Skip to content
Robotic Brain Universe

Synapses and plasticity

How neurons signal across synapses, how connections strengthen and weaken to store memories, what attacks them, and the connectomes and chips that copy them.

Intermediate · about 9 min · updated 2026-10-02 · awaiting clinical review

See it in 3D:

Illustrative simulation excitatory inhibitory

Chemical transmission from Loewi to Katz; quantal release and the NMDA receptor's magnesium block; Hebb's postulate, long-term potentiation, spike-timing-dependent plasticity and memories switched on and off by LTP and LTD; short-term depression; synaptogenesis and pruning; myasthenia, synapse loss in Alzheimer's disease and anti-NMDA-receptor encephalitis; binomial release, magnesium block and the BCM rule; fly and human connectomes and memristor synapses.

Contents
  1. Where learning lives
  2. What a synapse is
  3. Why synapses change
  4. How a synapse works and learns
  5. When synapses are made and lost
  6. When synapses fail
  7. The mathematics of synapses
  8. Technology: connectomes and synthetic synapses
  9. Milestones
  10. Frontiers
  11. Check yourself

Where learning lives

Your neocortex holds roughly 0.15 quadrillion synapses, the junctions where one neuron passes a signal to the next. Change their strength and you change what the brain does: in an animal experiment, researchers inactivated a fear memory by weakening a set of synapses with light, then restored it by strengthening them again.[1,2]

A synapse is not a wire but a tiny chemical factory. Signals cross it as packets of neurotransmitter, are received by receptors with remarkable properties, and leave behind changes that can last for hours or longer.[3,4,5]

This reading covers how chemical transmission was discovered, how release and reception work, how synapses strengthen and weaken, how they are built and pruned, what diseases attack them, the mathematics of plasticity, and how connectomics and synapse-like electronics are pushing the frontier. For spikes themselves see Neurons and signals; for learning rules in AI see Neural networks.[6,7]

What a synapse is

Chemical transmission. In 1921 Loewi showed that the effect of the heart's nerves could be transferred by a humoral (chemical) agent. Dale and Loewi shared the 1936 Nobel Prize 'for their discoveries relating to chemical transmission of nerve impulses'.[8,9]

Quanta. Del Castillo and Katz analysed the end-plate potential at the nerve–muscle junction as the sum of quantal components: transmitter is released in packets of fixed size. Katz shared the 1970 Nobel Prize for discoveries concerning the transmitters in nerve terminals and the mechanism of their storage, release and inactivation.[3,10]

A receptor that detects coincidence. The NMDA receptor is a glutamate receptor whose current, unusually, increases as the cell is depolarised. Nowak and colleagues showed why: at rest, magnesium ions block its channel, and depolarisation relieves the block. The receptor therefore passes much more current when glutamate is present and the postsynaptic cell is already depolarised.[4,11]

Key numbers

Synapses in the human neocortex
about 0.15 quadrillion (1.5 × 10¹⁴)[1]
Synapses in one cubic millimetre of human temporal cortex
about 150 million[12]
Synapses in the complete wiring diagram of an adult fruit fly brain
5 × 10⁷, between 139,255 neurons[6]
Physiological extracellular magnesium, which blocks NMDA receptors at rest
about 1 mM[4]

Why synapses change

Hebb's postulate. In 1949 Hebb proposed that when one cell repeatedly helps to fire another, the connection between them grows stronger.[13]

Long-term potentiation. In 1973 Bliss and Lømo found that trains of stimulation of the perforant path in the hippocampus of anaesthetised rabbits produced a long-lasting potentiation of synaptic transmission: a lasting, activity-dependent change of the kind Hebb had predicted.[5]

Timing matters. In cultured hippocampal neurons, synapses strengthened when the receiving neuron fired within about 20 ms after the sending neuron and weakened when it fired within about 20 ms before, a rule called spike-timing-dependent plasticity.[14]

From synapses to memory. Nabavi and colleagues conditioned animals to associate a foot shock with optogenetic stimulation of auditory inputs to the amygdala. Optogenetic long-term depression of those inputs then inactivated the memory, and long-term potentiation reactivated it, supporting a causal link between these synaptic changes and memory.[2]

How a synapse works and learns

Short-term dynamics. At many synapses, successive spikes release less and less transmitter: the synapse depresses. Recording from pairs of neocortical pyramidal neurons, Tsodyks and Markram showed that the rate of depression, set by the probability of release, decides whether the postsynaptic cell mainly receives the presynaptic firing rate or the timing of synchronous spikes.[15]

