Dendrites: The Neuron's Computing Input Tree

Abstract

Dendrites are the branched extensions of a neuron that receive synaptic input and conduct it toward the cell body, where the decision to fire is made. Far from passive wires, they are active, compartmentalized processors: their geometry shapes how signals attenuate, and voltage-gated channels let individual branches generate local spikes that transform how inputs are summed. This article traces the field from Wilfrid Rall's cable theory, which first showed how dendritic shape governs the spread of electrical signals, through the discovery of sodium, calcium, and NMDA dendritic spikes, to the modern view of the single dendritic branch as a functional computing unit. It covers dendritic structure, passive electrotonic signaling, active synaptic integration, and the dendritic spine as the elementary site of plasticity, with a worked example and three interactive demonstrations.

Keywords: dendrite, cable theory, synaptic integration, dendritic spike, dendritic spine

What Dendrites Are

A *dendrite* is a tapering, branched process that extends from the body of a neuron and carries electrochemical signals inward, from synaptic contacts toward the soma. A single neuron typically bears one axon, which sends signals out, and many dendrites, which bring signals in; together the dendrites form the dendritic tree, or arbor, that constitutes most of a neuron's receptive surface. The great majority of excitatory synapses a neuron receives land on its dendrites, so the dendritic tree is where the raw material of neural computation first arrives.

For much of the twentieth century dendrites were pictured as passive cables that merely funneled input to the soma. That picture has been overturned. Dendritic membrane carries voltage-gated sodium, calcium, and potassium channels, and the glutamate-gated NMDA receptor, and these active conductances let dendrites amplify, filter, and locally regenerate signals rather than only attenuate them (Häusser, Spruston, & Stuart, 2000). The modern account, synthesized across six decades of work, treats the dendritic tree as a layered set of semi-independent processors whose branching structure and channel complement jointly determine what the neuron computes (Stuart & Spruston, 2015).

Key Takeaways

Dendrites are the input side of the neuron, receiving most of its synapses and conducting signals toward the soma.
Cable theory, founded by Wilfrid Rall, explains how a dendrite's diameter and membrane properties set the distance a passive signal travels before it decays.
Active channels let dendrites generate local sodium, calcium, and NMDA spikes, so integration is often nonlinear rather than a simple sum.
A single dendritic branch can act as a functional unit, integrating clustered inputs on its own before passing a result to the rest of the cell.
Dendritic spines are the tiny protrusions that host most excitatory synapses and are the elementary sites of structural plasticity.

Types of Dendrites

The Medical Subject Headings vocabulary files *Dendrites* (D003712) beneath *Neurons* in tree A08.675.256 and lists the narrower descriptors shown in Table 1. This placement reflects indexing practice rather than a functional taxonomy: MeSH is a controlled vocabulary for retrieving literature, so its subtypes group anatomical structures for cataloguing, not the physiological classes (apical, basal, oblique) by which neuroscientists usually sort dendrites. The parent kind, *Neurons*, is the cell class to which dendrites belong; the three entries below are glossed from their ordinary anatomical meaning and do not yet have their own articles on this site.

Table 1. Direct subtypes of Dendrites in the MeSH classification (tree A08.675.256).
Subtype	In brief
Dendritic Spines	The micron-scale membranous protrusions that stud a dendrite and host most excitatory synapses, each a near-isolated biochemical compartment.
Growth Cones	The motile, sensing tips of a developing dendrite or axon that steer its outgrowth toward targets during wiring.
Neurites	The general term for any projection from a neuron before it is committed as an axon or a dendrite, used especially in culture and development.

Structure and Classes

Neuroscientists classify dendrites by their position and role on the cell rather than by the indexing scheme above. On a cortical pyramidal neuron the tree divides into two systems: a single thick *apical* dendrite that ascends from the top of the soma toward the cortical surface and ends in a spreading tuft, and several *basal* dendrites that radiate from the base. Shorter *oblique* branches leave the apical trunk along its length. This layout is not incidental. Because different cortical layers carry different information, a tree that samples several layers lets one neuron combine feedforward input arriving on its basal dendrites with feedback and contextual input arriving on its apical tuft (Spruston, 2008).

