Perceiving faces in variability: an eye-tracking study in infants

• In Event: 2-104 - Infant Attention and Face Learning: Insights from Measures of Brain and Behavior

Integrative Statement

Being able to learn and recognize faces despite a wide range of natural variation in appearance is critical to infants’ social interaction and their development of social cognition. Variations in facial characteristics include changes in expressions, make­up, viewing angles and changes that occur with aging and health conditions (e.g., stress, weight gain, or tiredness; Coetzee et al., 2009). Recent studies examining adult face perception suggest that variations in appearance can have a large influence on the facial configuration (e.g., spacing among facial features) and distinctiveness of a given identity (Hancock et al., 2000), creating a challenge for accurate face recognition (Burton, 2013). Many past studies investigating infants’ face recognition used highly controlled stimuli, precluding our understanding of how infants develop the ability to form stable representations of faces across variability in appearance. This investigation used infant eye-tracking methods to examine the attentional mechanisms underlying infants’ perception of faces that vary in appearance.

Six to 12-month-old infants (N=14) were tested in an eye-tracking procedure. Infants viewed two muted videos of two female models reading a storybook. One of the videos contained high variability in the model’s appearance (different filming locations, lighting conditions, viewing angles, hairstyle, make-up); the other video contained only minor changes in the model’s appearance (same filming locations etc.). Eight female models matched in physical similarity were selected for the study, resulting in four face pairs that were randomly assigned to the high or low variability condition. After each video, infants were shown a static face pair comprising a novel image of the learned identity and an image of a novel identity.

Analyses of looking behavior during presentation of the static face pairs suggests that infants show comparable fixation counts and dwell time for learned and novel faces and for faces learned from high and low variability video (ps > 0.1). However, we found a significant interaction between novelty (learned vs. novel faces) and learning order (High vs. Low Variability video first) on dwell time, F1,12= 11.82, p = .005, η2 = 0.50. Follow-up paired sample t-tests suggest that when presented with the low variability video first, infants fixate novel faces longer than learned faces (p= .008); in contrast, when presented with the high variability video first, infants fixate comparably long for learned and novel faces (p= .338). The difference in the dwell time for learned and novel faces likely reflects the difference in the way in which a face is initially perceived/learned. We argue that exposure to natural variation in appearance likely influences infants’ attention allocation when perceiving and learning faces, which then influences their later looking to the learned versus novel face. Recent studies investigating face learning in adults and children suggest that exposure to high variability in appearance facilitates face learning (Baker et al., 2016; Zhou et al., 2018). In line with these studies, our study highlights that visual exposure to natural variability in people’s appearance is critical to infants’ attention to and learning of newly encountered faces.

Authors

• Xiaomei Zhou, Ryerson University
  Presenting Author

• Shruti Vyas, Ryerson University
  Non-Presenting Author

• Margaret Moulson, Ryerson University
  Non-Presenting Author