Molecular memory. Kandel's work took a reductionist approach, studying elementary forms of learning in identified nerve cells to ask what changes in the brain when we learn and how the change is retained; it revealed a dialogue between genes and synapses. Carlsson, Greengard and Kandel shared the 2000 Nobel Prize for discoveries concerning signal transduction in the nervous system.[16,17]

Transmission and potentiation at a glutamate synapse (simplified)Presynaptic spikearrives at the terminalQuantal releasepackets of glutamate releasedwith some probabilityPostsynaptic responsedepolarises the postsynapticcellNMDA receptor unblockedonly if glutamate is bound andthe cell is depolarisedLong-term potentiationthe synapse strengthens forhours or longerMemory traceLTP and LTD can switch a memoryon and off
Transmission and potentiation at a glutamate synapse (simplified). A simplified sequence: quantal release, depolarisation that relieves the magnesium block of NMDA receptors, and long-lasting potentiation linked causally to memory in experiments.[2,3,4,5]
Text version of the diagram
  1. Presynaptic spike: arrives at the terminal. Leads to Quantal release.
  2. Quantal release: packets of glutamate released with some probability. Leads to Postsynaptic response; NMDA receptor unblocked.
  3. Postsynaptic response: depolarises the postsynaptic cell. Leads to NMDA receptor unblocked.
  4. NMDA receptor unblocked: only if glutamate is bound and the cell is depolarised. Leads to Long-term potentiation.
  5. Long-term potentiation: the synapse strengthens for hours or longer. Leads to Memory trace.
  6. Memory trace: LTP and LTD can switch a memory on and off.

When synapses are made and lost

Building and pruning. Synapse formation in the human cerebral cortex begins before birth. Synaptic density peaks near 3 months after birth in auditory cortex (Heschl's gyrus) but not until after 15 months in prefrontal cortex (middle frontal gyrus); a later phase of net elimination has ended by age 12 in auditory cortex but extends into mid-adolescence in prefrontal cortex.[18]

Across a lifetime. Ageing affects every structural element of the neocortex that has been counted, including synapses, neurons and myelinated fibres; the total length of myelinated fibres, 150,000–180,000 km in young people, shows a large reduction with age.[1]

When synapses fail

Myasthenia gravis is an autoimmune disease in which antibodies against the acetylcholine receptor, or related proteins of the postsynaptic muscle membrane, cause muscle weakness; it is treated with acetylcholinesterase inhibitors, thymectomy and immunotherapy. In 1973, rabbits immunised with purified acetylcholine receptor made antibodies against it and developed flaccid paralysis with the electrical signs of neuromuscular blockade, which anticholinesterase drugs dramatically relieved.[19,20]

Alzheimer's disease. In 15 patients and 9 controls, cognitive test scores correlated only weakly with plaques and tangles but powerfully with the density of neocortical synapses: synapse loss was the major correlate of cognitive impairment.[21]

Anti-NMDA-receptor encephalitis. In 100 patients with antibodies against the NMDA receptor, all had psychiatric symptoms or memory problems, and most had seizures, reduced consciousness and abnormal movements. The antibodies reduced the number of NMDA receptors at synapses, an effect reversed when the antibodies were removed; 75 patients recovered or had mild deficits.[22]

The mathematics of synapses

Transmission is probabilistic, the NMDA receptor follows a voltage-dependent law, and plasticity can be described by rules that keep learning stable.[3,11,23]

Quantal release[3,15]
P(k)=(nk)pk(1−p)n−k,m=n pP(k) = \binom{n}{k} p^{k} (1-p)^{n-k}, \qquad m = n\,p

If a terminal has n release sites, each releasing a quantum with probability p, the number of quanta released by one spike follows a binomial distribution with mean m (the quantal content). The postsynaptic response is then roughly m times the size of a single quantum.

Symbols in Quantal release
SymbolMeaningUnit
nnnumber of release sites—
pprelease probability per site—
kknumber of quanta released—
mmmean quantal content—
Magnesium block of the NMDA receptor[4,11]
gNMDA(V)=gˉ1+[Mg2+]o3.57 mM e−0.062 Vg_{\mathrm{NMDA}}(V) = \frac{\bar{g}}{1 + \dfrac{[\mathrm{Mg}^{2+}]_o}{3.57\ \mathrm{mM}}\, e^{-0.062\,V}}

The conductance of NMDA receptors rises steeply with depolarisation because magnesium block is relieved. With 1 mM magnesium, most of the conductance is blocked near the resting potential and much of it is available near 0 mV, which makes the receptor a detector of coincident input and output activity.

