Hebb’s Theory: How Neurons That Fire Together Wire Together

The phrase “neurons that fire together wire together” sums up an idea proposed in 1949 by psychologist Donald Hebb: brain cells that are active together become more strongly linked. In this view, learning isn’t stored in single cells but in the strengthened connections between them. What began as a guess about the mind has since been traced to a specific molecule that only switches on when two cells are active at once.1PubMed Central. A synaptic model of memory: long-term potentiation in the hippocampus The slogan leaves out two things, though: the order of firing matters, and the same rule can wire in bad habits as easily as good ones.

What Hebb Actually Proposed

Brain cells pass signals across tiny junctions called synapses, where one cell’s long output fibre meets the next cell. Hebb’s rule says that when cell A is close enough to excite cell B and repeatedly takes part in making it fire, a growth or chemical change happens in one or both cells.2Wikipedia. Hebbian theory Afterward, A becomes more effective at firing B.3Wikipedia. Hebbian theory The strength of a connection, then, is a record of how often the two cells worked together.

Take an everyday case. If you smell coffee every morning while you read the news, the cells handling the smell and the cells handling the reading are active together, and each repeat strengthens their shared connections a little more. Eventually the smell alone can bring the morning routine to mind. That’s how associations form without any deliberate effort.

The slogan slightly misstates Hebb. His own wording stresses that A takes part in causing B to fire, and two cells that happen to fire together at random aren’t the same as one cell driving the other.

Hebb built his idea on an earlier discovery. In 1944 he learned of work showing the brain is full of looping circuits.4PubMed Central. Donald O. Hebb and the Organization of Behavior: 17 years in the writing That work found one cell alone often cannot excite another, but can when another cell fires at the same moment.5PubMed Central. Donald O. Hebb and the Organization of Behavior: 17 years in the writing Hebb saw this as the building block for a theory of how the mind works.6PubMed Central. Donald O. Hebb and the Organization of Behavior: 17 years in the writing

His phrase “a growth or chemical change” was a deliberately open guess. He wrote in 1949, before anyone could watch a synapse change, so he named no mechanism. The rule says only that the change happens in one or both cells, not which, and later research found both sides can change. What matters is that the change persists after the activity ends, which is what makes this a theory of memory rather than of momentary firing.

How a Connection Physically Gets Stronger

The main lab model of learning at the level of connections is long-term potentiation (LTP), a lasting strengthening of a synapse, studied especially in the hippocampus, a memory hub.7PubMed Central. A synaptic model of memory: long-term potentiation in the hippocampus The best-understood form depends on the NMDA receptor, a sensor on the receiving cell that responds to the messenger chemical glutamate.8PubMed Central. A synaptic model of memory: long-term potentiation in the hippocampus Glutamate alone usually isn’t enough to open it, because a magnesium particle plugs the channel.9PubMed Central. NMDARs, Coincidence Detectors of Astrocytic and Neuronal Activities The plug is pushed out only when the receiving cell is already electrically excited.10PubMed Central. NMDARs, Coincidence Detectors of Astrocytic and Neuronal Activities So the receptor opens only when the sending cell (releasing glutamate) and the receiving cell (excited) are active at the same time, which is why it acts as a coincidence detector.11PubMed Central. NMDARs, Coincidence Detectors of Astrocytic and Neuronal Activities

The sequence runs like this. The sending cell releases glutamate into the gap, and it sticks to receptors on the receiving cell. If the receiving cell is already excited at that moment, the plug leaves the channel and calcium flows into the receiving side of the synapse.12PubMed. The role of CaMKII autophosphorylation for NMDA receptor-dependent synaptic potentiation That calcium surge inside the cell is the signal that says these two cells were active together.

