Related Experiment Video
Updated: Jan 10, 2026
08:45
Mapping Cortical Dynamics Using Simultaneous MEG/EEG and Anatomically-constrained Minimum-norm Estimates: an Auditory Attention Example
Published on: October 24, 2012
15.1K
No Single Best Pipeline: Multiverse Analysis of EEG Preprocessing for N-Back Working Memory Tasks
Haijing Huang1,2, Adriano H Moffa1,2, Colleen Loo1,2
1Discipline of Psychiatry and Mental Health, Faculty of Medicine and Health, School of Clinical Medicine, University of New South Wales, Sydney, New South Wales, Australia.
Psychophysiology
|November 28, 2025
Summary
Preprocessing electroencephalography (EEG) data for working memory tasks requires careful pipeline selection. Optimal EEG signal quality depends on specific research goals, as no single pipeline excels across all metrics.
Area of Science:
- Cognitive Neuroscience
- Neuroscience
- Signal Processing
Background:
- Working memory (WM) is crucial for cognitive functions and often assessed using n-back tasks.
- Electroencephalography (EEG) is widely used to study WM, but preprocessing significantly impacts event-related potentials (ERPs).
- Existing research lacks systematic comparisons of EEG preprocessing pipelines for n-back tasks.
Purpose of the Study:
- To systematically compare 43 EEG preprocessing pipelines for n-back tasks.
- To evaluate the impact of filtering, Independent Component Analysis, and re-referencing on ERPs (P300, N200, P200).
- To assess data quality across target trials, difference waveforms, and signal-to-noise ratio (SNR) consistency in clinical populations.
Main Methods:
- A multiverse analysis of 43 EEG preprocessing pipelines was conducted.
- Pipelines varied in high-pass filtering (0.5-4 Hz), low-pass filtering (10-100 Hz), ICA, and re-referencing (average, REST, linked-mastoids).
- Data from 164 participants across three n-back datasets (2-back, 3-back) were analyzed.
Main Results:
- No single pipeline was optimal across all data quality dimensions.
- A pipeline with 0.5 Hz high-pass, 100 Hz low-pass filtering, and REST referencing showed strong SNR and target-nontarget discrimination.
- 0.5 Hz high-pass filtering specifically improved target vs. nontarget discrimination.
- Pipeline choices influenced ERP outcomes at individual and group levels, affecting amplitude preservation and reliability.
Conclusions:
- EEG preprocessing pipeline selection critically impacts n-back task ERP data quality and interpretation.
- Specific preprocessing choices, such as high-pass filtering and referencing, substantially influence results.
- Tailoring preprocessing pipelines to research objectives is essential for robust and reliable EEG findings in WM studies.
