Introduction to ERPs/ERFs

Initialize NeuroAnalyzer 
using NeuroAnalyzer

Event-Related Potentials (ERPs) and Event-Related Fields (ERFs)

Definition

ERPs: Averaged electrical activity time-locked to a stimulus (e.g., a click, flash, or cognitive task).

Hypothesis: ERPs reflect complex cognitive processes (e.g., attention, memory, perception).

Assumption: Cognitive components do not vary in time across trials.

ERFs: The MEG equivalent of ERPs, measuring magnetic fields instead of electrical activity.

Event-Related Potentials (ERP): averaged electrical activity time-locked to a stimulus (e.g. click or flash).

Methodology

Averaging: ERPs/ERFs are obtained by averaging EEG/MEG signals across multiple trials to reduce noise and isolate stimulus-related activity.

  • Example: Averaging 100 epochs reduces noise from 20 μV to 2 μV.
  • Averaging 200 epochs reduces noise to ~1.4 μV.To halve the noise level achieved with 100 epochs, 400 epochs are needed (noise reduces with the square root of the number of trials).

Noise vs. Signal:

  • Noise Level: Typically ~20 μV.
  • Signal of Interest: Typically ~5 μV.

Limitation: Longer experiments are required to achieve low noise levels, which can be impractical.

Reverse Inference Problem

Reverse Inference Chain:

Neural activity → ERP/ERF response → Experiment triggers ERP/ERF component → Inference about neural activity.

Challenge: ERP/ERF components may not be specific to a single cognitive process.

Example: The same ERP component (e.g., N170) might be activated by multiple neural processes.

Solution: Reverse inference is only as valid as the empirical evidence linking the component to a specific neural activity.

Baseline Correction

Purpose: Ensures that ERP/ERF amplitudes are measured relative to zero by removing pre-stimulus offsets.

Baseline Interval: Typically -200 ms to 0 ms (pre-stimulus period).

Process:

  • Compute the mean signal during the baseline interval for each channel.
  • Subtract this mean from all time points in the epoch or averaged time series.

Frequency Domain: Baseline correction can also involve subtracting baseline spectral power from event-related spectral responses.

Effects of choosing a baseline window: