Impact of face outline, parafoveal feature number and feature type on early face perception in a gaze-contingent paradigm: A mass-univariate re-analysis of ERP data

Abstract

Recent ERP research using a gaze-contingent paradigm suggests the face-sensitive N170 component is modulated by the presence of a face outline, the number of parafoveal facial features, and the type of feature in parafovea (Parkington and Itier, 2019). The present study re-analyzed these data using robust mass univariate statistics available through the LIMO toolbox, allowing the examination of the ERP signal across all electrodes and time points. We replicated the finding that the presence of a face outline significantly reduced ERP latencies and amplitudes, suggesting it is an important cue to the prototypical face template. However, we found that this effect began around 114 ms, and was maximal during the P1-N170 and N170-P2 intervals. The number of features present in parafovea also impacted the entire waveform, with systematic reductions in amplitude and latency as the number of features increased. This effect was maximal around 120 ms during the P1-N170 interval and around 170 ms between the N170 and P2. The ERP response was also modulated by feature type; contrary to previous findings this effect was maximal around 200 ms and the P2 peak. Although we provide partial replication of the previous results on the N170, the effects were more temporally distributed in the present analysis. These effects were generally maximal before and after the N170 and were the weakest at the N170 peak itself. This re-analysis demonstrates that classical ERP analysis can obscure important aspects of face processing beyond the N170 peak, and that tools like mass univariate statistics are needed to shed light on the whole time-course of face processing.

Keywords: Event-related potentials; Eye tracking; Face processing; Facial features; Holistic processing; Mass univariate analysis.

Fig. 1

A. Examples of stimuli from each condition presented with left and right eye fixation. Note that directions are referenced from an observer's perspective (left eye means the eye on the left of the face; right eye means the eye on the right of the face). Red box: Conditions analyzed in Analysis 1 to assess the effect of face outline. Blue box: Conditions with one parafoveal feature, compared in Analysis 3. Yellow box: Conditions with two parafoveal features, compared in Analysis 4. In Analysis 2, the effect of the number of parafoveal feature was assessed by comparing the One eye condition (0 parafoveal feature), an aggregate of the blue box conditions (1 parafoveal feature), an aggregate of the yellow box conditions (2 parafoveal features) and the Full face condition (3 parafoveal features). B. Sample trial progression showing a full-face stimulus with left eye fixation (so the face was presented such that its left eye was in the center of the screen and in fovea). Each trial began with a central fixation cross. Participants were required to maintain fixation on the cross for 250 ms, which initiated stimulus presentation for 250 ms. Average time taken to activate the trigger was 481 ms ± 100 ms. If 10s elapsed without trigger activation, the trial was aborted and a drift correction was implemented. A response screen was presented following stimulus presentation until a response was made (for oddball trials) or 700 ms elapsed. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 2

A. Main effects of Eye Fixation, Category and their interaction for Analysis 1 (testing for the effect of face outline) with α = 0.05 and TFCE correction applied. Each raster plot depicts the significant F-values of the various effects at every electrode and time-point analyzed (magnitude of the F-values according to the coloured scale). Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered after TFCE correction (see text for details). B. ERP and amplitude difference plots illustrating the main effect of Category on two relevant electrodes. Note the large shift in latency between the Isolated Eye and the One Eye categories highlighted with the black lines (∼35 ms at F1 and 33 ms at P10), which drives the large amplitude differences seen with the difference waves. The red bar at the top of the difference wave plots represents the time points at which the main effect of Category was significant for these two electrodes. Note that the confidence intervals for the difference waves are very restricted, as would be expected for highly significant effects, appearing as a single blurred line on relevant electrodes. For all ERP plots, the ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3

A. Contrast raster plots for Analysis 1 using a Bonferroni-corrected α = 0.025 with TFCE correction applied. Each plot depicts the significant F-values corresponding to the effect of eye fixation analyzed separately for the Isolated Eye (left panel) and the One Eye conditions (right panel). The topographic map of the unthresholded and uncorrected F-values is displayed at the time where maximum values were registered after TFCE correction (see text for details). B. ERP and amplitude difference plots illustrating the lack of eye fixation effect for the Isolated Eye category (blue lines) and a small but clear effect of eye fixation for the One Eye category (red lines) on relevant electrodes. The red significance bar for the difference wave plots represents the time points at which the Eye Fixation by Category interaction was significant, for these electrodes. For all ERP plots, the ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4

A. The main effect of Eye Fixation from Analysis 2 (testing the effect of parafoveal feature number) with α = 0.05 and TFCE correction applied. The raster plot depicts the significant F-values at every electrode and time-point analyzed (F-value magnitude according to the coloured scale). Topographic maps of the unthresholded and uncorrected F-values are displayed at the time where maximum values were registered after TFCE correction (see text for details). B. ERP and amplitude difference plots comparing left and right eye fixation conditions on relevant electrodes. The red significance bar at the top of the difference wave plots represents the time points at which the main effect of Eye Fixation was significant for these electrodes. ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 5

