For the sake of simplicity, the preceding exercises in this chapter were performed on grand average ERP waveforms. This is fine for visualization, but you will ordinarily need to re-reference the single-subject data. The kinds of re-referencing we’ve been discussing so far in this chapter can be applied to the continuous EEG, the epoched EEG, or the averaged ERPs. In many cases, however, you will need to re-reference the EEG prior to artifact rejection or correction. In this exercise, we’ll see how to re-reference the continuous EEG.
Before you start this exercise, I recommend quitting and restarting EEGLAB so that everything is fresh. You won’t need any of the data you created in the previous exercises.
In the Chapter_5 folder, you’ll find a dataset named 6_N400_unreferenced.set. This is EEG dataset for the participant we looked at in Chapter 2. In that chapter, we looked at a dataset that had already been referenced, but this version has not be referenced. However, it has been filtered (0.1–30 Hz). Launch EEGLAB and load this file (EEGLAB > File > Load existing dataset) and take a look at the EEG (EEGLAB > Plot > Channel data (scroll)). Set the vertical scale to 100 µV, and click the >> button once to scroll to the 5 second point. You should see something like Screenshot 5.6.
Screenshot 5.6
These data were recorded with a BioSemi ActiveTwo EEG recording system, which saves the single-ended data rather than saving the referenced data. So, the voltages that you’re looking at are the raw voltages between each active electrode and the ground electrode (or, more precisely, the common mode sense electrode, which is BioSemi’s equivalent of ground). You can see EEG from the P9 and P10 channels, which were used as the reference in data from the previous exercises. In the present exercise, we’ll reference (not re-reference) the data to the average of P9 and P10.
We'll also create special bipolar versions of the horizontal EOG and vertical EOG (HEOG and VEOG) channels. To understand why this is useful, take a look at the channel labeled VEOG-lower at the bottom of the plot. This electrode was located just below the right eye, and the negative-going deflection that you can see in this channel shortly after the 5 second mark is an eyeblink. If you look at the FP1 and FP2 channels at the same time, you’ll see a positive-going deflection. This pattern occurs because eyeblinks (and vertical eye movements) arise from a dipole located inside the eyes, and electrodes under versus over the eyes are on opposite sides of this dipole, yielding opposite polarities.
Brain activity spreads over the entire head, so the VEOG-lower electrode picks up brain activity as well as the electrooculogram (EOG) voltage produced by eyeblinks. However, most brain activity will be quite similar at electrodes just above and just below the eye. We can therefore isolate the EOG activity and largely eliminate the brain activity by subtracting the FP2 signal (just above the right eye) from the VEOG-lower signal (just below the right eye). In addition, because the blink activity is positive at FP2 and negative at VEOG-lower, this subtraction also increases the size of the blink activity. When we try to reject trials with blinks, this subtraction makes our job much easier, because it makes the blinks bigger and makes non-blink EEG activity smaller. Thus, when we reference the data from our EEG electrodes, we’ll also create a bipolar VEOG channel in which we subtract FP2 from VEOG-lower.
All EEG voltages are actually “bipolar” in the sense of having two poles (active and reference). However, the term bipolar is used in EEG recordings when a channel uses a special reference that is different from the other channels.
We also usually create a bipolar HEOG signal to isolate horizontal eye movements. During a typical EEG recording, we place one HEOG electrode next to the left eye (HEOG-left) and another next to the right eye (HEOG-right). When the eyes move leftward, this produces a negative voltage at HEOG-left and a positive voltage at HEOG-right. This reverses for rightward eye movements. By creating a bipolar channel (HEOG-right minus HEOG-left), we can effectively double the size of the eye movement voltage. In addition, brain activity is usually quite similar at these two sites, so this subtraction also eliminates most of the brain activity. Thus, the bipolar HEOG signal is very useful when we try to reject trials with horizontal eye movements.
Enough talk—let’s try it! We’ll reference each scalp channel to the average of P9 and P10, and we’ll create bipolar VEOG and HEOG channels. With the 6_N400_unreferenced.set dataset loaded and active, select EEGLAB > ERPLAB > EEG Channel Operations. This is nearly identical to ERP Channel Operations, but it operates on the EEG (whether continuous or epoched). Clear out any existing equations and set the mode to Create new dataset.
Click the Reference assistant button and type (ch9+ch27)/2 into the Ch_REF text box. P9 is in Channel 9, and P10 is in Channel 27, so this expression gives us the average of P9 and P10. Check the boxes shown in Screenshot 5.7. Even the EOG channels will benefit from having a reference electrode, so indicate that All channels should be included. The click OK to create the equations. In the main EEG Channel Operations GUI, you should now see that each channel being created will be computed as the original channel minus the average of P9 and P10.
Screenshot 5.7
Now we need to add equations for creating the bipolar EOG signals. For reasons that will become clear in the chapter on artifact correction, it’s often a good idea to have both the bipolar signals and the signals referenced to the average of P9 and P10. To make this happen, add the following two equations to the list of equations in the EEG Channel Operations GUI:
nch34 = ch31 - ch32 Label HEOG-bipolar
nch35 = ch33 - ch16 Label VEOG-bipolar
The new Channel 34 will be HEOG-right minus HEOG-left, and the new Channel 35 will be VEOG-lower minus FP2 (which is just above the right eye). The list of equations should look like that shown in Screenshot 5.8. Click RUN, and name the new dataset 6_N400_ref to indicate that it has now been referenced. You’ll want to refer to this dataset in the next exercise, so save it as a file if you’re not going to do the next exercise right away.
Now plot the data with EEGLAB > Plot > Channel data (scroll). The most obvious change is that the bipolar EOG channels are now present. You can see that blinks are larger in the VEOG-bipolar channel than in the VEOG-lower or FP2 channels. The N400 task used stimuli presented in the center of the monitor, so there aren’t any obvious horizontal eye movements. We'll see what those look like in the chapter on artifact rejection.