The tree's fine geometry matters as much as its gross plan. Dendrites taper as they branch, and the diameter of a branch, together with the specific membrane resistance and the internal resistance of the cytoplasm, sets how far a signal spreads before it fades. Rall captured this with the *equivalent cylinder*, a mathematical reduction showing that a branched tree obeying a simple diameter rule behaves, electrically, like a single unbranched cable, which made the otherwise intractable geometry analytically tractable (Rall, 1959).

Figure 1

The Anatomy of a Pyramidal Neuron's Dendritic Tree

The apical trunk samples superficial cortical layers through its tuft, while basal dendrites gather input near the soma; the axon alone carries output away.

Cable Theory and Passive Signaling

Even a dendrite with no active channels is not a perfect conductor. A synaptic current injected at one point spreads along the membrane while leaking outward across it, so the resulting voltage falls with distance. For a long, uniform passive cable this decay is exponential in the steady state, described by the *length constant* or electrotonic length, written as the Greek letter lambda. Lambda equals the square root of the quantity (branch diameter times specific membrane resistance) divided by (four times the internal resistivity). A larger diameter or a leakier-resistant membrane lengthens lambda, letting signals travel farther; a thin, leaky branch shortens it, confining signals locally. Rall's insight was that this passive framework, applied to the real branched geometry of a neuron, quantitatively predicts how a synaptic potential recorded at the soma depends on where on the tree it originated (Rall, 1959).

Passive cable properties therefore impose a bias: a synapse close to the soma has a larger, faster impact on the cell's output than an identical synapse far out on a thin branch, whose signal arrives smaller and more smeared in time. The first demonstration below lets the reader vary distance and diameter and watch the passive attenuation predicted by lambda; the Worked Example puts numbers to it. This distance-dependent filtering is the baseline that the neuron's active channels then partly correct, so that distal inputs are not simply lost.

Demo 1 · Passive attenuation along a dendrite

A 10 mV synaptic potential decays exponentially as it travels toward the soma along a passive cable. Vary the branch diameter and the distance and watch how much survives. The length constant is λ = √(d R_m / 4R_i), with R_m = 20,000 Ω·cm² and R_i = 150 Ω·cm.

Branch diameter: 1.0 µm

Distance from soma: 200 µm

λ = 577 µm · at 200 µm the potential arrives as 7.1 mV of the original 10 mV (70.7% remains, a 29% loss).

An exact steady-state cable model with representative fixed resistances; real dendrites taper and carry active channels that offset this decay. Values computed locally, not stored.

Synaptic Integration and Dendritic Spikes

If dendrites were purely passive, a neuron would simply add up the attenuated contributions of its synapses. In reality, dendritic membrane is studded with voltage-gated channels that make integration nonlinear. When enough excitatory input arrives close together in space and time on a thin branch, it can trigger a regenerative event, a *dendritic spike*, in which the branch's own channels amplify the response far beyond a linear sum (Magee, 2000). Three classes of dendritic spike are now recognized, distinguished by their ionic basis and location, and summarized in Table 2.

The most consequential of these is the interaction between the soma and the apical tuft. An action potential initiated at the axon does not only travel outward; it also propagates backward into the dendrites as a *back-propagating action potential*. When such a back-propagating spike coincides with strong input to the apical tuft, it can ignite a dendritic calcium spike, coupling the two ends of the cell and driving a burst of output. Larkum and colleagues showed that this coincidence mechanism lets a pyramidal neuron associate input arriving in different cortical layers, effectively using the apical dendrite as an associative element (Larkum, Zhu, & Sakmann, 1999). A separate class of event, the NMDA spike, arises in thin terminal branches when clustered glutamatergic inputs recruit the regenerative current of NMDA receptors, producing a local plateau of depolarization (Major, Larkum, & Schiller, 2013).

Table 2. The three principal classes of dendritic spike in cortical pyramidal neurons.
Spike class	Ionic basis	Typical location	Functional role
Sodium spike	Voltage-gated Na⁺ channels	Apical trunk and proximal dendrites	Fast, brief boosting of strong local input and back-propagation of the axonal spike.
Calcium spike	Voltage-gated Ca²⁺ channels	Apical tuft and distal trunk	Sustained plateau that couples cortical layers and drives bursting when input coincides with a back-propagating spike.
NMDA spike	NMDA-receptor current	Thin basal and tuft branches	Local supralinear integration of clustered inputs, letting a single branch act as a computing unit.

