In an N2pc (or CDA or LRP) experiment, we’re mainly concerned about systematic eye movements in the direction of the target, not random eye movements. A good way to assess these systematic eye movements is to make one averaged HEOG waveform for all the left-target trials and another averaged waveform for all the right-target trials. That’s what we’ll do in this exercise, before we work on rejecting trials with artifacts.
To begin the exercise, make sure that 15_N2pc_ICA_preprocessed is still loaded and active. Select EEGLAB > ERPLAB > Compute averaged ERPs, and run it with the default settings. You can name the resulting ERPset 15_N2pc_no_rejection. Select EEGLAB > ERPLAB > Plot ERPs > Plot ERP waveforms to plot the data, but telling the routine to plot only Channel 32 (HEOG-bipolar). It should look something like Screenshot 8.7.A.
Screenshot 8.7
Starting just after 200 ms, you can see a positive voltage deviation (leftward eye movement) for the left-target trials and a negative voltage deviation (rightward eye movement) for the right-target trials. (There are also small differences between left- and right-target trials at earlier latencies, but they must be random noise because it takes at least 150 ms for the visual system to find a color-defined target in a visual search array and execute a saccade to it.)
The negative voltage in the right-target average is larger than the positive voltage for the left-target average. This is a fairly common pattern, and it may be related to the widely observed right hemifield bias for spatial attention. The main issue, however, is the overall difference in HEOG activity between left-target trials and right-target trials. To visualize this, we can make a left-target-minus-right-target difference wave. Select EEGLAB > ERPLAB > ERP Operations > ERP Bin Operations, and create an equation that reads bin3 = bin1 - bin2 label Left Target Minus Right Target. Make sure that the Mode is set to Modify existing ERPset, and then click RUN.
If you plot the resulting difference wave, you’ll see that the difference between the left- and right-target averages is approximately 8 µV between approximately 250 and 450 ms (see Screenshot 8.7B). However, keep in mind that this reflects a mixture of some trials with a large eye movement and some trials without.
Is this a large enough difference to be problematic? To answer this question, we need to think about why horizontal eye movements are problematic for interpreting an N2pc experiment (or a CDA experiment, LRP experiment, etc.). First, we need to ask whether our N2pc effect might actually be HEOG voltage that has propagated from the eyes to the posterior electrodes where the N2pc is ordinarily observed. Fortunately, a great paper by Lins et al. (1993) provides propagation factors for blinks, vertical eye movements, and horizontal eye movements. They don’t provide values for the PO7 and PO8 electrodes, but they do provide values for the O1 and O2 electrodes and the P5 and P6 electrodes. The PO7 and PO8 electrodes that we used to measure N2pc amplitude in the ERP CORE paper are halfway between O1/P5 and O2/P6. Lins et al. reported a propagation of 1% from HEOG-bipolar to O1 and O2 and a propagation of 3% from HEOG-bipolar to P5 and P6. If we split the difference and assume a 2% propagation to PO7 and PO8, the 8 µV voltage deflection we observed at the HEOG channel in this participant would be expected to lead to a voltage of 0.16 µV at P5 and P6. That’s quite a bit smaller than the N2pc, which is typically 1-2 µV, but still as big as some between-condition differences in N2pc amplitude. And in many participants, the HEOG-bipolar voltage would be quite a bit larger.
We also need to consider whether the eye movements caused the target to shift toward the center of the retina, reducing the lateralization that is necessary for seeing an N2pc (because it is defined by the difference between contralateral and ipsilateral channels). Again, the paper by Lins et al. (1993) provides useful information. Specifically, they found that the HEOG-bipolar signal increases by 16 µV for every degree of lateral eye rotation (which my lab has confirmed several times). Because the difference between left-target and right-target trials in the average HEOG-bipolar signal was approximately 8 µV, we can conclude that there was a 0.5° difference in eye rotation between left-target and right-target trials, on average. However, this is just an average. It’s quite plausible that the target was fully foveated by ~250 ms on a substantial proportion of trials, with little or no eye movement on other trials. On the subset of trials with large eye movements, we can’t be certain that a target on the left side of the video display was actually in the left visual field and that a target on the right side of the video display was actually in the right visual field.
If we want to be careful, we need to both decrease the artifactual EOG voltage produced by the eye movements and eliminate trials with large deviations in eye position. Artifact correction can be used to reduce the artifactual EOG voltage (as described in the next chapter), but we need to reject trials with large eye movements to deal with large deviations in eye position.