Symbols in Magnesium block of the NMDA receptor
SymbolMeaningUnit
gˉ\bar{g}maximal conductance—
[Mg2+]o[\mathrm{Mg}^{2+}]_oextracellular magnesium concentrationmM
VVmembrane potentialmV
The BCM rule[23]
dwidt=ϕ ⁣(y,θM)xi,θM=f ⁣(⟨y⟩)\frac{dw_i}{dt} = \phi\!\left(y, \theta_M\right) x_i, \qquad \theta_M = f\!\left(\langle y \rangle\right)

A synapse's change depends on presynaptic activity x and a non-linear function φ of postsynaptic activity y: below a threshold θ_M, φ is negative (weakening); above it, positive (strengthening). The threshold slides with the time-averaged postsynaptic activity, which keeps learning stable and lets neurons develop selectivity, as for orientation in visual cortex.

Symbols in The BCM rule
SymbolMeaningUnit
wiw_istrength of synapse i—
xix_ipresynaptic activity at synapse i—
yypostsynaptic activity—
θM\theta_Mmodification threshold—
⟨y⟩\langle y \rangleslowly varying time average of y—

Technology: connectomes and synthetic synapses

Mapping every synapse. Electron microscopy now reconstructs synapses at scale. The FlyWire project mapped a whole adult fruit fly brain: 50 million chemical synapses between 139,255 neurons, with predicted neurotransmitters, openly available for browsing and analysis. A cubic millimetre of human temporal cortex, removed during epilepsy surgery, revealed about 150 million synapses, and rare powerful inputs of up to 50 synapses among thousands of weak ones.[6,12]

Synapses in silicon. A nanoscale memristor, a two-terminal device whose conductance can be adjusted by the charge passing through it, can act as a synapse and supports spike-timing-dependent plasticity when combined with CMOS neurons. Neuromorphic processors such as Loihi implement on-chip learning with spiking neurons.[7,24]

Input neuronsMemristor synapsesOutput neurons
A memristor crossbar (schematic). A schematic of a neuromorphic circuit: each memristor's conductance plays the part of a synaptic weight and can be changed by spike timing.[7]
Synaptic plasticity at a glance[2,5,14,15,18]
FormTimescaleEvidence here
Short-term depressionMilliseconds to secondsPaired recordings in neocortex
Long-term potentiationHours or longerRabbit hippocampus, 1973
Spike-timing-dependent plasticitySet by millisecond spike orderCultured hippocampal neurons
LTD and LTP of a memorySwitches memory off and onOptogenetics in an animal model
Developmental pruningMonths to yearsHuman cortex

Milestones

From the heart's nerves to a fly's connectome

  1. 1921Loewi shows that nerve effects on the heart are transmitted chemically.[8]
  2. 1936Dale and Loewi share the Nobel Prize for chemical transmission.[9]
  3. 1949Hebb's postulate of synaptic strengthening.[13]
  4. 1954Del Castillo and Katz describe quantal release.[3]
  5. 1970Katz, von Euler and Axelrod share the Nobel Prize.[10]
  6. 1973Long-term potentiation; an autoimmune model of myasthenia.[5,20]
  7. 1982The BCM theory of synaptic modification.[23]
  8. 1984Magnesium block explains the NMDA receptor's voltage dependence.[4]
  9. 1991Synapse loss is the main correlate of cognitive impairment in Alzheimer's disease.[21]
  10. 1998Spike-timing-dependent plasticity.[14]
  11. 2000Carlsson, Greengard and Kandel share the Nobel Prize.[17]
  12. 2008Anti-NMDA-receptor encephalitis characterised in 100 patients.[22]
  13. 2014A memory is switched off and on with LTD and LTP.[2]
  14. 2024A whole fly brain and a cubic millimetre of human cortex mapped synapse by synapse.[6,12]

Frontiers

Whole-brain wiring. The fly connectome lets researchers trace pathways from photoreceptors to descending motor neurons and analyse information flow across the whole brain; its authors present its technologies and open ecosystem as groundwork for connectomes of other species.[6]

The human synapse map. In the human cortical fragment, glia outnumbered neurons two to one, and most connections were weak, with rare strong ones; its authors expect such resources to bring valuable insights into the human brain.[12]

Check yourself

Check yourself

  1. What did Loewi's 1921 experiment show?
    Show answer

    That the effect of nerves on the heart is transmitted by a chemical agent.

