Abstract
Electroencephalography (EEG) is the recording of the brain's electrical activity from electrodes placed on the scalp, capturing the summed postsynaptic activity of large populations of cortical neurons. Its defining virtue is temporal resolution: EEG follows the brain on the millisecond timescale at which cognition unfolds. This article covers how the signal is generated and recorded since Hans Berger's first human recording in 1929, and the two complementary ways it is read — as frequency-domain rhythms (delta, theta, alpha, beta, gamma) whose power tracks states from sleep to focused attention, and as time-domain event-related potentials that index the stages of perception and cognition. What EEG cannot do well is localise its sources, an ill-posed inverse problem. Three interactive demonstrations let readers explore alpha blocking, the frequency bands, and how averaging extracts an event-related potential.
Keywords: electroencephalography, EEG, event-related potentials, neural oscillations, brain rhythms
Electroencephalography is the oldest continuously used method for observing the living human brain at work, and it remains among the most informative. It records, from electrodes resting on the scalp, the tiny fluctuating voltages produced by the brain's own electrical activity — the electroencephalogram, or EEG. What makes those voltages worth recording is not their size, which is measured in microvolts, but their timing: EEG resolves neural events on the scale of milliseconds, the scale on which perception, attention, and thought actually operate (Cohen, 2017).
That temporal precision is the reason EEG belongs to cognitive psychology and not only to clinical neurology. Where methods such as functional MRI ask where in the brain a process happens, EEG is unusually well suited to asking when, and in what order — whether one process precedes another, how long a computation takes, and how the brain's ongoing rhythms shape the processing of an incoming stimulus. The sections below set out what EEG is, the forms it takes, how the signal is generated and recorded, the brain rhythms it reveals in the frequency domain, the event-related potentials it reveals in the time domain, and how the raw signal is analysed.
- EEG records the summed postsynaptic potentials of populations of cortical neurons from scalp electrodes, at millisecond temporal resolution (Cohen, 2017).
- Hans Berger recorded the first human EEG in 1929 and named the alpha rhythm; Adrian and Matthews confirmed it in 1934 (Berger, 1929; Adrian & Matthews, 1934).
- The ongoing signal decomposes into frequency bands — delta, theta, alpha, beta, gamma — whose oscillations coordinate neural activity and track cognitive state (Buzsáki & Draguhn, 2004; Klimesch, 1999).
- Averaging the signal time-locked to an event extracts the event-related potential, whose components — such as the P300 and the N400 — index specific cognitive operations (Luck, 2014; Sutton et al., 1965; Kutas & Hillyard, 1980).
- EEG offers excellent temporal but poor spatial resolution: inferring the cortical sources of a scalp signal is a mathematically ill-posed inverse problem (Michel & Murray, 2012).
Figure 1
Two Ways of Reading the EEG Signal
What Electroencephalography Is
Electroencephalography is the measurement of the electrical activity of the brain from electrodes placed on the scalp; the record it produces is an electroencephalogram, abbreviated EEG for both the technique and its output. The voltages measured are small — typically tens of microvolts — and fluctuate continuously, yielding the familiar wavering traces that were, for decades, the only window onto the electrical life of the intact human brain (Berger, 1929).
The signal does not arise from action potentials, which are too brief and too spatially disorganised to sum into a scalp-measurable voltage. It arises instead from postsynaptic potentials: the slower changes in membrane voltage produced when neurons receive input. When a large population of similarly oriented cortical pyramidal neurons — aligned in parallel, perpendicular to the cortical surface — is active together, their postsynaptic potentials add to produce an electric field large enough to be detected at the scalp (Cohen, 2017). EEG is therefore a population measure by nature: it reports the synchronised synaptic activity of tens of thousands of neurons at once, not the firing of single cells.