Because these regenerative events are confined to the branches that host them, a large dendritic tree behaves less like one summing point and more like a set of semi-independent subunits, each capable of a local nonlinear operation before it reports to the rest of the cell. This is the empirical basis for treating the single dendritic branch as a fundamental functional unit of the nervous system (Branco & Häusser, 2010). The second demonstration contrasts a linear, passive sum of inputs with the supralinear jump that occurs when the same inputs are clustered on one active branch.

Demo 2 · Linear sum versus a dendritic spike

Add excitatory inputs clustered on a single thin branch. Below threshold they sum roughly linearly; once enough arrive together, the branch fires an NMDA spike and the response jumps far above the linear prediction — the branch acting as its own computing unit.

Coactive clustered inputs: 3

Below the spike threshold (4 inputs): the branch sums roughly linearly to 12 mV. Add more clustered inputs to ignite the plateau.

Illustrative threshold model of NMDA-spike integration; real thresholds and gains vary with branch, receptor density, and input timing. Values computed locally, not stored.

Dendritic Spines and Plasticity

Most excitatory synapses in the cortex do not contact the dendritic shaft directly; they sit on *dendritic spines*, tiny membranous protrusions, each typically less than a micron across, connected to the shaft by a thin neck. Yuste and Denk used two-photon imaging to show that calcium entering through a synapse on a spine is largely confined to that spine head, so each spine acts as a separate biochemical compartment and, they proposed, a basic functional unit of neuronal integration (Yuste & Denk, 1995). The spine neck is the key: its narrow geometry throttles the diffusion of ions and messengers between head and shaft, isolating the chemistry of one synapse from its neighbors.

That compartmentalization is what makes input-specific plasticity possible. Because the calcium signal that triggers strengthening or weakening is kept local, one synapse can change without dragging its neighbors along, which is a physical prerequisite for the synapse-specific long-term potentiation thought to underlie memory. Spines are also structurally dynamic: their size tracks synaptic strength, larger heads holding stronger, more stable synapses, and this structure-stability-function relationship links a spine's shape to how long its synapse is likely to last (Kasai, Matsuzaki, Noguchi, Yasumatsu, & Nakahara, 2003). When many potentiated synapses cluster on the same stretch of dendrite, their combined local depolarization can recruit dendritic nonlinearities, a synaptic-clustering arrangement that computational and experimental work links to how memories are allocated within a tree (Kastellakis & Poirazi, 2019). The third demonstration lets the reader vary spine-neck resistance and watch how tightly a calcium signal is held in the spine head.

Demo 3 · The spine neck as a compartment

Calcium entering a spine head can either stay local or leak down the neck into the shaft. A thin, high-resistance neck confines the signal, isolating one synapse’s chemistry from its neighbors. Vary the neck resistance and watch how much calcium is retained in the head.

Spine-neck resistance: 50 (relative)

50% of the calcium signal is retained in the spine head; 50% leaks to the shaft. A thinner, higher-resistance neck keeps the synapse’s chemistry more isolated.

Illustrative single-compartment leak model of spine-neck confinement; real spines vary in neck geometry and buffering. Values computed locally, not stored.

Worked Example: Passive Attenuation Along a Dendrite

Consider an excitatory postsynaptic potential of 10 mV generated on a passive dendritic branch, and ask how much of it survives the trip to the soma. Take a branch of diameter 1 micron (1 × 10⁻⁴ cm), a specific membrane resistance of 20,000 ohm-centimeter-squared, and an internal resistivity of 150 ohm-centimeter, values in the usual range for cortical dendrites.

The length constant is the square root of (diameter times membrane resistance) divided by (four times internal resistivity). Numerically, the numerator is 1 × 10⁻⁴ multiplied by 20,000, which is 2.0; the denominator is four times 150, which is 600. The quotient is 0.00333 centimeter-squared, whose square root is 0.0577 centimeter, that is, lambda is about 577 micrometers.

In the steady state the voltage decays as the starting voltage times the exponential of minus distance over lambda. At 200 micrometers the exponent is minus 200 over 577, or minus 0.347, and its exponential is 0.707, so the 10 mV potential arrives as 7.1 mV, a 29 percent loss. At 500 micrometers the exponent is minus 0.866, its exponential is 0.421, and the potential arrives as just 4.2 mV, a 58 percent loss. The lesson is quantitative: a synapse 500 micrometers out on a thin passive branch delivers well under half the somatic impact of an identical synapse at the cell body, which is exactly the distance-dependent penalty that active dendritic channels evolved to offset. The first demonstration reproduces these figures as the distance and diameter are varied.

