Fig 1.
Example stimulus under each experimental condition of picture filtering (broadband / high spatial frequency / low spatial frequency) and gaze direction (direct / averted).
These avatar faces were created with FaceGen Modeller 3.5 (see [45] for details).
Fig 2.
Result of the additional behavioral experiment.
Using the same material, the same paradigm than the EEG experiment and the same number of participants (N = 15), we asked participants to answer as quickly as possible to the gender of each pictures. Critically, the direct gaze condition with broadband filtering is performed significantly faster than its averted counterpart.
Fig 3.
Grand mean ERPs from the pooled left and right occipito-temporal electrodes (highlighted on the cap at the top left) are shown for the broadband (top row), high spatial frequency (middle row), low spatial frequency (bottom row) faces with direct gaze (in black) and averted gaze (in red). Both P1 and N170 peak latencies were delayed with filtered pictures. The dashed lines represent the peak latency of the N170 under each picture filtering condition.
Fig 4.
Early differences between topographical ERP patterns in response to direct and averted gaze, in the broadband (BB) picture condition.
A) Overview of ERPs in response to direct and averted gaze BB pictures and sample-by-sample t-test of ERP differences between these two conditions. The upper graph represents the overlay of grand mean ERP waveforms obtained on every electrode in response to direct and averted gaze BB pictures. The lower graph represents the number of electrodes for which the p value of the exploratory t-test of ERP differences between direct and averted gaze conditions (performed on each electrode and each time point) reached significance (at an uncorrected threshold of p < .05). The time window delimited by a dashed green rectangle represents the 41–80 ms time window in which different topographical maps were identified for the direct and averted gaze BB picture conditions (see part C of this figure and text). On the right of the graph, a topographical map shows the electrodes that reached the .05 threshold p-value and a bar plot illustrates the early effect showing a scatterplot of the individual participants’ mean amplitude values between 41 and 80 ms for direct and averted gaze at the electrode of maximal t-value (represented by a black dot on topographical map). This showed that the early effect was present in every participant but one. B) Global topographic dissimilarity analysis. The graph represents the statistical level of the global dissimilarity index between the ERP scalp distributions for direct and averted gaze, in the BB picture condition. C) Micro-state segmentation. The five stable topographical maps obtained using the microstate segmentation procedure are represented above the Global Field Power (GFP) for direct and averted gaze BB pictures. The areas under the curves of the GFP are coloured to represent the period of time where each map was stable. Between 41 and 80 ms, distinct maps (maps 1 and 2) were identified for the direct and averted gaze conditions.
Fig 5.
Result of the fitting procedure for the two early maps (Map1 / Map 2) identified in the 41–80 ms time window.
There was a significant interaction between MAP and GAZE in the BB picture condition. The fitting procedure showed that the Map 1 explained more variance in the direct than in the averted gaze condition and that Map 2 explained better the averted gaze condition than Map 1 did. In addition, post-hoc t-test showed that Map 1 explained better the direct gaze than the averted gaze condition, t(14) = 3.33, p < .005, and that Map 2 explained better the averted gaze condition than Map 1 did, t(14) = 2.73, p < .016. Since normality condition was not verified for this explained variance measure, we ran additional non-parametrical Wilcoxon signed-rank tests; this confirmed the results reported with Zs < -2.04, ps < .05.
Fig 6.
Source localization in the 41–80 ms time window.
A) Result of the node-wise ANOVA performed on source estimation in the 41–80 ms time window. The nodes showing a significant interaction between FREQUENCIES and GAZE (p < .05 over a minimum of 8ms with a cluster size > 32 solution points) are represented in red over left and right lateral views and horizontal sections of the template brain. B) Bar plot of the estimated source activity over the left parietal cluster identified in A, under each condition of picture filtering (Broadband, HSF, LSF) and gaze direction (direct gaze in black, averted gaze in red), showing that the left parietal region was significantly more activated for direct than averted gaze in the broadband face condition.