Two properties define EEG's place among the methods of cognitive neuroscience. Its temporal resolution is excellent, on the order of a millisecond, because the electric field propagates to the scalp effectively instantaneously; the EEG tracks neural activity as fast as that activity changes. Its spatial resolution, by contrast, is poor: the skull and scalp smear the field, and many different configurations of cortical activity could in principle produce the same pattern of scalp voltages, so localising the sources is a genuinely hard inverse problem (Michel & Murray, 2012). This trade-off — superb timing, coarse localisation — is the single most important fact about what EEG can and cannot tell us.
Types of Electroencephalography
In the Medical Subject Headings (MeSH) vocabulary maintained by the U.S. National Library of Medicine, Electroencephalography is a diagnostic technique classified under both Diagnostic Techniques, Neurological and Electrodiagnosis. Beneath it, MeSH files two narrower descriptors, listed in Table 1. It is worth stressing what this classification is and is not: MeSH is an indexing vocabulary built to organise the biomedical literature, not a theory of the EEG signal, so its narrower terms pick out topics that literature is indexed under rather than a clean taxonomy of measurement types. The two children are also not alternatives to one another — they describe orthogonal aspects of the same recordings. Brain waves are the rhythmic components of the ongoing signal itself, while EEG phase synchronization is an analysis of the temporal relationship between signals recorded at different sites.
| Descriptor | MeSH UI | What it names |
|---|---|---|
| Brain Waves | D058256 | The rhythmic oscillatory components of the ongoing EEG, grouped into frequency bands (delta, theta, alpha, beta, gamma). |
| Electroencephalography Phase Synchronization | D058407 | The consistency of the phase relationship between signals at different electrodes, used to index functional coupling between regions. |
Beyond the MeSH hierarchy, EEG is described in practice by how it is used. Spontaneous or resting EEG records the ongoing signal without a controlled stimulus and is read largely in the frequency domain; event-related EEG time-locks the recording to repeated stimuli or responses and is read in the time domain as event-related potentials. Recordings also differ by setting — routine clinical EEG, long-term monitoring, high-density research montages — but these are variations in application and instrumentation, not distinct MeSH categories.
How the EEG Signal Is Generated and Recorded
The scalp EEG is a volume-conducted signal. The synchronised postsynaptic potentials of a patch of cortex set up an electric field that spreads passively through the brain, cerebrospinal fluid, skull, and scalp to reach the recording electrodes — a process called volume conduction. Because the intervening tissues, especially the skull, are poor and unevenly resistive conductors, the field is attenuated and spatially blurred by the time it reaches the surface, which is the physical origin of EEG's limited spatial resolution (Michel & Murray, 2012). A voltage is always a difference between two points, so every EEG channel is measured relative to a reference electrode; the choice of reference shapes the recorded waveforms and is a standing methodological concern.
Electrodes are placed at standardised scalp locations. The international 10–20 system, and its high-density extensions, position electrodes at fixed proportions of the distance between skull landmarks so that recordings are comparable across people and laboratories. Research montages range from a handful of electrodes to arrays of 256 or more. The recorded signal must then be separated from artifacts — eye blinks and movements, muscle activity, heartbeat, and line noise — which are often far larger than the neural signal of interest and are a major part of the practical work of EEG (Jas et al., 2017).
The method's history begins with Hans Berger, a psychiatrist at the University of Jena, who in 1929 published the first recordings of the human electroencephalogram and described a prominent rhythm, around 10 Hz, that appeared when his participants closed their eyes and rested — the rhythm he named alpha (Berger, 1929). Berger's claims were met with scepticism until Edgar Adrian and Bryan Matthews replicated them in 1934, confirming that the “Berger rhythm” originated in the occipital cortex and was suppressed by visual attention (Adrian & Matthews, 1934). That the alpha rhythm is strong with the eyes closed and vanishes when the eyes open — alpha blocking — was the first demonstration that a scalp rhythm tracks a cognitive state, and it remains the simplest way to see the EEG respond to what the mind is doing.
Alpha blocking: the Berger effect
The posterior alpha rhythm is large when the eyes are closed and the mind is at rest, and collapses the moment the eyes open. This was the first sign that a scalp rhythm tracks a cognitive state.
Amplitude and rhythm are schematic and drawn deterministically to illustrate synchronisation versus desynchronisation, not a specific recording.