Discussion

The arc from Rall to the present is a shift in what a neuron is taken to be. The classical view placed all of the computation at the axon's trigger zone, with dendrites as passive antennae; the modern view distributes computation across the tree, so that a pyramidal neuron performs many local operations in its branches before committing to an output (London & Häusser, 2005). This has consequences beyond cellular physiology. If a single neuron with an actively integrating tree can carry out operations that once seemed to require a small network, then the brain's computational capacity has been systematically underestimated, and the mapping between biological neurons and the simple summing units of artificial neural networks is looser than often assumed (Poirazi & Papoutsi, 2020).

The dendritic view also reframes classic questions in cognitive psychology. Synapse-specific plasticity, the cellular substrate most often invoked for learning, depends on the spine's compartmentalization; the capacity of a neuron to bind together inputs from different cortical layers depends on apical calcium spikes; and the storage of related memories may exploit the clustering of synapses on shared branches. Each of these is a case where a psychological function has a specific dendritic mechanism, rather than a diffuse cellular one. The dendrite is, in this sense, a bridge between the biophysics of a single membrane and the information processing that supports working memory and thought.

Current Directions

The most striking recent development is the study of human dendrites directly, using tissue resected in neurosurgery. Human cortical neurons are larger than those of rodents, and their long dendrites appear to be electrically more compartmentalized, so that distal and proximal regions are more electrically isolated from one another than in the rodent, potentially giving each human neuron more independent processing subunits (Beaulieu-Laroche et al., 2018). A related discovery is a previously unknown dendritic event in human layer 2/3 pyramidal cells, a calcium-mediated dendritic action potential whose graded, tuned responses let a single dendrite compute operations, such as the exclusive-or, that a classic pointlike neuron cannot (Gidon et al., 2020).

These findings feed an active exchange with machine learning and computational modeling. Detailed models are being used both to interpret the flood of dendritic recordings and imaging data and to ask what abstract computations the biology affords, and there is renewed interest in whether artificial networks that incorporate dendrite-like nonlinearities learn more efficiently than those built from simple units (Poirazi & Papoutsi, 2020). Open questions remain about how consistently these mechanisms operate in the intact, behaving brain, and how the many spike types interact during natural activity rather than under controlled stimulation.

Common Misconceptions

Dendrites are passive wires that only pass signals along.: Dendritic membrane carries voltage-gated sodium, calcium, and potassium channels and NMDA receptors, and can generate local regenerative spikes; integration is frequently nonlinear rather than a passive relay (Häusser, Spruston, & Stuart, 2000).
Every synapse contributes equally to whether a neuron fires.: Passive cable properties mean a distal synapse on a thin branch has far less somatic impact than an identical proximal one; location on the tree, not just synaptic strength, sets a synapse's weight (Rall, 1959).
Action potentials only travel away from the cell body.: Axonal spikes also propagate backward into the dendrites, and this back-propagating action potential is a signal that a neuron has fired, used to time plasticity and to trigger dendritic calcium spikes (Larkum, Zhu, & Sakmann, 1999).
A dendritic spine is just a passive anchor point for a synapse.: The spine is a distinct biochemical compartment whose thin neck confines calcium and messengers, making it a functional unit of integration and the site of synapse-specific plasticity (Yuste & Denk, 1995).

