Effects of EEG reference-electrode selection on visual oddball experiments

In so-called “oddball” experiments, which are commonly used to study cognitive processes such as attention, subjects are presented a series of frequent “standard” stimuli and must identify infrequent “target” stimuli. Upon detection of target stimuli, widespread neural activity can be measured, which in electroencephalography (EEG) studies presents as the P3 event-related potential (ERP). In such studies, undesirable noise in the EEG data can typically be removed by “referencing” the data to a set of electrodes, the choice of which can have differential effects on the analyses of resulting ERPs. The purpose of this study was therefore to investigate whether the selection of two common reference methods (average-reference and mastoids-reference) has any measurable effects on ERPs evoked during a canonical visual oddball experiment. As a methodological study, findings from traditional parametric statistical analysis were further evaluated against an increasingly popular nonparametric technique. Although it was predicted based on prior studies that a P3 effect would be observed in the target condition for both reference-selection methods, a trending P3 effect was only observed using the mastoids-reference method, and the nonparametric technique indicated a trending effect observed over different scalp sites in an earlier latency range. Potential reasons and implications are discussed, such as the small sample size of this study, when permutation tests may ultimately be preferred, and why the choice of processing and analysis methods such as these should be made with strong justification.

Figure 3. Grand average ERP waveforms for channels P3 and P4. Mastoids-reference waveforms are shown on top, and average reference waveforms are shown on bottom. The target stimulus waveform is shown in gray, the standard stimulus waveform is shown in red, and the difference wave is shown in blue. A larger amplitude ERP can be seen in the mastoids-reference waveforms, and a similar positive voltage difference can be seen in both sets of waveforms in the P3 latency range, although the morphologies appear different.