Chapter IV: Task-dependent Modulation Of Face Gaze
4.2 Don’t Look Study in Autism
4.2.3 Results
Finally, for the analysis of center bias, we analyzed the two ROIs defining the socially salient regions of the face (Eye Region Total and Mouth). For the Eye and Mouth ROIs, we calculated the Euclidean distance for each point within the ROI to the vertical and horizontal midlines of the ROI, weighting each distance by gaze duration (i.e., the heatmap value for the point). Next, we summed these values, and then divided by the sum of gaze in the ROI to obtain the average gaze distance to the horizontal and vertical midlines.
Statistical analyses were conducted by carrying out repeated-measures ANOVAs for each ROI in the analysis subset, with a between-subjects factor of group (ASD, Control) and within- subjects factors of condition (Free View, Avoid Eyes, Avoid Mouth) and stimulus type (Open Eyes, Closed Eyes). In order to investigate the relationship of the two restricted conditions (Avoid Eyes, Avoid Mouth) against the unrestricted condition (Free View), appropriate contrasts were also conducted for gaze to the Right Eye, Left Eye, Eye Region Total, and Mouth. Lastly, we calculated Pearson correlations between AQ and proportion of gaze to ROIs across all subjects, and ADOS-CSS-Overall scores and proportion of gaze to ROIs for the autism subjects only.
A Control Autism
Op en e ye s Cl os ed e ye s
B
Op en e ye s Cl os ed e ye s
C
Op en e ye s Cl os ed e ye s
Figure 4.5. Heat maps for the Autism and Control groups, and condition with Open Eyes and Closed Eyes stimuli. Maps were generated by using a Gaussian kernel
FREE
function to spatially smooth each duration-weighted fixation point, then maps were averaged across trials and subjects. Warmer colors represent longer total fixation time.
4.2.3.1 Gaze proportions in all regions of interest
The first set of analyses focused on gaze to the different regions of interest (see Figure 4.4-A for ROIs). A repeated-measures ANOVA with between-subjects factor of group and within- subjects factors of condition and stimulus was conducted for each of the seven ROIs (see Figure 4.6).
Figure 4.6. Proportion of fixation time in the ROIs in the ASD (red) and control (blue) groups. Error bars denote standard error.
Right Eye
The ANOVA for the right eye revealed a significant main effect of group on gaze time, F(1,23)
= 5.56, p = .027, η2 = .195. Overall, ASD looked less at the right eye than Controls regardless of stimulus type or condition (ASD: M = 7.7%, SE = 1.1; Controls: M = 11.4%, SE = 1.1).
While the interaction between group and stimulus did not reach significance, F(1,23) = 2.77, p
Open Eyes Closed Eyes 0
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Off−Head
Open Eyes Closed Eyes 0
0.2 0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type Head Remainder
Open Eyes Closed Eyes 0
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Nose
Open Eyes Closed Eyes 0
0.2 0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type Eye Region Total
Open Eyes Closed Eyes 0
0.05 0.1 0.15 0.2
Proportion of Time (%)
Stimulus Type Left Eye
Open Eyes Closed Eyes 0
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Right Eye
Open Eyes Closed Eyes 0
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Mouth
Open Eyes Closed Eyes 0
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.2
0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.1
0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.2
0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.05
0.1 0.15 0.2
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.1
0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.1
0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type
Open Eyes Closed Eyes 0
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.2
0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.1
0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.2
0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.05
0.1 0.15 0.2
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.1
0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0Open Eyes Closed Eyes 0.1
0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Control Autism
Free View
Avoid Eyes
Avoid Mouth
FREE
= .109, η2 = .108, the effect size indicates there is likely a greater sensitivity of stimulus type on Controls than ASD, such that the decrease in gaze between open and closed eyes is larger in the Control group than the ASD group (ASD: 2.1%; Controls: 4.0%).
There was also a marginally significant interaction between condition and group, F(1.59,36.61)
= 2.86, p = .081, η2 = .111. A planned contrast comparing the unrestricted condition to Avoid Eyes and Avoid Mouth indicated there was a significant interaction between condition and group for Avoid Eyes – Free View, F(1,23) = 10.78, p = .003, η2 = .319, but not for Avoid Mouth – Free View, F(1,23) = 0.42, p = .521, η2 = .018. While Controls looked more to the right eye than ASD in the Free View condition (ASD: M = 9.3%, SE = 1.6; Controls: M = 15.5%, SE = 1.5; t(23) = 3.05, p = .006), there was no group difference in “Avoid Eyes,” (p = .992).