Glossary

Active conductance.: A voltage- or ligand-gated ion channel in the dendritic membrane that lets the dendrite amplify or regenerate signals rather than only attenuate them.
Apical dendrite.: The single thick dendrite that ascends from the top of a pyramidal soma toward the cortical surface, ending in a tuft.
Axon.: The single output process of a neuron that carries action potentials away from the soma, complementary to the input-receiving dendrites.
Back-propagating action potential.: An axonal action potential that travels backward into the dendritic tree, signaling that the neuron has fired and helping trigger dendritic spikes and plasticity.
Basal dendrite.: One of the dendrites that radiate from the base of a pyramidal soma and gather input near the cell body.
Cable theory.: The mathematical framework, founded by Rall, that treats a dendrite as a leaky electrical cable to predict how voltage spreads and decays.
Compartmentalization.: The electrical and biochemical isolation of one part of a neuron from another, so that a signal in a spine or branch stays local.
Dendritic spike.: A regenerative, locally generated depolarization in a dendrite, driven by sodium, calcium, or NMDA-receptor currents.
Dendritic spine.: A small membranous protrusion on a dendrite, connected by a thin neck, that hosts an excitatory synapse and acts as a separate compartment.
Electrotonic length (lambda).: The length constant of a passive cable, the distance over which a steady voltage decays to about 37 percent of its starting value.
Equivalent cylinder.: Rall's reduction of a branched dendritic tree, under a diameter rule, to a single electrically equivalent unbranched cable.
Excitatory postsynaptic potential (EPSP).: The transient depolarization produced in a dendrite by an excitatory synaptic input.
NMDA spike.: A dendritic spike carried by NMDA-receptor current, triggered by clustered glutamatergic input on a thin branch, producing a local plateau.
Passive membrane.: Membrane whose response is set only by fixed resistance and capacitance, without voltage-gated channels, so it can only attenuate signals.
Pyramidal neuron.: The principal excitatory cortical cell type, named for its triangular soma, with the characteristic apical-and-basal dendritic layout.
Synaptic integration.: The process by which a neuron combines its many synaptic inputs, across space and time, into a decision to fire or not.

Key Researchers

Michael Häusser

(b. 1966). A neuroscientist at University College London who has studied dendritic integration and pioneered two-photon and optogenetic methods for reading and writing activity in single dendrites; a Fellow of the Royal Society.
Wikipedia - Google Scholar - UCL profile

Matthew Larkum

(b. 1966). A neurophysiologist at the Humboldt University of Berlin who discovered dendritic calcium spikes and the coincidence mechanism that couples input arriving in different cortical layers of pyramidal neurons.
ORCID - Google Scholar - Faculty page

Panayiota Poirazi

A computational neuroscientist at IMBB-FORTH in Crete whose models of dendritic nonlinearities have clarified their role in learning, memory, and the computational power of single neurons.
ORCID - Wikipedia - Google Scholar - Faculty page

Wilfrid Rall

(1922-2018). The founder of cable theory and compartmental modeling of dendrites, who showed how dendritic geometry shapes the spread and integration of synaptic signals, laying the theoretical foundation for the whole field.
Wikipedia - Wikidata

Nelson Spruston

A neuroscientist and executive director at HHMI's Janelia Research Campus whose work on pyramidal-neuron dendritic structure and synaptic integration, especially in the hippocampus, has mapped how dendrites process input.
ORCID - Google Scholar - Janelia profile

Greg Stuart

A neuroscientist at Monash University who pioneered dendritic patch-clamp recording, enabling the first direct electrical measurements from dendrites and the study of back-propagation and dendritic spikes.
ORCID - Google Scholar - Faculty page

Rafael Yuste

A neuroscientist at Columbia University who imaged calcium in single dendritic spines, proposed the spine as a functional unit of integration, and helped initiate the BRAIN Initiative.
Wikipedia - Wikidata - Columbia faculty page

Frequently Asked Questions

What are dendrites?

Dendrites are the branched extensions of a neuron that receive synaptic inputs from other cells and conduct the resulting electrical signals toward the cell body. They form the dendritic tree, which carries most of a neuron's synapses and is where incoming information is first integrated (Stuart & Spruston, 2015).

What is the difference between dendrites and axons?

Dendrites are the input side of a neuron and are usually multiple and branched, while the axon is the single output process that carries action potentials away from the soma. Signals normally flow inward along dendrites to the cell body and outward along the axon, though action potentials also back-propagate into dendrites (Spruston, 2008).

What are dendritic spines?

Dendritic spines are tiny membranous protrusions, typically under a micron across, that stud a dendrite and host most excitatory synapses. Each spine is a separate biochemical compartment whose thin neck confines calcium, making it a functional unit of integration and a site of synapse-specific plasticity (Yuste & Denk, 1995).

How do dendrites process information?

A dendrite combines its inputs across space and time, but not always by simple addition. Passive cable properties weight inputs by their distance from the soma, and active channels can make clustered inputs sum supralinearly, so a dendrite performs genuine integration rather than passive relay (Magee, 2000).

What is a dendritic spike?

A dendritic spike is a regenerative depolarization generated within a dendrite by its own voltage- or ligand-gated channels. Sodium, calcium, and NMDA-receptor spikes each occur in characteristic locations and let a branch amplify strong or clustered input rather than merely passing it on (Major, Larkum, & Schiller, 2013).