Brain Rhythms and Frequency Bands
Read in the frequency domain, the ongoing EEG is not noise but a superposition of rhythms. Neural populations oscillate, and their oscillations are conventionally grouped into frequency bands that recur across people and that relate, if imperfectly, to distinct functions (Buzsáki & Draguhn, 2004). The bands, from slowest to fastest, are delta (roughly 1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (above about 30 Hz). The boundaries are conventions, not sharp natural divisions, but the grouping is useful because power in each band behaves distinctively.
Delta dominates deep sleep. Theta appears in the hippocampus and cortex during memory encoding and navigation. Alpha, the rhythm Berger found, is prominent over posterior cortex during relaxed wakefulness and is suppressed by engagement; it is now understood less as mere idling than as an active mechanism of inhibition and attentional gating, its power rising over cortical regions to be suppressed (Klimesch, 2012). Beta is associated with active, alert processing and with motor control. Gamma has been linked to feature binding and local cortical computation. Across these bands, oscillations are thought to coordinate the timing of neural activity — providing a temporal framework within which distributed populations can communicate — which is why they are central to theories that connect EEG rhythms to memory and cognition (Klimesch, 1999).
The EEG frequency bands
The ongoing signal is a superposition of rhythms. Select a band to see a representative wave at that rate, its conventional range, and what it is typically associated with.
Band boundaries are conventions, not sharp natural divisions, and the mapping from a band to a cognitive function is many-to-many; the wave shown is schematic.
The relationship between bands and functions is real but must be stated carefully. A frequency band is defined purely by rate; it is not a thing but a range, and the same band can serve different functions in different regions and tasks. The mapping from band to cognitive process is many-to-many, and treating any single band as the signature of a single faculty is a persistent oversimplification that the careful literature resists (Buzsáki & Draguhn, 2004).
Event-Related Potentials
The second way to read the EEG is in the time domain, by extracting the event-related potential (ERP). The problem the ERP technique solves is that the response to any single stimulus is buried in the far larger ongoing EEG. The solution is signal averaging: the stimulus is presented many times, a segment of EEG is extracted time-locked to each presentation, and these segments are averaged. Activity that is not time-locked to the stimulus — the ongoing rhythms and the noise — varies randomly from trial to trial and averages toward zero, while the brain's consistent, time-locked response to the stimulus survives and emerges as the ERP waveform (Luck, 2014).
An ERP is a sequence of positive and negative voltage deflections, called components, labelled by polarity and timing: P for positive and N for negative, with a number giving the approximate latency in milliseconds or the ordinal position. Two components have been especially important to cognitive psychology. The P300, discovered by Sutton and colleagues in 1965, is a large positive deflection around 300 ms that grows when a stimulus is improbable or task-relevant, and is read as an index of the updating of attention and working memory by a meaningful event (Sutton et al., 1965; Polich, 2007). The N400, discovered by Kutas and Hillyard in 1980, is a negative deflection around 400 ms that grows when a word is semantically unexpected — larger for the last word of “he spread the warm bread with socks” than of “… with butter” — and has become the standard electrophysiological measure of semantic processing (Kutas & Hillyard, 1980; Kutas & Federmeier, 2011).
Signal averaging: pulling the ERP out of the noise
A single trial (faint) is dominated by background EEG. Averaging many trials time-locked to the stimulus cancels the random activity and lets the consistent event-related potential — here a P300 — emerge (bold).
Single-trial noise is generated deterministically (standard deviation 30 µV) and added to a fixed ERP waveform (a small N-dip and a 34 µV P300), so the figure is identical on every render.
The power of the ERP technique is that component latency and amplitude give separate, millisecond-resolved measures of mental processing. Because a component's latency reflects the timing of the operation it indexes, ERPs can establish the order and duration of cognitive stages — showing, for instance, that a stimulus was categorised before it was fully identified — in a way that slower methods cannot. The cost is that averaging discards everything not time-locked to the event, including the trial-to-trial variability and the ongoing oscillations that the frequency-domain analysis is built to study.
