A seizure is a temporary state in which brain networks enter an abnormal pattern of electrical and chemical activity. It is not simply ‘too much electricity’ and it is not one identical event repeated across every type of epilepsy. Different seizures can begin through different cellular and network mechanisms, involve different brain regions, and produce very different symptoms.
Modern recordings from scalp EEG, intracranial EEG, single neurons and experimental models show that a seizure is a process with phases: conditions build that allow seizure activity to begin; local networks are recruited; activity may remain confined or spread; the seizure eventually terminates; and the brain then enters a postictal state while normal network function is restored.
A seizure is a network event
Individual neurons fire action potentials all the time. That is normal brain function. A seizure occurs when populations of neurons begin interacting in an abnormal, self-sustaining pattern that disrupts normal processing. The defining problem is therefore not one neuron firing, but altered behaviour across a network.
That network may be highly localised at first, as in many focal seizures, or involve distributed circuits from the beginning, as in some generalised seizure types. Even in a focal seizure, the region generating the earliest detectable activity may not be identical to the whole epileptogenic network.
There is no single universal seizure-onset pattern
Intracranial recordings show several recognised patterns at the start of focal seizures. Some begin with low-voltage, fast-frequency activity. Others begin with rhythmic spikes or high-amplitude discharges, sometimes described as hypersynchronous onset patterns. These patterns can reflect different underlying circuit mechanisms and are associated with different brain regions and pathologies.
This is important because the phrase ‘the seizure starts’ can sound as though one universal switch has been thrown. In reality, seizure initiation can arise through different combinations of altered synaptic transmission, intrinsic excitability, extracellular ion changes, network connectivity and local pathology.
What changes just before a seizure?
The seconds or minutes before an electrographic seizure can contain subtle changes that are difficult to detect clinically. Research has found shifts in firing patterns, synchrony, extracellular potassium, intracellular chloride and inhibitory activity in some models and human recordings. These changes are not identical across all seizures and should not be interpreted as a single universal ‘pre-seizure signature’.
Longer-timescale factors also affect the probability of seizure occurrence. Sleep–wake state, circadian and multi-day rhythms, hormones, medication levels and illness can alter network susceptibility without directly causing a seizure every time they occur.
The role of extracellular potassium
When neurons fire intensely, potassium leaves cells and accumulates in the extracellular space. If extracellular potassium rises sufficiently, surrounding neurons become easier to depolarise. This can create a positive-feedback loop in which neuronal firing changes the ionic environment in a way that promotes further firing.
Normally, neurons and astrocytes regulate extracellular potassium tightly. During seizure activity, however, ion movements can exceed normal buffering capacity. Experimental and computational work has therefore placed dynamic ion concentrations among the important contributors to seizure initiation, progression and termination.
Chloride can change inhibitory signalling
Intense GABAA-receptor activation drives chloride movement across neuronal membranes. During sustained activity, intracellular chloride can accumulate in some neurons. This shifts the chloride equilibrium potential and can make GABAergic currents less inhibitory than they were at baseline.
This does not mean GABA universally becomes excitatory during seizures. The effect depends on cell type, location, transporter function and timing. But it helps explain why an inhibitory system that initially restrains seizure activity may become less effective as a seizure develops.
The paradoxical role of inhibition at seizure onset
One of the most important changes in modern seizure science is the recognition that inhibitory interneurons can be highly active at seizure onset. In some focal seizure models, interneurons initially fire intensely and hold back principal neurons around the developing seizure core.
This surrounding inhibition can create a kind of restraint around the seizure. As the seizure progresses, ionic shifts, synaptic fatigue or other mechanisms can weaken that restraint. The transition from restrained abnormal activity to wider recruitment is an active area of research.
What is the seizure core?
Intracranial microelectrode studies have described a seizure core in which neurons become intensely recruited into the ictal rhythm, surrounded by tissue that may show strong field-potential activity without the same degree of local neuronal recruitment. This distinction helps explain why electrical activity recorded at a distance does not always mean that every nearby neuron is fully participating in the seizure.
The border between recruited and unrecruited tissue can move. Researchers have described this advancing boundary as an ictal wavefront. The mechanisms governing that recruitment are central to understanding seizure spread, which is treated in more detail on the separate network-spread page.
Synchrony matters — but seizures are not simply ‘everything firing together’
Seizures are often described as hypersynchronous activity, but that shorthand can be misleading. Some seizure onsets can initially involve relatively desynchronised high-frequency firing before more organised rhythmic activity appears. Synchronisation can also change substantially as a seizure evolves.
The clinically visible seizure is therefore the result of changing spatial and temporal coordination across networks, not a single moment when every neuron suddenly fires in lockstep.
Why symptoms depend on which network is recruited
The outward form of a seizure depends on which brain networks are involved and in what sequence. Activity involving visual cortex may produce visual symptoms; temporal-limbic networks can produce experiential, emotional or memory-related phenomena; motor networks can produce tonic, clonic or other movements; autonomic networks can alter heart rate, breathing, gastrointestinal function or other body systems.
This is why seizure semiology can provide localisation clues without functioning as a perfect map. Symptoms reflect network participation, and the area producing the first symptom may not always be the exact site where the seizure began.
Focal seizures can remain focal
A focal seizure does not inevitably spread through the whole brain. Local inhibitory circuits, network geometry, white-matter pathways and the state of connected regions can restrict propagation. Some seizures remain confined enough to produce only a limited symptom or even little obvious outward change.
When seizure activity recruits bilateral networks extensively enough, a focal seizure can evolve to a focal-to-bilateral tonic-clonic seizure. The mechanisms and pathways of propagation are covered separately so this page remains focused on the seizure process itself.
