Although the choice of the measurement reference strongly affects the measurement sensitivity, validity and comparability, selection is often based on tradition, convenience and comparability to earlier results [Dien in Behav Res Methods Ins C 30(1):34–43, 1998; Femi and Sundor in Int J Psychosom 36(1-4):23–33; 1989]. Artificial means can be applied to compensate for the referential issues, but they cannot restore any lost data. The validity of the recorded data is ultimately defined by the hardware setup. In this simulation study, common average ground reference (AR) is characterized and compared to two alternative common ground reference schemes in respect to their influence on the sensitivity distribution and technical feasibility of scalp EEG recording. It was found that, despite the polar average reference effect [Junghöfer et al. in Clin Neurophysiol 110(6):1149–1155; 1999], AR merits a significantly higher symmetricity and should be promoted generally not only in high-electrode-density studies, but also in low-channel-count studies if the stringent design requirements can be met. In low-electrode-density studies, balancing the setup may prove challenging, but successful implementation can provide nearly undistorted data. Isolation of the system is a critical parameter, but technological advances enable the requirements to be fulfilled. A physical ground should be applied if high isolation is not applicable or if it is defined by the application. The results will apply for the applied homogenous concentric 3-sphere model, but should be further studied in a realistic context if more detailed and case-sensitive information is required; the underlying phenomena are generally applicable.
EEGReference selectionAsymmetricityCommon average ground referenceSensitivityFeasibility