  2. What is a quantum in synaptic transmission?
    Show answer

    A packet of transmitter of fixed size; release consists of whole numbers of quanta.

  3. Why does the NMDA receptor act as a coincidence detector?
    Show answer

    Magnesium blocks it at rest, so it conducts only when glutamate is bound and the cell is depolarised.

  4. What did Bliss and Lømo discover in 1973?
    Show answer

    Long-term potentiation: a long-lasting increase in synaptic transmission after high-frequency stimulation.

  5. How did Nabavi and colleagues show that LTP and LTD are linked to memory?
    Show answer

    Optogenetic LTD of amygdala inputs inactivated a fear memory and LTP reactivated it.

  6. What best correlated with cognitive impairment in Terry's Alzheimer's study?
    Show answer

    The density of neocortical synapses (synapse loss), not plaques or tangles.

  7. In the binomial model, what is the mean quantal content?
    Show answer

    m = n p, the number of release sites times the release probability.

  8. In the BCM rule, what happens to the threshold when a neuron is very active?
    Show answer

    It slides upwards, because it depends on the time-averaged postsynaptic activity.

Glossary[2,3,4,5,6,7,8,15,18]

Synapse
A junction where one neuron passes a signal to another cell.
Neurotransmitter
A chemical released at a synapse to signal to the next cell.
Quantum
A packet of neurotransmitter of fixed size.
Release probability
The chance that a release site releases a quantum when a spike arrives.
NMDA receptor
A glutamate receptor blocked by magnesium at rest, central to many forms of plasticity.
Long-term potentiation
A long-lasting, activity-dependent strengthening of synaptic transmission.
Long-term depression
A long-lasting, activity-dependent weakening of synaptic transmission.
Short-term depression
A temporary fall in transmission during repeated spikes.
Synaptic pruning
The elimination of synapses during development.
Connectome
A complete map of neurons and their synaptic connections.
Memristor
An electronic device whose conductance can be set by the charge passed through it.