Analysing the EEG Signal
Turning raw recordings into either rhythms or ERPs requires substantial processing, and the analysis pipeline is itself a subject of active methodological work. A typical workflow filters the signal to the frequencies of interest, identifies and removes artifacts, segments the data around events, and only then averages or transforms it. Independent component analysis is widely used to separate the recording into statistically independent sources, isolating stereotyped artifacts such as eye blinks from genuine neural components; it was made broadly accessible by the open-source EEGLAB toolbox (Delorme & Makeig, 2004). Artifact handling has itself been increasingly automated, so that the rejection of bad data segments and channels need not depend on subjective visual inspection (Jas et al., 2017).
Two developments have reshaped the analysis landscape. The first is a move toward standardisation and reproducibility: shared data formats such as EEG-BIDS impose a common organisation on datasets so that pipelines can be applied and compared across laboratories, addressing the fragmentation that long made EEG results hard to replicate (Pernet et al., 2019). The second is a sharper account of the frequency spectrum itself. The EEG spectrum contains both genuine oscillatory peaks and a broadband, non-oscillatory background that falls off with frequency; conflating the two inflates and distorts band-power measures, and separating the periodic from the aperiodic components has become an important correction to decades of oscillation research (Donoghue et al., 2020). Recovering the cortical sources of a scalp signal — source localisation — remains constrained by the inverse problem, but improved head models and dense electrode arrays have made it a usable, if always assumption-laden, tool (Michel & Murray, 2012).
Worked Example
Consider why the ERP technique needs so many trials. The logic rests on how signal and noise behave differently under averaging. Suppose the true, time-locked ERP has an amplitude of S = 3 µV at some latency, while the ongoing background EEG contributes zero-mean random noise with a standard deviation of σ = 30 µV on each trial — ten times the size of the signal we want to see.
On a single trial the signal-to-noise ratio (SNR) is S / σ = 3 / 30 = 0.1; the response is invisible, swamped by noise thirty times its own standard deviation from the mean. Now average n trials. The time-locked signal is identical on every trial, so its averaged amplitude stays at S = 3 µV. The noise is random and independent across trials, so the standard deviation of the average of n independent samples shrinks as σ / √n. The averaged SNR is therefore S / (σ / √n) = (S / σ) × √n = 0.1 × √n.
The improvement grows only with the square root of the number of trials. To reach an SNR of 1 — signal equal to noise — we need 0.1 × √n = 1, so √n = 10 and n = 100 trials. To reach a comfortable SNR of 3 we need 0.1 × √n = 3, so √n = 30 and n = 900 trials. Doubling the SNR again, to 6, would require n = 3,600 trials. This square-root law is the central practical constraint of ERP research: because each halving of the noise costs four times as many trials, experiments are built around presenting a small set of conditions dozens or hundreds of times, and rare or fatiguing events are expensive to measure precisely (Luck, 2014).
Discussion
EEG's enduring value comes from a specific complementarity with the other tools of cognitive neuroscience. Its temporal resolution is its comparative advantage: no widely available non-invasive method follows the brain as fast, so EEG is the method of choice whenever the question is about the timing, sequence, or duration of mental events rather than their anatomical location (Cohen, 2017). The frequency-domain and time-domain readings of the signal answer different questions from the same data — the ongoing rhythms that coordinate processing, and the transient responses that index particular operations — and much of the method's reach comes from holding both in view.
The limits are equally definite. The inverse problem means EEG cannot, on its own, say with confidence where in the brain a signal arises; scalp topography constrains but does not determine the sources, and localisation always rests on assumptions (Michel & Murray, 2012). The signal is dominated by superficial, radially oriented cortex and is comparatively blind to deep and sulcal sources. And EEG is only ever a population measure of synchronised synaptic activity, several steps removed from the spiking of individual neurons that ultimately carries information. These are not flaws to be fixed but the boundaries of what the method observes.
