Assessing two methods of webcam-based eye-tracking for child language research

Area of Science:

• Developmental Psychology
• Cognitive Science
• Human-Computer Interaction

Background:

• Web-based eye-tracking offers a scalable method for child research.
• Assessing the feasibility and accuracy of webcam-based eye-tracking in children is crucial for online studies.

Purpose of the Study:

• To evaluate two webcam eye-tracking methods (WebGazer.js and hand annotation) for feasibility and accuracy in children.
• To compare the sensitivity and temporal precision of automatic gaze estimation versus manual video annotation for child eye-tracking experiments.

Main Methods:

• Experiment 1: Compared WebGazer.js and hand annotation in a visual-world language task with 5-6 year olds.
• Experiment 2: Assessed WebGazer.js spatiotemporal resolution in a visual-fixation task with 4-12 year olds.
• Both experiments utilized webcam recordings for eye-tracking data collection.

Main Results:

• Web-based eye-tracking is feasible for children in both supervised and unsupervised settings.
• Hand annotation demonstrated higher sensitivity and accuracy for fine-grained effects relevant to child language research.
• WebGazer.js provided noisier gaze estimates with lower temporal precision compared to hand annotation.

Conclusions:

• Webcam video annotation is recommended for detailed analysis in child language research conducted online.
• WebGazer.js may be suitable for less precise gaze tracking but requires careful consideration of its limitations.
• Recommendations are provided for researchers designing online child eye-tracking studies using webcam methodologies.