8.15: Exercise- Visualizing the Eye Movements

Now that you know about the ERP CORE N2pc paradigm, let’s take a look at the eye movements. As you can imagine, it’s difficult to maintain gaze on the central fixation point and not look toward the target. But as described above, it’s important to make sure that the N2pc results aren’t impacted by eye movements toward the target, which will change the location of the target relative to the center of gaze and also create a negative voltage over the contralateral hemisphere.

In this exercise, we’ll focus on the data from Subject 15. If you look in the Chapter_8 folder, you’ll see two datasets from this participant, one containing the continuous data (15_N2pc_ICA_preprocessed.set) and one containing the epoched data (15_N2pc_ICA_preprocessed_epoched.set). To make this exercise simpler, ICA-based artifact correction was already applied to these datasets to eliminate blinks. However, the datasets include a VEOG-uncorrected channel that contains the original uncorrected VEOG-bipolar signal. We like to keep this uncorrected channel so that we can reject any trials on which the participant blinked during the time period of the stimulus and therefore could not see the stimulus. This particular participant never blinked during that time period, however, so we don’t need to worry about rejecting those epochs in this exercise. I also applied several other preprocessing operations, including filtering (bandpass 0.1–30 Hz, 12 dB/octave). The EEG channels and the HEOG-left and HEOG-right channels have all been referenced to the average of P9 and P10, and there is also an HEOG-bipolar channel (HEOG-left minus HEOG-right).

Before we get started on the exercise, reset everything with the Reset button, load the continuous dataset (15_N2pc_ICA_preprocessed.set) using the EEGsets panel, and scroll through the data. You should always scroll through the continuous data as the first step so that you know what’s in the file (as we did in the video demonstration in the previous chapter). Subject 15 has some weird stuff in the F7 channel near the end of the session, but we won’t worry about that for this exercise.

Now load the epoched dataset (15_N2pc_ICA_preprocessed_epoched.set), select only the EOG channels (29-33) in the Channel and IC Selection panel, and scroll through the data. This participant did a good job of following the fixation instructions initially, but you’ll start to see a fair number of eye movements beginning at Epoch 70 (see Screenshot 8.19, with the eye movement circled in Epoch 70).

If you look at the event codes and the polarity of the HEOG deflections, you’ll see that most of the clear eye movements are toward the side containing the target. Remember that the HEOG-bipolar channel was calculated as HEOG-left minus HEOG-right, and the dipole is positive at the front of the eyes, so a leftward eye movement produces a positive deflection and a rightward eye movement produces a negative deflection. For example, Epoch 70 has a negative deflection, indicating a rightward eye movement. You can see that the time-locking event code for this epoch is labeled “B2(121)”, indicating that the event code was 121 (attend-blue, target on the right, gap on the top) and was assigned to Bin 2 (right-side targets). This is important because it indicates that we had a rightward eye movement on a trial with a right-side target.

When participants try to maintain central fixation but occasionally fail, eye movements will typically consist of a rapid shift away from the fixation point, a period of stability at the new location for a few hundred milliseconds, and then a rapid shift back to the fixation point. This leads to a “boxcar” shape in the HEOG signal (a flat signal at one voltage level corresponding to the location of the fixation point, a sudden change to a different voltage level for 100–500 milliseconds, and then a return back to the original voltage level corresponding to the fixation point). You can see this pattern in Epoch 70. The voltage level changes suddenly at approximately 285 ms after the time-locking event and then returns to the original level approximately 180 ms later.

If you look at epoch 73, you’ll see a voltage deviation in the HEOG-bipolar channel, but it isn’t a saccadic eye movement. It doesn’t have the classic boxcar shape. Instead, it has the typical shape of a blink and matches the time course of the much larger deflection in the VEOG-bipolar channel. This indicates that the deflection in HEOG-bipolar is some blink voltage that has leaked through to the HEOG electrodes.

There’s a small leftward (positive) eye movement at approximately 570 ms after the time-locking event in Epoch 78, which had a left-side target. Epoch 81, with a right-side target, has a somewhat complicated pattern that looks like a brief leftward movement followed by a clearer rightward movement. Epoch 82, also with a right-side target, has a clear rightward eye movement.

There’s also a brief spike in voltage at the onset of the eye movement in Epoch 82 (circled in Screenshot 8.20). That’s probably the spike potential, which is caused by the muscle activity that produces the eye movement. Ordinarily, the muscle contraction that produces an eye movement occurs briefly at the start of the eye movement (to overcome the inertia in eye position), but then the muscle activity becomes too small to see as the eyes maintain their new location (which requires very little muscle activity). The present data have been low-pass filtered, so the spike potential isn’t very clear. Without the filtering, the spike potential is quite large in some participants. It’s fairly localized to frontal electrode sites, and it’s easy to filter out, so I don’t usually worry about it as an artifact. However, if you perform time-frequency analyses, it’s easy to mistake this EMG burst for gamma-band EEG activity (Yuval-Greenberg et al., 2008), so be cautious when someone says they’re seeing gamma-band EEG oscillations at frontal electrode sites.

If you keep scrolling through the data, you’ll see quite a few eye movements (mainly in the direction of the target, and mainly starting around 200 ms after stimulus onset). We clearly have a lot of work to do to make sure that these eye movements don’t confound our N2pc data!

---