A. The main effect of Parafoveal Feature Number from Analysis 2, with α = 0.05 and TFCE correction applied. The raster plot depicts the significant F-values corresponding to the differences between stimuli with 0, 1, 2, or 3 features within parafovea at every electrode and time-point analyzed (F-value magnitude according to the coloured scale). Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered after TFCE correction (see text for details). B. ERP and amplitude difference plots comparing stimuli with 0, 1, 2, and 3 parafoveal features on relevant electrodes, with each difference wave corresponding to a specific contrast comparing stimuli with different numbers of parafoveal features according to the legend on the right. ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). F-value plots depict the time course of each F-contrast comparing conditions with different numbers of parafoveal features, at the same electrodes as those displaying the ERPs (note that contrasts were Bonferroni-corrected and were run using α = 0.008). The red significance bar at the top of each difference wave and F-value plot represents the time points at which the main effect of Parafoveal Feature Number was significant for these electrodes. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6

A. The interaction between the effects of Eye Fixation and Parafoveal Feature Number from Analysis 2 with and α = 0.05 and TFCE correction applied. The raster plot depicts all electrodes and time-points at which a significant interaction between these factors occurred (F-value magnitude according to the coloured scale). Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered after TFCE correction (see text for details). B. ERP plots comparing the amplitude and latency of the ERP response to the Parafoveal Feature Number conditions separately for the left (top plots) and right (bottom plots) Eye Fixation conditions. ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). The red significance bar at the top of each difference wave represents the time points at which the interaction between Eye Fixation and Parafoveal Feature Number was significant for these electrodes. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 7

A. Main effects of Eye fixation from Analysis 3 (faces with 1 parafoveal feature) with α = 0.05 and TFCE correction applied. The raster plot depicts the significant F-values obtained at every electrode and time-point analyzed (magnitude according to the coloured scale). Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered after TFCE correction (see text for details). B. ERP plots comparing left and right eye fixation conditions on relevant electrodes, with associated difference wave. The red significance bar at the top of each difference wave plot represents the time points at which the main effect of Eye Fixation was significant for each electrode. ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 8

A. Main effects of Parafoveal feature type from Analysis 3 (faces with 1 parafoveal feature) with α = 0.05 and TFCE correction applied. The raster plot depicts the significant F-values obtained at every electrode and time-point analyzed (magnitude according to the coloured scale). Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered after TFCE correction (see text for details). B. Post-hoc contrasts comparing the various parafoveal feature type conditions (collapsed across left and right eye fixations), Bonferroni-corrected with α = 0.016 and TFCE correction applied. Each raster plot depicts the significant F-values obtained for each contrast, with magnitude according to the coloured scale. Left panel: faces with an eye in parafovea compared to faces with the nose in parafovea. Right panel: faces with an eye in parafovea versus faces with the mouth in parafovea. No significant differences were found between faces with a nose in parafovea versus faces with a mouth in parafovea so this contrast is not shown. Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered in each contrast after TFCE correction (see text for details). C. ERP waves of each parafoveal feature condition (eye, nose or mouth in parafovea) at relevant electrodes. Amplitude difference plots depict the three contrasts (comparing two conditions at a time). The red significance bar at the top of each difference wave plot represents the time points at which the main effect of Parafoveal Feature Type was significant for each electrode. All ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 9

A. Main effects of Eye fixation, Parafoveal feature type and their interaction from Analysis 4 (faces with 2 parafoveal features) with α = 0.05 and TFCE correction applied. Each raster plot depicts the significant F-values corresponding to the differences between levels of each factor at every electrode and time-point analyzed (magnitude of the F-values according to the coloured scale). Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered after TFCE correction (see text for details). B. ERP plots comparing left and right eye fixation conditions on relevant electrodes, with associated difference wave. The red significance bar at the top of each difference wave represents the time points at which the main effect of Eye Fixation was significant for these electrodes. All ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 10

A. Contrast raster plots from Analysis 4 (faces with 2 parafoveal features) with α = 0.016 and TFCE correction applied. Each raster plot depicts the significant F-values corresponding to differences between the various parafoveal feature type conditions (collapsed across left and right eye fixations). Left panel: faces with an eye and a nose in parafovea compared to faces with the nose and mouth in parafovea. Middle panel: faces with an eye and a nose in parafovea compared to faces with an eye and the mouth in parafovea. Right panel: faces with the nose and mouth in parafovea versus faces with an eye and the mouth in parafovea. Topographic maps of the unthresholded and uncorrected F-values are displayed at time points where maximum values were registered in each contrast after TFCE correction (see text for details). B. ERP waves of each parafoveal feature condition (eye-nose, nose-mouth or eye-mouth in parafovea) at relevant electrodes. Amplitude difference plots depict the three contrasts (comparing two conditions at a time). The red significance bar at the top of each difference wave represents the time points at which the main effect of Parafoveal Feature Type was significant for these electrodes. All ERPs represent the central tendency computed using raw means for single subject trials and 20% trimmed means across subjects; the difference waves were created with a 95% Bayesian Highest Density Interval (HDI) confidence interval (CI). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)