What is cable theory and the length constant?

Cable theory treats a dendrite as a leaky electrical cable and predicts how a voltage spreads along it. The length constant, lambda, is the distance over which a steady signal decays to about 37 percent of its starting value, set by the branch's diameter and its membrane and internal resistances (Rall, 1959).

Can a single dendrite perform a computation on its own?

Yes. Because regenerative events stay confined to the branches that host them, a single dendritic branch can integrate clustered inputs nonlinearly and act as a functional computing unit before reporting to the rest of the cell (Branco & Häusser, 2010).

Are human dendrites different from those of other animals?

Human cortical neurons have unusually long dendrites that appear to be more electrically compartmentalized than a rodent's, and human layer 2/3 cells show a distinctive calcium-mediated dendritic spike, suggesting each human neuron may support more independent processing than previously assumed (Beaulieu-Laroche et al., 2018).

References

Beaulieu-Laroche, L., Toloza, E. H. S., van der Goes, M. S., Lafourcade, M., Barnagian, D., Williams, Z. M., Eskandar, E. N., Frosch, M. P., Cash, S. S., & Harnett, M. T. (2018). Enhanced dendritic compartmentalization in human cortical neurons. Cell, 175(3), 643-651.e14. https://doi.org/10.1016/j.cell.2018.08.045

Branco, T., & Häusser, M. (2010). The single dendritic branch as a fundamental functional unit in the nervous system. Current Opinion in Neurobiology, 20(4), 494-502. https://doi.org/10.1016/j.conb.2010.07.009

Gidon, A., Zolnik, T. A., Fidzinski, P., Bolduan, F., Papoutsi, A., Poirazi, P., Holtkamp, M., Vida, I., & Larkum, M. E. (2020). Dendritic action potentials and computation in human layer 2/3 cortical neurons. Science, 367(6473), 83-87. https://doi.org/10.1126/science.aax6239

Häusser, M., Spruston, N., & Stuart, G. J. (2000). Diversity and dynamics of dendritic signaling. Science, 290(5492), 739-744. https://doi.org/10.1126/science.290.5492.739

Kasai, H., Matsuzaki, M., Noguchi, J., Yasumatsu, N., & Nakahara, H. (2003). Structure-stability-function relationships of dendritic spines. Trends in Neurosciences, 26(7), 360-368. https://doi.org/10.1016/S0166-2236(03)00162-0

Kastellakis, G., & Poirazi, P. (2019). Synaptic clustering and memory formation. Frontiers in Molecular Neuroscience, 12, 300. https://doi.org/10.3389/fnmol.2019.00300

Larkum, M. E., Zhu, J. J., & Sakmann, B. (1999). A new cellular mechanism for coupling inputs arriving at different cortical layers. Nature, 398(6725), 338-341. https://doi.org/10.1038/18686

London, M., & Häusser, M. (2005). Dendritic computation. Annual Review of Neuroscience, 28, 503-532. https://doi.org/10.1146/annurev.neuro.28.061604.135703

Magee, J. C. (2000). Dendritic integration of excitatory synaptic input. Nature Reviews Neuroscience, 1(3), 181-190. https://doi.org/10.1038/35044552

Major, G., Larkum, M. E., & Schiller, J. (2013). Active properties of neocortical pyramidal neuron dendrites. Annual Review of Neuroscience, 36, 1-24. https://doi.org/10.1146/annurev-neuro-062111-150343

Poirazi, P., & Papoutsi, A. (2020). Illuminating dendritic function with computational models. Nature Reviews Neuroscience, 21(6), 303-321. https://doi.org/10.1038/s41583-020-0301-7

Rall, W. (1959). Branching dendritic trees and motoneuron membrane resistivity. Experimental Neurology, 1(5), 491-527. https://doi.org/10.1016/0014-4886(59)90046-9

Spruston, N. (2008). Pyramidal neurons: Dendritic structure and synaptic integration. Nature Reviews Neuroscience, 9(3), 206-221. https://doi.org/10.1038/nrn2286

Stuart, G. J., & Spruston, N. (2015). Dendritic integration: 60 years of progress. Nature Neuroscience, 18(12), 1713-1721. https://doi.org/10.1038/nn.4157

Yuste, R., & Denk, W. (1995). Dendritic spines as basic functional units of neuronal integration. Nature, 375(6533), 682-684. https://doi.org/10.1038/375682a0

Dendrites