Generalised seizures are not simply focal seizures that spread faster
Generalised epilepsies involve distributed networks differently from focal epilepsies. For example, absence seizures depend strongly on abnormal thalamocortical oscillations involving reciprocal connections between cortex, thalamic relay cells and the thalamic reticular nucleus. Other generalised seizure types have different network mechanisms.
The old idea that all generalised seizures must begin everywhere in the brain at exactly the same instant has also become more nuanced. Modern network studies show that generalised epilepsies can involve specific distributed circuits rather than indiscriminate activation of the entire brain.
What happens during a tonic-clonic seizure?
In a bilateral tonic-clonic seizure, widespread network recruitment produces changing motor output over time. The tonic phase reflects sustained muscle contraction, while the clonic phase consists of repeated bursts separated by periods of muscle relaxation. These visible phases correspond to evolving patterns of neuronal activity rather than two completely separate seizures.
Breathing, heart rate and autonomic function can also change during the seizure. These changes are clinically important, but their detailed mechanisms are covered on the separate breathing, autonomic and SUDEP pages rather than duplicated here.
Why does a seizure stop?
Seizure termination is not simply the brain ‘running out of electricity’. Several mechanisms can contribute, and their importance varies with seizure type and experimental model.
Proposed mechanisms include activation of potassium currents, adenosine release, synaptic depression, depletion of readily releasable neurotransmitter, changes in extracellular and intracellular ion concentrations, recovery of inhibitory restraint, activation of neuromodulatory systems and network-level desynchronisation.
Modern work suggests that seizure progression itself can create conditions that favour termination. In focal seizure models, a reproducible sequence of excitability and ionic changes can gradually shift the network away from the state that sustained the seizure.
Why some seizures do not stop normally
If the processes that normally terminate a seizure fail, or if seizure-driving mechanisms remain dominant, seizure activity can become prolonged. This is one route into status epilepticus. Prolonged seizure activity can itself alter receptors, ion gradients and network inhibition, making termination progressively more difficult.
Status epilepticus is treated separately because its biology, risks and emergency management go beyond the mechanisms of an ordinary self-terminating seizure.
The postictal state begins when the seizure ends — but the brain is not instantly back to baseline
Electrographic seizure termination does not mean immediate restoration of normal function. After a seizure, affected networks can show suppression, altered connectivity, metabolic changes and changes in cerebral blood flow. Neuromodulators such as adenosine may remain elevated, and neurons may temporarily be less excitable.
This helps explain why confusion, sleepiness, headache, language difficulty, weakness, memory gaps or behavioural changes can persist after the visible seizure has stopped. Different postictal symptoms reflect different networks and mechanisms, which are explored in the dedicated Postictal Recovery section.
Blood flow and metabolism change during and after seizures
Seizure activity is metabolically demanding. Active tissue requires increased oxygen and glucose, and cerebral blood flow usually rises in involved regions during the ictal period. After some seizures, blood-flow patterns can reverse or become regionally reduced.
These haemodynamic changes matter because imaging techniques such as ictal SPECT and postictal perfusion studies can sometimes help localise seizure networks. They also show that a seizure involves vascular and metabolic responses as well as neuronal electrical activity.
Glial cells participate too
Astrocytes regulate extracellular potassium, remove neurotransmitters such as glutamate, control water balance and interact with blood vessels. During intense seizure activity, these homeostatic systems are heavily challenged.
Microglia and inflammatory pathways can also influence neuronal excitability, particularly in some acquired epilepsies. Seizures are therefore not purely neuronal events, even though neuronal firing is central to their expression.
What EEG actually records during a seizure
EEG records summed electrical activity generated mainly by populations of cortical neurons, not individual action potentials. During a seizure, EEG patterns can change in frequency, amplitude, rhythmicity and spatial distribution.
Scalp EEG has limited spatial resolution and can miss deep, small or unfavourably oriented seizure sources. Intracranial EEG can record closer to the generating network but samples only the implanted regions. No recording method gives a complete view of every neuron involved.
Why two seizures in the same person can look different
Seizures from the same epileptic network can vary because the brain state is different each time. Sleep stage, medication concentration, recent seizures, hormonal state and ongoing network activity can influence which pathways are recruited and how far activity spreads.
That means variation does not necessarily imply a different diagnosis or a new seizure focus. Equally, major changes in seizure pattern can be clinically important and sometimes warrant reassessment.
What science still does not know
Researchers still cannot explain precisely why one susceptible network produces a seizure at a particular moment while another does not. We also do not have one mechanism that explains every seizure type or every epilepsy.
Much of the detailed cellular evidence comes from animal models, resected human tissue or intracranial recordings in people with drug-resistant focal epilepsy. Those findings are invaluable but do not automatically describe every epilepsy syndrome. Distinguishing established mechanisms from model-specific findings is therefore essential.
What this page does — and does not — explain
This page follows the biological sequence of a seizure itself: initiation, recruitment, ictal activity, termination and immediate recovery. The electrical signal of an individual neuron, excitation–inhibition balance, detailed seizure propagation and the long-term development of epilepsy are covered on separate Science & Discovery pages so that each subject has one clear home.
Sources
de Curtis et al. — Mechanisms Leading to Initiation, Development, and Termination of Focal Seizures, Jasper’s Basic Mechanisms of the Epilepsies, 5th ed. (2024)
Role of Ion Concentration Dynamics in Epileptic Seizures, Jasper’s Basic Mechanisms of the Epilepsies, 5th ed. (2024)
Cho & Paz — Convergence of Thalamic Mechanisms in Genetic Epilepsies, Jasper’s Basic Mechanisms of the Epilepsies, 5th ed. (2024)
Jasper’s Basic Mechanisms of the Epilepsies, 5th ed. — full reference work
Evidence checked: September 2026.