References

  1. Pakkenberg B, Pelvig D, Marner L, Bundgaard MJ, Gundersen HJG, Nyengaard JR, Regeur L. Aging and the human neocortex. Experimental Gerontology 2003;38(1-2):95-99. doi:10.1016/S0531-5565(02)00151-1
  2. Nabavi S, Fox R, Proulx CD, Lin JY, Tsien RY, Malinow R. Engineering a memory with LTD and LTP. Nature 2014;511(7509):348-352. doi:10.1038/nature13294
  3. del Castillo J, Katz B. Quantal components of the end-plate potential. The Journal of Physiology 1954;124(3):560-573. doi:10.1113/jphysiol.1954.sp005129
  4. Nowak L, Bregestovski P, Ascher P, Herbet A, Prochiantz A. Magnesium gates glutamate-activated channels in mouse central neurones. Nature 1984;307(5950):462-465. doi:10.1038/307462a0
  5. Bliss TVP, Lømo T. Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. The Journal of Physiology 1973;232(2):331-356. doi:10.1113/jphysiol.1973.sp010273
  6. Dorkenwald S, Matsliah A, Sterling AR, Schlegel P, Yu SC, McKellar CE, et al.. Neuronal wiring diagram of an adult brain. Nature 2024;634(8032):124-138. doi:10.1038/s41586-024-07558-y
  7. Jo SH, Chang T, Ebong I, Bhadviya BB, Mazumder P, Lu W. Nanoscale memristor device as synapse in neuromorphic systems. Nano Letters 2010;10(4):1297-1301. doi:10.1021/nl904092h
  8. Loewi O. Über humorale Übertragbarkeit der Herznervenwirkung. Pflügers Archiv für die Gesamte Physiologie des Menschen und der Tiere 1921;189(1):239-242. doi:10.1007/BF01738910
  9. Nobel Prize Outreach. The Nobel Prize in Physiology or Medicine 1936. NobelPrize.org 1936. https://www.nobelprize.org/prizes/medicine/1936/summary/
  10. Nobel Prize Outreach. The Nobel Prize in Physiology or Medicine 1970. NobelPrize.org 1970. https://www.nobelprize.org/prizes/medicine/1970/summary/
  11. Jahr CE, Stevens CF. Voltage dependence of NMDA-activated macroscopic conductances predicted by single-channel kinetics. The Journal of Neuroscience 1990;10(9):3178-3182. doi:10.1523/JNEUROSCI.10-09-03178.1990
  12. Shapson-Coe A, Januszewski M, Berger DR, Pope A, Wu Y, Blakely T, et al.. A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution. Science 2024;384(6696):eadk4858. doi:10.1126/science.adk4858
  13. Hebb DO. The Organization of Behavior. Psychology Press 2005. doi:10.4324/9781410612403
  14. Bi GQ, Poo MM. Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type. The Journal of Neuroscience 1998;18(24):10464-10472. doi:10.1523/JNEUROSCI.18-24-10464.1998
  15. Tsodyks MV, Markram H. The neural code between neocortical pyramidal neurons depends on neurotransmitter release probability. Proceedings of the National Academy of Sciences 1997;94(2):719-723. doi:10.1073/pnas.94.2.719
  16. Kandel ER. The molecular biology of memory storage: a dialogue between genes and synapses. Science 2001;294(5544):1030-1038. doi:10.1126/science.1067020
  17. Nobel Prize Outreach. The Nobel Prize in Physiology or Medicine 2000. NobelPrize.org 2000. https://www.nobelprize.org/prizes/medicine/2000/summary/
  18. Huttenlocher PR, Dabholkar AS. Regional differences in synaptogenesis in human cerebral cortex. The Journal of Comparative Neurology 1997;387(2):167-178. doi:10.1002/(SICI)1096-9861(19971020)387:2<167::AID-CNE1>3.0.CO;2-Z
  19. Gilhus NE, Tzartos S, Evoli A, Palace J, Burns TM, Verschuuren JJGM. Myasthenia gravis. Nature Reviews Disease Primers 2019;5:30. doi:10.1038/s41572-019-0079-y
  20. Patrick J, Lindstrom J. Autoimmune response to acetylcholine receptor. Science 1973;180(4088):871-872. doi:10.1126/science.180.4088.871
  21. Terry RD, Masliah E, Salmon DP, Butters N, DeTeresa R, Hill R, et al.. Physical basis of cognitive alterations in Alzheimer's disease: synapse loss is the major correlate of cognitive impairment. Annals of Neurology 1991;30(4):572-580. doi:10.1002/ana.410300410
  22. Dalmau J, Gleichman AJ, Hughes EG, Rossi JE, Peng X, Lai M, et al.. Anti-NMDA-receptor encephalitis: case series and analysis of the effects of antibodies. The Lancet Neurology 2008;7(12):1091-1098. doi:10.1016/S1474-4422(08)70224-2
  23. Bienenstock EL, Cooper LN, Munro PW. Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex. The Journal of Neuroscience 1982;2(1):32-48. doi:10.1523/JNEUROSCI.02-01-00032.1982
  24. Davies M, Srinivasa N, Lin TH, Chinya G, Cao Y, Choday SH, et al.. Loihi: a neuromorphic manycore processor with on-chip learning. IEEE Micro 2018;38(1):82-99. doi:10.1109/MM.2018.112130359

Related readings

  • Neurons and electrical signals

    How a nerve cell makes a 100-millivolt spike, the equations that describe it, and why one neuron may be a deep network in disguise.

    Introductory

  • Neural networks, biological and artificial

    How brains and machines learn from experience, the maths both share, and why AI and neuroscience keep borrowing from each other.

    Intermediate

  • The hippocampus and memory

    How the brain records new memories and maps space, what patient H.M. revealed, and how AI navigators and memory implants learn from it.

    Intermediate

  • The limbic system, cingulate and insula

    How the brain's inner rim and the hidden insula sense the body, shape feelings, register pain and conflict, and how stimulators and decoders now target mood.

    Intermediate

  • The cerebral cortex and its maps

    What the folded outer layer of the brain is made of, why it folds, and how scientists map it from Brodmann's microscope to petabyte connectomes.

    Introductory

  • Mapping the brain: the breakthroughs

    From Golgi and Brodmann to whole-brain connectomes, cell atlases and AI foundation models: how the brain is mapped and where mapping is heading.

    Introductory

Template anatomy for education. Not patient-specific. Not for clinical decision-making.