Static estimates of latency variability

To estimate relative latency, we first calculated template waveforms separately for each subject in each ROI (the calcarine fissure and fusiform gyrus bilaterally). The template waveform was obtained by averaging dSPM source waveforms across trials and across vertices within a given ROI. To cope with the low signal-to-noise ratio from single-epoch waveform, we estimated evoked responses using a bootstrap approach. Specifically, we obtained one bootstrap sample by randomly selecting 30 single-trial source estimates (approximately 15% of the trials) and averaging over these trials. The bootstrap procedure was repeated 100 times for each condition, ROI, and subject. These responses were then low-pass filtered at 40 Hz.

We also compared noise levels across ROIs. Noise levels were estimated by the standard deviation over the pre-stimulus or baseline period (−200 to 0 ms) and averaged across bootstrap samples and across subjects. To avoid potential confounds caused by regional differences in noise levels, we equalized noise levels across regions by adding Gaussian random noise (mean = 0 and standard deviation = differences in noise levels between regions) to the bootstrap samples across time.

Then, we calculated cross-correlations between the template waveform and each bootstrap sample in each ROI. The template waveform was temporally shifted (−200 and +200 ms in steps of 10 ms). The shift corresponding to the maximum correlation coefficient was defined as the relative latency. With 100 relative latencies across bootstrap samples, we calculated intra-subject latency variability or the standard deviation of relative latencies across bootstrap samples. Intra-subject latency variability in the calcarine fissure and fusiform gyrus was compared using Wilcoxon signed-rank test.

To further understand the relationship between the static estimates of latency variability and the traditional measures of peak latency, Pearson’s correlation coefficients were calculated between the static estimates of latency variability and the peak latency obtained using the traditional method as described in the previous section.