Rereferencing

Rereferencing is the process of changing the reference electrode or reference scheme used to compute electrical potentials in electrophysiological recordings, recalculating each channel relative to a new reference. It is commonly applied to electroencephalography (EEG) data and, to a lesser extent, to magnetoencephalography (MEG) data, to reduce reference bias and improve the interpretability of scalp topographies.

In practice, rereferencing involves selecting a new reference signal and transforming the data so that the recorded potentials represent differences with respect to that reference. Common schemes include the average reference (the mean across all active electrodes), linked mastoids or earlobes, nose reference, and customary electrode locations such as Cz or FCz. More advanced methods, such as the Reference Electrode Standardization Technique (REST), attempt to estimate an reference that approximates an ideal, neutral point.

Rereferencing is typically performed during preprocessing after data collection and can occur before or after filtering, artifact rejection, and epoching, depending on the goals of the analysis. The choice of reference scheme can materially affect the apparent distribution and timing of signals, so it should be matched to the study design and the spatial coverage of the sensor array.

Considerations and caveats include the influence of the reference on topographic maps and source localization, potential bias when many channels are missing or noisy, and the fact that rereferencing cannot recover information that was not captured by the recording configuration. In MEG, reference schemes are used less often because the data are less affected by a common reference; in EEG/MEG combinations, average reference is a common default. Software tools (for example, EEG/MEG analysis packages) provide options to apply and customize rereferencing in preprocessing pipelines.