For cognitive psychology, then, EEG is best understood as an instrument for the chronometry of mind. It made the timing of mental processes measurable a half-century before functional imaging made their location measurable, and it continues to do what imaging cannot: watch cognition unfold at the speed it actually happens.
Current Directions
Contemporary EEG research is driven as much by advances in analysis as by advances in hardware. A prominent methodological correction concerns the spectrum itself: the recognition that the EEG power spectrum mixes genuine oscillations with an aperiodic, broadband background whose slope carries its own physiological meaning has prompted a reanalysis of how band power should be measured, and tools that parameterise the spectrum into periodic and aperiodic parts are now widely adopted (Donoghue et al., 2020). This matters because a change once attributed to an oscillation may instead reflect a shift in the aperiodic background — a distinction earlier methods could not make.
A second front is reproducibility and scale. Standardised data structures such as EEG-BIDS, together with open-source analysis toolboxes and automated artifact-rejection methods, are moving the field from bespoke single-laboratory pipelines toward shared, comparable, and reusable analyses across large datasets (Pernet et al., 2019; Jas et al., 2017; Delorme & Makeig, 2004). A third is the continued refinement of what oscillations do: rather than treating a band as a fixed correlate of a faculty, current work examines how the phase and power of rhythms shape perception moment to moment, developing alpha's role from passive idling into active inhibitory control of information flow (Klimesch, 2012). Alongside these, low-cost and wearable EEG systems are extending recording out of the laboratory, raising both new applications and new questions about signal quality.
Common Misconceptions
- “EEG records the firing of individual neurons.”
- It does not. The scalp EEG reflects the summed postsynaptic potentials of large, synchronously active populations of similarly oriented cortical neurons — tens of thousands of cells at once. Action potentials are too brief and too disorganised to contribute meaningfully to the scalp signal (Cohen, 2017).
- “EEG shows where in the brain activity is happening.”
- Only very coarsely. Volume conduction blurs the signal, and infinitely many source configurations can produce the same scalp pattern — the inverse problem. EEG's strength is timing, not location; precise anatomical localisation requires strong assumptions or complementary methods (Michel & Murray, 2012).
- “Each frequency band corresponds to one mental function.”
- A band is a range of rates, not a faculty. The same band serves different roles in different regions and tasks, and the mapping between bands and cognitive processes is many-to-many. Treating alpha as “attention” or gamma as “binding” outright oversimplifies the evidence (Buzsáki & Draguhn, 2004).
- “An ERP component is a thing that switches on in the brain.”
- A component is a feature of an averaged waveform, defined by polarity, latency, and scalp distribution — not a discrete generator. It is extracted by averaging away everything not time-locked to the event, so it summarises a consistent response, not a single switch (Luck, 2014).
Glossary
- 10–20 system.
- The international standard for placing scalp electrodes at fixed proportions of the distances between skull landmarks, making recordings comparable across people and laboratories.
- Alpha rhythm.
- An oscillation of roughly 8–13 Hz, prominent over posterior cortex during relaxed wakefulness and suppressed by engagement; the rhythm Berger first described, now linked to attentional inhibition.
- Artifact.
- A non-neural signal contaminating the EEG — eye blinks and movements, muscle activity, heartbeat, or line noise — often much larger than the neural signal and requiring removal before analysis.
- Beta rhythm.
- An oscillation of roughly 13–30 Hz associated with active, alert processing and with motor control.
- Delta rhythm.
- The slowest EEG band, roughly 1–4 Hz, dominant during deep (slow-wave) sleep.
- Electroencephalography (EEG).
- The recording of the brain's electrical activity from scalp electrodes; also the record itself, the electroencephalogram.
- Event-related potential (ERP).
- The brain's electrical response time-locked to an event, extracted by averaging the EEG across many repetitions so that non-time-locked activity cancels.
- Frequency band.
- A conventional range of oscillation rates (delta, theta, alpha, beta, gamma) into which the ongoing EEG spectrum is divided for analysis.
- Gamma rhythm.
- Fast oscillations above roughly 30 Hz, linked to local cortical computation and to the binding of stimulus features.
