University of Central Florida Undergraduate Research Journal - Perception of Facial Expressions in Social Anxiety and Gaze Anxiety
US tab


The purpose of this study was to explore the relationship between gaze anxiety and facial recognition by selfreported socially anxious individuals. It was predicted that gaze anxiety would be related to poorer performance on a facial recognition task, and my results support this hypothesis. Gaze anxiety appears to be related to a socially anxious individual's ability to identify emotional intensity. Additionally, I found relationships between gaze anxiety and the identification accuracy of specific emotions.

Gaze Anxiety

When individuals with high trait social anxiety judged emotional intensity, they perceived images of faces as being more intense the more anxious they were about making eye-contact. The reported tendency to avoid eye-contact (a subscale of the GARS) accounted for the largest amount of variation in response. Past research has demonstrated that socially anxious individuals are hypervigilant towards negative emotions, and these data suggest that these biases are moderated by more specific fear. This might explain why some studies fail to replicate these biases or have small effect sizes.

The way these intensity ratings were derived should be noted. All of the images in the recognition task had either a mild expression or an extreme expression. These represented opposite ends of my 5-point intensity scale and there was no subtle gradation in between the two extremes. I expected average intensity ratings to be around 3, an assumption supported by the results. This result suggests that participants were generally interpreting the faces and the scales correctly. Average intensity ratings above the midpoint of the scale reflect a tendency to rate the mild images as more intense than they actually were. Given this result, my model represents a significant increase in perceived intensity. It was also an effect dependent on the amount of self-reported social anxiety.

While overall accuracy could not be predicted by gaze anxiety, these relationships change when separating emotion by type. Higher gaze anxiety was related to increased accuracy at identifying angry expressions and decreased accuracy at identifying neutral expressions. Similar findings for social anxiety are well documented, and it is typically attributed to hypervigilance. Socially anxious individuals are better at identifying angry faces but more likely to misjudge a neutral face as negative. Because of this, my findings could be due to changes in social anxiety. However, the relationship between GARS scores, neutral emotion accuracy, and angry emotion accuracy was strengthened when controlling for changes in social anxiety. This suggests that gaze anxiety has a significant relationship to the perceptual biases often experienced by socially anxious individuals.

Facial Recognition

In addition to gaze anxiety, I wanted to compare recognition accuracy when presented with the entire face to accuracy when presented with only the eyes. There is evidence that individuals might pay preferential attention to the eye region depending on a number of factors such as mood or social anxiety (Darby & Harris, 2010; Hills and Lewis, 2011; Gutierrez-Garcia & Calvo, 2014). Avoiding or fixating on one region does not necessarily mean that the rest of the face is visually ignored, but I was interested in measuring how much accuracy is lost when contextual information from the face was absent. When presented with only the eye region, participants were about 8-9% less accurate at identifying emotion. Nevertheless, participants were surprisingly accurate (71%) with only the eyes and the immediate areas around them. An even larger influence on recognition accuracy was gender. Participants were much better at identifying emotions in female faces than they were at identifying emotions in male faces—by about 14%. There were more females than males in our sample, but the difference between the genders of the participants was negligible.


Significant results were found despite several factors limiting the strength of my findings. The survey was conducted entirely online with a convenience sample of psychology students—most of which were females. Variance in the data could have been impacted by a lack of attentiveness or motivation on the participants' part. This is expected with online surveys. It is possible that my results would have been more robust if I could have controlled for these factors with a face to face session.

Another limitation was the survey software I used to collect data. I was unable to limit the amount of time participants spent on each facial image. Past research on attention, perception, and social anxiety has been conducted by only briefly exposing participants to facial stimuli. In my questionnaire, by contrast, partiipants could have lingered on each image. I could control for this afterwards by excluding outliers based on completion time, but I don't know how long participants fixated on each particular image. In addition, the GARS is a newer form with less validation than the other self-reports used in the study. There have been a couple large studies to support the scale's use in college populations (Langer et al., 2014; Schneier et al., 2011), but an objective measurement would have been better. Using an eyetracker in combination with the self-report would have provided a metric for a participant's perceived anxiety and actual avoidance.