Statistical Analysis within the masked cortical source space
First, the oscillatory ssVEF responses between both pre-cue conditions (attend the central fixation cross vs. attend the peripheral rings) were compared using a non-parametric cluster-based permutation approach (Nichols and Holmes 2002) restricted to the cortical sources that oscillated at the stimulus driven frequencies significantly different from noise (see Figure 1). From these sources the relative power of the 12 Hz and 15 Hz driving frequencies were extracted by dividing the power values of interest by the mean power across the noise power bins (31 frequency bins between 17 Hz and 20 Hz in 0.1 Hz steps). Then, within the cortical source mask (Figure 1) source clusters were formed when at least one adjacent source neighbor indicated a significant difference based on a dependent t-test (cluster alpha threshold of p = 0.025 testing for both directions). Then, the t-values were summed across the cluster. Thereafter, 1000 random permutations between both pre-cue conditions for each cortical source under the Null hypothesis of no differences between attending the central fixation cross or the peripheral rings were done. At each permutation step the same cluster rule was applied and the maximum t cluster sum entered a permutation distribution. Finally, the empirically observed t clusters obtained by the first step (see above) with a sum exceeding the 97.5 percentile (test in both directions) of the permutation distribution were considered as significant. In order to test possible interactions of cortical hemisphere, mean relative ssVEF power were extracted from the resulting significant clusters (see Results) and compared with the mean ssVEF activity in the homologous opposite hemisphere by using a repeated measures ANOVA with within-subject factors pre-cue condition (fixation cross and peripheral rings) and hemisphere (left and right). Greenhouse-Geisser corrections were applied when necessary.
However, the main objective of this study was to characterize how stimulus driven ssVEF responses are modulated by shifts of spatial attention when spatial attention had been already directed towards the peripheral or central visual field (rings vs. cross). Using the same non-parametric cluster-based mass-univariate approach, the interaction of relative ssVEF power changes between experimental phases (pre-cue baseline, post-cue hemifield not attended or attended) and pre-cue baseline conditions (attend central fixation cross or peripheral rings) was assessed for the left and right visual hemifield stimulations, separately. However, clusters were derived from a dependent F-test (equivalent to a repeated measures ANOVA) comparing the differences in relative ssVEF power between the pre-cue peripheral rings minus central fixation condition, post-cue visual hemifield not attended when rings were attended before minus visual hemifield not attended when central fixation cross was attended before, and post-cue visual hemifield attended when rings were attended before minus visual hemifield attended when central fixation cross was attended before. Note, that the comparison of these differences is equivalent to test the interaction experimental phase by pre-cue condition. Source clusters were formed when at least one adjacent source neighbor indicated a significant interaction based on a dependent F-test (cluster alpha threshold of p = 0.05 testing for one direction). F-values were summed across the cluster. Then, the same permutation approach as outlined above was applied. However, the empirically observed F clusters with a sum exceeding the 95 percentile (F test in only one direction) of the permutation distribution were considered as significant. All non-parametric cluster-based permutation tests were implemented using the fieldtrip toolbox (Oostenveld et al., 2011;https://www.fieldtriptoolbox.org).
In order to test any interactions with hemisphere, mean relative ssVEF power values of cortical source clusters of both hemispheres as obtained by the permutation statistics were entered in a repeated measures ANOVA with within-subject factors of pre-cue baseline, experimental phase and hemisphere. Greenhouse-Geisser corrections were applied when necessary.