- Independent component analysis (ICA).
- A statistical method that separates a multichannel recording into independent sources, widely used to isolate and remove stereotyped artifacts such as eye blinks.
- Inverse problem.
- The mathematically ill-posed task of inferring the cortical sources of a scalp signal, which has no unique solution because many source configurations produce the same surface voltages.
- N400.
- A negative ERP component peaking near 400 ms that grows to semantically unexpected words; the standard electrophysiological index of semantic processing.
- Neural oscillation.
- A rhythmic fluctuation in the activity of a neural population, thought to coordinate the timing of processing and communication across regions.
- P300.
- A large positive ERP component near 300 ms that grows to improbable or task-relevant stimuli; read as an index of attentional and working-memory updating.
- Postsynaptic potential.
- The change in a neuron's membrane voltage caused by synaptic input; the summed postsynaptic potentials of aligned populations are the source of the EEG signal.
- Signal averaging.
- Averaging EEG segments time-locked to repeated events so that random background activity cancels and the consistent event-related response emerges; noise falls with the square root of the number of trials.
- Theta rhythm.
- An oscillation of roughly 4–8 Hz, prominent in the hippocampus and cortex during memory encoding and spatial navigation.
- Volume conduction.
- The passive spread of an electric field from its neural source through the brain, skull, and scalp to the electrodes; the cause of EEG's spatial blurring.
Key Researchers
Hans Berger
(1873–1941). Psychiatrist at the University of Jena; recorded the first human electroencephalogram and named the alpha rhythm, founding human EEG (Berger, 1929). Wikipedia · Wikidata
György Buzsáki
(living). Systems neuroscientist at New York University; foundational work on how neuronal oscillations coordinate activity in cortical networks, linking brain rhythms to neural computation (Buzsáki & Draguhn, 2004). ORCID · Wikipedia · Google Scholar
Marta Kutas
(living). Cognitive neuroscientist at the University of California, San Diego; co-discovered the N400 component indexing semantic processing and authored its definitive review (Kutas & Hillyard, 1980; Kutas & Federmeier, 2011). Wikipedia · Google Scholar
Steven J. Luck
(living). Cognitive neuroscientist at the University of California, Davis; author of the standard methods text on the event-related potential technique, which codified best practice for ERP research (Luck, 2014). ORCID · Google Scholar
Scott Makeig
(living). Computational neuroscientist at the University of California, San Diego; created EEGLAB, the open-source toolbox that brought independent component analysis of single-trial EEG into wide use (Delorme & Makeig, 2004). ORCID · Google Scholar
Frequently Asked Questions
What is electroencephalography?
Electroencephalography (EEG) is the recording of the brain's electrical activity from electrodes placed on the scalp. The signal reflects the summed postsynaptic potentials of large populations of cortical neurons, and its great strength is temporal resolution: it follows brain activity on the millisecond timescale.
What does the EEG signal actually measure?
It measures the small voltage fluctuations produced when many similarly oriented cortical neurons are synaptically active at the same time. It is a population measure of synchronised postsynaptic activity, not a recording of individual neurons firing, which are too brief and disorganised to reach the scalp.
What is the difference between EEG and an event-related potential?
The EEG is the continuous ongoing signal. An event-related potential (ERP) is extracted from it by presenting an event many times and averaging the segments time-locked to each presentation, so that background activity cancels and the brain's consistent response to the event emerges.
What are the EEG frequency bands?
The ongoing signal is conventionally divided into delta (about 1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (above about 30 Hz). Power in each band varies with state and task, though the bands are conventions rather than sharp natural divisions.
Why does EEG have poor spatial resolution?
Because the skull and scalp blur the electric field as it spreads from its source to the electrodes (volume conduction), and because many different arrangements of cortical activity can produce the same scalp pattern. Recovering the sources from the surface signal is a mathematically ill-posed inverse problem.
What is the P300?
The P300 is a large positive ERP deflection near 300 ms after a stimulus that is improbable or task-relevant. Discovered in 1965, it is interpreted as an index of the updating of attention and working memory when a meaningful or unexpected event occurs.