Left Eye
The ANOVA for the left eye revealed no significant interactions involving group (all ps >
.627), nor a main effect of group (p = .228).
Eye Region Total
The ANOVA for the total eye region revealed no significant interactions involving group (all ps > .309), and no main effect of group (p = .783).
Mouth
The ANOVA for the mouth region indicated there was a marginally significant interaction between group and condition, F(1.08,25.03) = 3.58, p = .067, η2 = .135. There was also a marginally significant interaction between group and stimulus, F(1,23) = 3.36, p = .080, η2 = .127, suggesting a greater influence of stimulus type on gaze to the mouth in Controls than ASD . There was not a significant main effect of group (p = .473).
Contrasts comparing the unrestricted condition to Avoid Eyes and Avoid Mouth revealed a significant interaction between group and condition for both Avoid Eyes – Free View, F(1,23)
= 4.90, p = .037, η2 = .176, and Avoid Mouth – Free View, F(1,23) = 5.93, p = .023, η2 = .205.
Both effects were driven by a significant group difference in the Free View condition, with
ASD looking less at the mouth than Controls (ASD: M = 4.5%, SE = 1.0; Controls: M = 9.6%, SE = 1.0; t(23) = 3.05, p = .006), but no group differences in either of the other two conditions (ps > .153). In other words, while ASD looked less at the mouth in Free View than Controls, both groups significantly increased gaze to the mouth in the Avoid Eyes condition and decreased gaze to the mouth in the Avoid Mouth condition relative to Free View.
Nose
The ANOVA for gaze to the nose revealed a marginally significant interaction between group and condition F(2,46) = 3.19, p = .074, η2 = .122. There were no other significant interactions involving group (all ps > .381) and no main effect of group (p = .871).
The contrast comparing Free View to Avoid Mouth revealed a significant interaction between group and condition, F(1,23) = 5.41, p = .029, η2 = .190, indicating that the ASD group looked more at the nose than Controls in Free View, (ASD: M = 18.7%, SE = 4.1; Controls: M = 12.7%, SE = 1.7) but less than Controls in Avoid Mouth (ASD: M = 5.6%, SE = 1.1; Controls:
M = 8.4%, SE = 2.0). Both Controls and ASD also decreased gaze to the nose region in the Avoid Mouth condition compared to Free View, but that decrease was greater in ASD.
Head Remainder
The ANOVA for the head remainder revealed no significant interaction effects involving group (all ps > .327), and no main effect of group (p = .343).
Off-Head
The ANOVA for gaze off-head revealed no significant interaction effects involving group (all ps > .438), and no main effect of group (p = .934).
4.2.3.2 AQ and ADOS correlations with gaze to ROIs AQ Correlations Across Groups
In the Free View condition, AQ was negatively correlated with gaze to the right eye for Open Eyes stimuli (r = -.459, n = 25, p = .021), and gaze to the mouth for Closed Eyes stimuli (r = - .463, n = 25, p = .020). Also, in Free View there was a negative correlation with gaze to the
Head Remainder for Closed Eyes stimuli (r = -.401, n = 25, p = .047). In the Avoid Eyes condition, AQ was positively correlated with gaze to the left eye for Open Eyes stimuli (r = .460, n = 25 p = .021). No correlations were significant in the Avoid Mouth condition.
ADOS CSS-Overall Correlations in Autism Group
In the Avoid Mouth condition, there was a negative correlation between ADOS severity scores and gaze to the right eye for Open Eyes stimuli (r = -.681, n = 11, p = .021), as well as a correlation for gaze off-head, for both Open Eyes stimuli (r = .670, n = 11, p = .024), and Closed Eyes stimuli (r = .664, n = 11, p = .026). Also in Avoid Mouth, there was a positive correlation for gaze to the head remainder for Open Eyes stimuli, (r = .615, n = 11, p = .044), and a negative correlation with gaze to the nose for Closed Eyes stimuli (r = -.679, n = 11, p = .022). None of the remaining correlations reached significance in any of the conditions.