The calcium activates an enzyme that keeps itself switched on.13PubMed. The role of CaMKII autophosphorylation for NMDA receptor-dependent synaptic potentiation Because that enzyme outlasts the calcium surge, the change can persist after the original signal is gone. The cell inserts more receptors into the synapse, making it more responsive to the same signal.14PubMed Central. CaMKII phosphorylation of TARPγ-8 is a mediator of LTP and learning and memory The synapse itself grows larger too.15PubMed. The role of CaMKII autophosphorylation for NMDA receptor-dependent synaptic potentiation When glutamate was applied while the cell was excited, spines physically expanded and their responses grew.16PubMed Central. Structural plasticity of dendritic spines

One connection can change without changing all of them because tiny bumps called spines on receiving cells act as separate chambers, so each can detect coincidence on its own.17PubMed Central. Mechanisms of calcium influx into hippocampal spines: heterogeneity among spines, coincidence detection by NMDA receptors, and optical quantal analysis Each spine handles its own calcium, so only synapses that were active at the right moment strengthen. A single cell can store many different associations this way.

Why Timing and Order Matter

Connections can also weaken, a process called long-term depression.18PubMed Central. Targeting homeostatic synaptic plasticity for treatment of mood disorders The exact order and gap between the two cells’ signals decide which happens and by how much.19PubMed Central. The spike timing dependence of plasticity When the sender fires just before the receiver, the link strengthens, while receiver-first weakens it.20PubMed Central. The spike timing dependence of plasticity This works only when the two signals come within a window of roughly 10 to 100 milliseconds of each other.21PubMed Central. The spike timing dependence of plasticity The rule has been found in creatures from insects to humans.22PubMed. Spike timing-dependent plasticity: a Hebbian learning rule

That fits Hebb’s original wording. A sender that fires just before plausibly helped cause the response. One that fires just after couldn’t have, so the link fades. In this way the brain rewards prediction, not coincidence alone, and the slogan is better read as “fire in sequence, wire together.”

Weakening isn’t a failure. It lets the brain prune links that no longer predict anything useful. Without it, every connection could only ever go up, and the network would lose the contrast that makes patterns stand out.

Hebbian Wiring in Memory, Skills and Fear

One part of the hippocampus is richly connected back onto itself, which lets it store experiences as linked patterns.23PubMed Central. The mechanisms for pattern completion and pattern separation in the hippocampus Activating a small part of a memory can then bring back the whole thing.24PubMed Central. The mechanisms for pattern completion and pattern separation in the hippocampus In recordings from inside people’s brains, remembering a pairing re-created the brain activity from when it was first learned.25eLife. Hippocampal pattern completion is linked to gamma power increases and alpha power decreases during recollection When many cells have been strengthened together into a pattern, activating a few of them sends signals along those strengthened links to the rest, so the network completes itself from partial input.

The cost is that recall is partly reconstruction rather than playback. The network fills in what’s missing, which can produce memories that feel vivid and certain yet contain details added or blended from other experiences. Confidence in a memory is not a reliable measure of its accuracy.

Practising a skill builds new connections, but later than you might expect. Long skill training builds new synapses in the adult movement-control area of the brain.26PubMed Central. Cortical synaptogenesis and motor map reorganization occur during late, but not early, phase of motor skill learning In animals, this growth came in the late phase of training, not in early learning.27PubMed Central. Cortical synaptogenesis and motor map reorganization occur during late, but not early, phase of motor skill learning New wiring seems to lock in a skill rather than create the first improvement.28PubMed Central. Cortical synaptogenesis and motor map reorganization occur during late, but not early, phase of motor skill learning In people, magnetic stimulation timed so input arrived just before output improved motor learning, while other timings did not.29PubMed Central. Hebbian priming of human motor learning

Fear is a clear case of two things becoming wired together. Learning to fear a sound paired with a shock depends on changes in the amygdala, an emotion centre.30PubMed. Synaptic plasticity in the lateral amygdala: a cellular hypothesis of fear conditioning The pathway carrying the sound strengthens, and drugs blocking that strengthening there also block the fear learning in animals.31PubMed. Synaptic plasticity in the lateral amygdala: a cellular hypothesis of fear conditioning

