Binocular microsaccades were added as “events” into the EEG data, and epochs locked to them (−0.5 to 0.5 seconds) were extracted, baseline corrected (−0.2 to 0 seconds), and manually screened for artifacts and noisy channels (on average, 70.1 ± 85.1 epochs per subject discarded; 7.4 ± 5.2 channels per subject discarded and interpolated based on the activity of surrounding channels). On average, 117.9 ± 76.3 epochs in the young group, 147.1 ± 62.2 epochs in the middle-aged group, and 137.9 ± 82.5 epochs in the elderly group were analyzed. We quantified two typical microsaccade-related potentials (i.e., MLR and SP). Potential latencies were defined in grand averaged ERPs for each of the three groups. Within a short window around it (20 ms for MLR; 10 ms for SP), individual peak potential amplitude and latency were located. The EEGLab function “newtimef” was used to compute ERSP and ITC across 25 linearly spaced frequencies ranging from 6 to 30 Hz and 200 linearly time points spanning −220 to 220 ms around microsaccade onset. Representative electrodes were selected for the occipital region (E75, E70, and E83), central region (E7, E31, E55, E80 and E106), and frontal region (E11, E19, and E4) (
Fig. 1). We subdivided the time range into eight small time windows (−220 to −150, −150 to −100, −100 to −50, −50 to 0, 0 to 50, 50 to 100, 100 to 150, and 150 to 220 ms), and the frequency range into six bands (alpha: 7–14 Hz, low alpha: 7–11 Hz, high alpha: 11–14 Hz, beta: 14–30 Hz, low beta: 14–18 Hz, and high beta: 18–30 Hz). For source estimation, the forward model and the inverse model were calculated with an open access software Brainstorm.
32 We calculated the forward model using the symmetric boundary element method (BEM)
33,34 and default Montreal Neurological Institute (MNI) magnetic resonance imaging (MRI) template.
35 The inverse model was estimated using the weighted minimum norm estimate (wMNE).
36 When computing the inverse model, (1) the source orientations were constrained to be normal to the cortical surface; (2) a depth weighting algorithm was used to compensate for any bias affecting the superficial sources calculation
37; and (3) a regularization parameter, λ2 = 0.1 was used to minimize numerical instability, reduce the sensitivity of the wMNE to noise, and effectively obtain a spatially smoothed solution.
36 In current density distribution visualization, a 1-Hz high-pass filter and 40-Hz low-pass filter were applied.