4.2.3.3 Gaze distribution in eye region
To examine the distribution of gaze in the eye region, we calculated the proportion of gaze time in each eye ROI relative to the total gaze time spent in the entire eye region (see Figure 4.4-B for ROIs). Results are summarized in Figure 4.7.
A Control Autism
B
Figure 4.7. (A) Heatmaps and gaze proportions in the eye-related ROIs in the ASD (red) and control (blue) groups collapsed across condition and stimulus type. (B) Gaze
Open Eyes Closed Eyes
0 0.2 0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type Eye Region Remainder/Total Eye Region
Open Eyes Closed Eyes
0 0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Left Eye/Total Eye Region
Open Eyes Closed Eyes
0 0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Right Eye/Total Eye Region
Open Eyes Closed Eyes
0 0.2 0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type 0 Open Eyes Closed Eyes
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0 Open Eyes Closed Eyes
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type
Open Eyes Closed Eyes
0 0.2 0.4 0.6 0.8 1
Proportion of Time (%)
Stimulus Type 0 Open Eyes Closed Eyes
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type 0 Open Eyes Closed Eyes
0.1 0.2 0.3 0.4
Proportion of Time (%)
Stimulus Type Control Autism
Free View
Avoid Eyes
Avoid Mouth
FREE
proportions in the eye-related ROIs in the ASD (red) and control (blue) groups divided by condition and stimulus type. Error bars denote standard error.
Right Eye
The ANOVA for the right eye revealed a marginally significant interaction between group and condition, F(2,22) = 3.09, p = .066, η2 = .219. There was also a significant main effect of group, F(1,23) = 9.07, p = .006, η2 = .283, indicating the ASD group spent a significantly smaller proportion of gaze to the eye region fixating on the right eye relative to Controls, regardless of condition (ASD: M = 18.0%, SE = 2.5; Controls: M = 28.4%, SE = 2.4).
Planned contrasts comparing the unrestricted condition to the two restricted conditions revealed that there was a significant interaction between group and condition in Avoid Eyes – Free View, F(1,23) = 6.45, p = .018, η2 = .219, indicating the ASD group spent a smaller proportion of time in the eye region looking at the right eye in Free View (ASD: M = 18.2%, SE = 2.8; Controls: M = 33.6%, SE = 1.9; t(23) = 3.52, p < .001) but not in Avoid Eyes, p
= .507.
Left Eye
The ANOVA for the left eye revealed no significant interactions involving group (all ps > .448) and no main effect of group (p = .927).
Eye Region Remainder
The ANOVA for the remainder of the eye region revealed no significant interactions involving group (all ps > .168). There was, however, a significant main effect of group, F(1,23) = 5.54, p
= .027, η2 = .194, indicating that the ASD group spent a greater proportion of their gaze in the eye region fixating between the eyes (ASD: M = 64.9%, SE = 3.1; Controls: M = 54.8%, SE = 3.0).
4.2.3.4 Center bias to horizontal and vertical midlines
To compare average gaze distance (measured in degrees of visual angle) from the vertical and horizontal midlines in the socially salient regions of the face, we conducted two repeated- measures ANOVAs for the Eye Region and Mouth ROIs (see Figure 4.4-C for ROIs).
Eye Region
The ANOVA for the eye region revealed a main effect of group for distance to the vertical midline, F(1,23) = 5.89, p = .024, η2 = .204, indicating the average horizontal distance to the midline was significantly less in the ASD group relative to the Control group (ASD: M = 1.49 dva, SE = .08; Controls: M = 1.75 dva, SE = .07). Center bias for gaze to the vertical midline in the eye region is shown in Figure 4.8.
There were no significant interactions involving group for distance to the horizontal midline (all ps > .497). There was, however, a marginally significant effect of group, F(1,23) = 3.32, p
= .081, η2 = .126, for greater vertical distance to the midline in the ASD group (ASD: M = .84 dva, SE = .03; Controls: M = .77 dva, SE = .03).
Figure 4.8. Mean distance to the vertical midline in the Eye Region, collapsed across stimulus type and condition. Distance is measured in degrees of visual angle. Dots represent individual subjects and solid lines denote group mean.
Mouth
The ANOVA for the mouth did not reveal any significant interactions involving group for distance to the vertical midline (all ps > .291), nor a main effect of group (p = .799). There were also no significant interactions involving group for distance to the horizontal midline (all ps >
.737), nor a main effect of group (p = .970).