Remote, tablet-based assessment of gaze following: a nationwide infant twin study

Frederick Shic, Kelsey Jackson Dommer, Jessica Benton, Beibin Li, James C. Snider, Pär Nyström, Terje Falck‐YtterFrontiers in Psychology 2023 · DOI

Abstract: Together, these results highlight the potential of remote assessment of infants' psychological development, which can improve generalizability, inclusion, and scalability in developmental research.

Introduction

This study, published in Frontiers in Psychology, represents a significant advancement in the "democratization" of developmental science. By transitioning infant gaze-tracking from specialized laboratory environments to consumer-grade tablets, researchers have demonstrated a scalable, remote methodology for investigating early cognitive development and its underlying genetic factors.

The Challenge of Traditional "Babylabs"

Historically, our understanding of infant psychology has been predominantly shaped by data from "Babylabs"—specialized facilities typically located within major urban university centers. This centralized model presents two significant challenges. First, it introduces a selection bias, as participants are frequently drawn from high-socioeconomic, urban "WEIRD" (Western, Educated, Industrialized, Rich, and Democratic) populations. Second, the sterile, unfamiliar environment of a laboratory can induce stress in infants, potentially masking natural behaviors and reducing ecological validity.

Methods: The KiiTs App and Remote Testing

To address these issues, the research team developed KiiTs (Karolinska Infant iPad Twin study), a custom iOS application. Parents across Sweden used their own iPads to facilitate the study at home. The app presented 32 trials featuring an actress looking toward one of two toys.

Example trial flow showing the Eyes and Head condition versus the Eyes only condition.

The experimental design manipulated two key variables: cue type (Eyes Only vs. Eyes + Head movement) and auditory cue (a social "hey hey" vs. a non-social bell sound). The iPad's front-facing camera recorded the infant’s face at 30 frames per second, precisely time-locked to the stimuli.

Data Processing and Coding

Because the study utilized standard consumer hardware rather than specialized infrared eye-trackers, the researchers employed a rigorous manual coding process. Trained coders, blind to the experimental conditions, performed frame-by-frame analysis of the video files to determine the infant's gaze direction.

Flowchart of the manual coding process used to filter trials and derive reaction time metrics.

This pipeline ensured that only trials where the infant was clearly attending to the cue were included in the final analysis. Of the 3,296 trials delivered, 77.5% were retained—a quality rate comparable to high-end laboratory eye-tracking studies (typically around 78.8%).

Key Results and Heritability

The study successfully replicated classic laboratory findings in a home setting. The "Head + Eyes" condition elicited significantly higher gaze-following rates (60.2%) compared to the "Eyes Only" condition (51.5%, $p < 0.001$). Furthermore, developmental trends were evident: older infants (>7.5 months) exhibited a 58.4% hit rate, compared to 53.4% in younger infants.

A primary goal was to explore the genetic basis of social attention using a twin cohort (104 infants, including both monozygotic and dizygotic twins). While the accuracy of gaze-following did not show strong evidence of heritability, the latency to disengage—the speed at which an infant looks away from the face toward the target—showed a significant genetic signal. Monozygotic (identical) twins demonstrated a strong correlation in reaction time ($ICC = 0.621, p < 0.001$), whereas dizygotic (fraternal) twins showed a weak, non-significant correlation ($ICC = 0.218$). This suggests that the speed of oculomotor processing may be more biologically "hard-wired" than the social inclination to follow a gaze.

Discussion and Future Directions

The remote nature of the study allowed researchers to reach a more diverse demographic, including families with lower average salaries than those typically represented in Swedish lab studies. However, the requirement for an iPad remains a barrier to entry for the lowest-income households.

Future work will focus on automating the gaze estimation process using computer vision and AI to eliminate the manual coding bottleneck. Additionally, expanding the platform to Android and web-based systems will be essential for achieving true equity in developmental research. While the current sample size was insufficient for complex "ACE" twin modeling, these provocative findings on the heritability of response latency warrant replication in larger, more diverse cohorts.

Summary updated

Abstract

Together, these results highlight the potential of remote assessment of infants' psychological development, which can improve generalizability, inclusion, and scalability in developmental research.
BibTeX
@inproceedings{shic2023remote,
  author = {Frederick Shic and Kelsey Jackson Dommer and Jessica Benton and Beibin Li and James C. Snider and Pär Nyström and Terje Falck‐Ytter},
  title = {{Remote, tablet-based assessment of gaze following: a nationwide infant twin study}},
  booktitle = {Frontiers in Psychology},
  year = {2023},
  doi = {10.3389/fpsyg.2023.1223267},
}
Figure