When Wiring Goes Wrong

Drugs of abuse can co-opt the strengthening mechanisms in the brain’s reward circuits.32Nature Reviews Neuroscience. Synaptic plasticity and addiction A leading view therefore treats addiction as a harmful but powerful form of learning.33Nature Reviews Neuroscience. Synaptic plasticity and addiction These drugs trigger a dopamine surge that causes widespread changes in connections, and those changes may be an early step toward compulsive drug-seeking in some users.34PubMed. Synaptic plasticity and addiction: learning mechanisms gone awry

Cravings get triggered by places and people through the same pairing process. Anything reliably present when a drug was taken, such as a location, a person or a time of day, becomes linked to the drug’s effect. Later, meeting that cue can activate the linked network and bring on craving without the drug being present.

Pain can be wired in too. Phantom limb pain tracks closely with how much the brain’s map of the missing limb has reorganised.35PubMed. Maladaptive plasticity, memory for pain and phantom limb pain: review and suggestions for new therapies Lost inhibition, LTP-like strengthening and new nerve sprouting reshape those circuits.36PubMed. Maladaptive plasticity, memory for pain and phantom limb pain: review and suggestions for new therapies

What Keeps the Brain From Overwiring

Pure strengthening would make busy circuits ever busier. Each strengthened connection would make its circuit more active, which would strengthen it further, until such circuits dominated and fired out of control, drowning out quieter ones. So the brain also uses stabilising processes that scale connection strengths to keep firing rates steady.37eLife. The interplay between homeostatic synaptic scaling and homeostatic structural plasticity maintains the robust firing rate of neural networks They work like a thermostat, nudging overall strengths back into a workable range. Some brain regions even use reversed rules, where co-activity weakens links, to learn sequences.38PubMed Central. Anti-Hebbian plasticity drives sequence learning in striatum

Scaling adjusts connections across a cell in proportion, so the differences between strong and weak links are preserved while the overall level comes down. A memory stored as a pattern of relative strengths can therefore survive being turned down overall. That differs from the specific weakening of one link caused by bad timing.

Sleep plays a major role in forming long-term memory.39PubMed. Mechanisms of systems memory consolidation during sleep During deep sleep, the hippocampus replays recent patterns, gradually building them into the outer brain.40PubMed. Mechanisms of systems memory consolidation during sleep This happens alongside a general scaling down of connections.41PubMed. Mechanisms of systems memory consolidation during sleep

Using Hebb’s Rule to Learn Better

Repeated co-activity is what strengthens links, so spaced repetition beats one cram. Present a cue directly before the thing you want to link it to, since order matters. Keep practising a skill past the early gains so new connections can form.42PubMed Central. Cortical synaptogenesis and motor map reorganization occur during late, but not early, phase of motor skill learning And protect your sleep after learning.

The same rule can help break a habit. Interrupt the cue-response sequence consistently, and insert a new response right after the cue. Repeat until the new link outcompetes the old. Expect it to take time.

Researchers are also testing timed stimulation. Pairing a wrist nerve pulse with magnetic stimulation of the brain increased motor responses in people within 30 minutes.43PubMed. Induction of plasticity in the human motor cortex by paired associative stimulation The effect lasted at least 30 to 60 minutes and was reversible.44PubMed. Induction of plasticity in the human motor cortex by paired associative stimulation A version is being trialled in small pilot studies for Alzheimer’s memory problems.45PubMed. Effects of repetitive paired associative stimulation on brain plasticity and working memory in Alzheimer’s disease: a pilot randomized double-blind-controlled trial Hebb’s rule also inspired learning methods in artificial intelligence.46Machine Learning and Knowledge Extraction. Delving into Unsupervised Hebbian Learning from Artificial Intelligence Perspectives