Single-subject analysis of N400 event-related potential component with five different methods

Area of Science:

• Neuroscience
• Cognitive Science
• Psychology
• Signal Processing

Background:

• Analyzing event-related potentials (ERPs) at the single-subject level is essential for understanding neural processes.
• Various statistical methods exist for ERP analysis, but their comparative performance in detecting effects at the individual level is not well-established.
• The N400 component, an event-related potential sensitive to semantic processing, is a common target for ERP research.

Purpose of the Study:

• To compare the detection rates of different single-subject ERP analysis methods for the auditory N400 effect.
• To evaluate factors influencing the results of these analysis methods.
• To provide guidance for researchers in selecting appropriate methods for single-subject ERP analysis.

Main Methods:

• Utilized electroencephalography (EEG) data from 79 healthy participants.
• Investigated the auditory N400 effect using five single-subject analysis methods: visual inspection, analysis of variance (ANOVA), cluster-based non-parametric testing, Bayesian approach, and Studentized continuous wavelet transform (t-CWT).
• Compared detection rates across different experimental paradigms (active responding and passive listening).

Main Results:

• The Bayesian method demonstrated the highest detection rate (89%) and greatest concordance between paradigms.
• Visual inspection achieved 85% detection, while ANOVA detected the effect in 68% and cluster-based testing in 59% of participants.
• Studentized continuous wavelet transform (t-CWT) methods showed the lowest detection rates (22-59%), indicating a more conservative approach. ANOVA results often required corroboration from other methods.

Conclusions:

• Different single-subject ERP analysis methods yield non-overlapping results, highlighting the complexity of choosing an optimal approach.
• The Bayesian method emerged as the most liberal and consistent across paradigms for N400 effect detection.
• Relying on a single statistical method may be insufficient for robust conclusions regarding single-subject ERPs, suggesting the need for methodological consensus or complementary analyses.