A brief introduction to EEG & ERP

Created: 2025-01-29

What is EEG?

Electroencephalography (EEG) records electrical activity produced by (populations of) neurons  

  • Measured through electrodes placed on the scalp

  • Captures real-time brain activity with millisecond precision

  • Non-invasive, relatively inexpensive method

Historical Development

  • Hans Berger records first human EEG (1929)

    • First human neuroimaging technique

  • Pauline and Hallowell Davis credited with first observations of auditory evoked ERPs in 1936

  • Jasper 1937, first visual ERPs

  • Walter et al. (1964) first report of the CNV (contingent negative variation, a frontal negative potential representing anticipatory attention)

  • 1960s-1970s, new computational techniques leads to ERP methodology (averaging)

Cellular Basis of EEG

Pyramidal neurons are the main source of EEG signals

  • Large cortical neurons oriented perpendicular to surface; dendrites extend toward cortical surface

  • Synchronized postsynaptic potentials create measurable fields, polarity depending on orientation of cortical surface

    • Scalp-recorded potentials only possible for layered structures with consistent orientations, i.e. primarily cerebral cortex

  • Requires ~10,000 neurons firing together to generate detectable signal –> low spatial resolution

EEG Frequency Bands

  • Delta (0.5-4 Hz): Deep sleep

  • Theta (4-8 Hz): Drowsiness, meditation

  • Alpha (8-13 Hz): Relaxed wakefulness

  • Beta (13-30 Hz): Active thinking

  • Gamma (>30 Hz): Complex processing

EEG vs fMRI

EEG fMRI
Temporal resolution Good (in milliseconds) Low (in seconds)
Spatial resolution Poor (in centimeters) High (in 1-2mm voxels)
Cost Low relative to fMRI Very high
Portability Portable systems available Requires fixed, dedicated installation
Comfort Minimal discomfort Loud, may be claustrophobic
Motion sensitivity Moderate High
Measured activity Direct measurement of neuronal electrical activity Indirect measurement via blood oxygen levels (BOLD signals)
Limitations

Low spatial resolution

Only measures cortex

 Low temporal resolution

Modern EEG Recording Systems

  • Ag/AgCl electrodes most common

  • Electrode-skin interface:

    • Conductive gel reduces impedance

  • Signal acquisition parameters

    • Sampling rate: typically 250-2000 Hz
    • Analog-to-digital conversion: 16-24 bit
    • Amplification: 1000-100,000x gain
    • Online filtering: high-pass 0.1-1 Hz; low-pass ~100-200 Hz
    • Notch: 50/60 Hz (power line)

BioSemi ActiveTwo vs. Brain Products actiCHamp

Feature BioSemi ActiveTwo Brain Products actiCHamp
Active Electrodes Yes (Active Pin-Type) Yes (actiCAP active)
Reference Scheme CMS/DRL feedback loop Traditional reference
Max Channels 256 160
Sampling Rate Up to 16384 Hz Up to 100 kHz
Resolution 24-bit 24-bit
Input Range ±262 mV ±400 mV
Bandwidth DC to 3.2 kHz DC to 7.5 kHz
Interface USB2/Optical fiber USB
Battery Operation Yes No (USB powered)
Trigger Input 16-bit parallel 8-bit parallel/serial
Special Features

Zero reference design

Replaceable electrode tips

Active shielding

Impedance measurement

Built-in calibration

Electrode position detection
Software ActiView BrainVision Recorder

Sources of Noise in EEG Recordings (1/2)

Biological Artifacts

  1. Muscle activity (EMG)
    • High frequency (>20 Hz)
    • Preventive: participant relaxation
    • Analysis: ICA, high-pass filtering
  2. Eye movements & blinks
    • Large amplitude deflections
    • Preventive: eye movement controls
    • Analysis: ICA, regression-based correction
  3. Cardiac activity (ECG)
    • Regular rhythmic activity
    • Analysis: ICA, template matching

Environmental Noise

  1. Power line interference (50/60 Hz)
    • Preventive: Proper grounding
    • Analysis: Notch filter
  2. Electronic equipment
    • Preventive: Shield recording area
    • Keep electronics away from EEG system

Sources of Noise in EEG Recordings (2/2)

Technical Noise

  1. Electrode issues
    • Poor contact: Check impedance
    • Bridging: Do not overgel
    • Cable movement: Secure cables
  2. Amplifier noise
    • Regular calibration
    • Proper maintenance
    • Temperature stability

Handling Noise

  1. Prevention during recording:
    • Proper shielding
    • Subject instruction
    • Equipment maintenance
    • Regular impedance checks
  2. Analysis solutions:
    • Filtering (appropriate frequency bands)
    • Artifact rejection
    • ICA decomposition
    • Robust averaging techniques
    • Reference choice optimization

Time-varying neural responses to specific events

  • Events could be external stimuli, or

  • Participant internal activity (visual, cognitive processing)

  • Relies on

    • Consistent marking of the events
    • Averaging, to remove noise independent of (cannot be removed by) pre-processing

  • ERPs are latent structures

Examples:

  • N100: Early sensory processing

  • P200: Feature detection

  • N200: Stimulus discrimination

  • P300: Target detection, decision-making

  • N400: Semantic processing

Anatomy of an ERP Component

  • Polarity: Positive (P) / Negative (N)

    • Depends on: neurotransmitter type (excitatory/inhibitory), synapse location, orientation of cortical surface, superposition of other sources
    • Polarity alone has no inherent meaning
    • Same cognitive process can produce different polarities at different sites - positive ≠ excitation, negative ≠ inhibition

  • Amplitude: Size of deflection (μV)

    • Depends on: number of neurons activated, geometry of neurons, synchronicity of post-synaptic potentials, depth of source, orientation of cortical surface, superposition of other sources
    • Voltage peaks are not the same as components!

  • Latency: Time from stimulus onset (ms)

  • Duration: Time course of the component

  • Topography: Scalp distribution

Interpreting ERP components (1/2)

Common Mistakes

  • Single component ≠ single process

  • Differences in peak amplitude ≠ change in component size

  • Larger amplitude of a component ≠ “more processing”

  • Peak latency ≠ process duration

  • Over-interpreting scalp distribution

  • Confusing correlation with causation

  • Average ERP may not reflect what happens on individual trials

  • Ignoring component overlap

Interpreting ERP components (2/2)

Component Overlap

An effect in one time period doesn’t necessarily mean a modulation of the component at that time period.

  • Difference waves can sometimes reveal the underlying component time course

  • Later components affected by earlier ones

  • Solutions:

    • Use difference waves

    • Principal Component Analysis (PCA)

    • Independent Component Analysis (ICA)

    • Careful experimental design to remove confounds

Component Overlap

Best Practices in ERP Research

  • Use appropriate control conditions

  • Maintain consistent trial numbers

  • Consider individual differences

  • Document and report all pre-processing steps

  • Use standardized electrode positions

  • Document analysis parameters

  • Consider alternative explanations

Considerations for Experimental Design

  • Sufficient trial numbers (>30-40 per condition)

  • Balanced conditions

  • Appropriate inter-stimulus intervals

  • Control for:

    • Physical stimulus properties

    • Motor responses

    • Attention and arousal

    • Order effects

Basic ERP data pipeline

  1. Reference
  2. Filter
  3. Artifact rejection
  4. Epoch
  5. Baseline correction
  6. Averaging across trials (single subject)
  7. Grand averaging across subjects

Questions?

Thank you for your attention!