Why do ERP experiments need so many trials?
Because averaging reduces random noise only in proportion to the square root of the number of trials. Cutting the noise in half requires four times as many trials, so isolating a small time-locked response from the much larger background typically demands dozens to hundreds of repetitions per condition.
How does EEG compare with functional MRI?
They are complementary. EEG has excellent temporal resolution but poor spatial resolution; functional MRI has good spatial resolution but poor temporal resolution. EEG is the better tool for questions about the timing and sequence of mental processes, fMRI for questions about their anatomical location.
References
Adrian, E. D., & Matthews, B. H. C. (1934). The Berger rhythm: Potential changes from the occipital lobes in man. Brain, 57(4), 355–385. https://doi.org/10.1093/brain/57.4.355
Berger, H. (1929). Über das Elektrenkephalogramm des Menschen. Archiv für Psychiatrie und Nervenkrankheiten, 87, 527–570. https://doi.org/10.1007/BF01797193
Buzsáki, G., & Draguhn, A. (2004). Neuronal oscillations in cortical networks. Science, 304(5679), 1926–1929. https://doi.org/10.1126/science.1099745
Cohen, M. X. (2017). Where does EEG come from and what does it mean? Trends in Neurosciences, 40(4), 208–218. https://doi.org/10.1016/j.tins.2017.02.004
Delorme, A., & Makeig, S. (2004). EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods, 134(1), 9–21. https://doi.org/10.1016/j.jneumeth.2003.10.009
Donoghue, T., Haller, M., Peterson, E. J., Varma, P., Sebastian, P., Gao, R., . . . Voytek, B. (2020). Parameterizing neural power spectra into periodic and aperiodic components. Nature Neuroscience, 23(12), 1655–1665. https://doi.org/10.1038/s41593-020-00744-x
Jas, M., Engemann, D. A., Bekhti, Y., Raimondo, F., & Gramfort, A. (2017). Autoreject: Automated artifact rejection for MEG and EEG data. NeuroImage, 159, 417–429. https://doi.org/10.1016/j.neuroimage.2017.06.030
Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and memory performance: A review and analysis. Brain Research Reviews, 29(2–3), 169–195. https://doi.org/10.1016/S0165-0173(98)00056-3
Klimesch, W. (2012). Alpha-band oscillations, attention, and controlled access to stored information. Trends in Cognitive Sciences, 16(12), 606–617. https://doi.org/10.1016/j.tics.2012.10.007
Kutas, M., & Federmeier, K. D. (2011). Thirty years and counting: Finding meaning in the N400 component of the event-related brain potential (ERP). Annual Review of Psychology, 62, 621–647. https://doi.org/10.1146/annurev.psych.093008.131123
Kutas, M., & Hillyard, S. A. (1980). Reading senseless sentences: Brain potentials reflect semantic incongruity. Science, 207(4427), 203–205. https://doi.org/10.1126/science.7350657
Luck, S. J. (2014). An introduction to the event-related potential technique (2nd ed.). MIT Press.
Michel, C. M., & Murray, M. M. (2012). Towards the utilization of EEG as a brain imaging tool. NeuroImage, 61(2), 371–385. https://doi.org/10.1016/j.neuroimage.2011.12.039
Pernet, C. R., Appelhoff, S., Gorgolewski, K. J., Flandin, G., Phillips, C., Delorme, A., & Oostenveld, R. (2019). EEG-BIDS, an extension to the brain imaging data structure for electroencephalography. Scientific Data, 6, 103. https://doi.org/10.1038/s41597-019-0104-8
Polich, J. (2007). Updating P300: An integrative theory of P3a and P3b. Clinical Neurophysiology, 118(10), 2128–2148. https://doi.org/10.1016/j.clinph.2007.04.019
Sutton, S., Braren, M., Zubin, J., & John, E. R. (1965). Evoked-potential correlates of stimulus uncertainty. Science, 150(3700), 1187–1188. https://doi.org/10.1126/science.150.3700.1187