischemiaimage-ppubs.uspto.gov
176. Therapy for subarachnoid hemorrhage and ischemiaimage-ppubs.uspto.gov
177. Stroke registry: hemorrhagic vs ischemic strokes - ScienceDirectsciencedirect.com
178. Distinct tau and alpha-synuclein molecular signatures in Alzheimer’s disease with and without Lewy bodies and Parkinson’s disease with dementialink.springer.com
179. The Contribution of Tau, Amyloid-Beta and Alpha-Synuclein Pathology to Dementia in Lewy Body Disordersncbi.nlm.nih.gov
180. Alpha synuclein co-pathology is associated with accelerated amyloid-driven tau accumulation in Alzheimer’s diseasencbi.nlm.nih.gov
181. Multiple sclerosis: molecular pathogenesis and therapeutic interventionnature.com
182. Recent Progress in Multiple Sclerosis Treatment Using Immune Cells as Targetsncbi.nlm.nih.gov
183. Innate Immune Responses and Viral-Induced Neurologic Diseasencbi.nlm.nih.gov
184. Walter Pittsen.wikipedia.org
185. A Logical Calculus of the Ideas Immanent in Nervous Activityen.wikipedia.org
186. McCulloch & Pitts Publish the First Mathematical Model ...historyofinformation.com
187. Accidental Revolution: The Quiet Origin of Thinking Machines - Constitutional Discourseconstitutionaldiscourse.com
188. Physics for Neuromorphic Computingarxiv.org
189. Energy in Deep Learning - Stanford Universitylarge.stanford.edu
190. Energy-Based Dynamical Models for Neurocomputation, Learning, and Optimizationarxiv.org
191. Sign-Symmetry Learning Rules are Robust Fine-Tunersarxiv.org
192. Artificial Neural Nets Finally Yield Clues to How Brains Learnquantamagazine.org
193. Towards Biologically Plausible Deep Learningarxiv.org
194. Whole-brain annotation and multi-connectome cell typing of Drosophilanature.com
195. Complete wiring map of an adult fruit fly brainnih.gov
196. Cubic Millimeter — MICrONS Explorermicrons-explorer.org
197. Functional connectomics spanning multiple areas of mouse visual cortexnature.com
198. Inhibitory cells orchestrate neural activity in mouse brainthetransmitter.org
199. Regulation of axonal regeneration after mammalian spinal cord injurynature.com
200. Neuronal Redevelopment and the Regeneration of Neuromodulatory Axons in the Adult Mammalian Central Nervous Systemncbi.nlm.nih.gov
201. Moving CNS axon growth and regeneration research into human model systemsncbi.nlm.nih.gov
202. Updates and challenges of axon regeneration in the mammalian central nervous system - PMCpmc.ncbi.nlm.nih.gov