Leiden University Student Repository

Lange, Heleen de (2023)

Eye-Tracking Data Quality and Fixation Classification Algorithms: Influence on Key Dependent Variables in Implicit Theory of Mind Research

Research master thesis | Developmental Psychopathology in Education and Child Studies (research) (MSc)

This thesis is part of a multi-lab eye-tracking project investigating implicit Theory of Mind (ToM) in toddlers and adults (ManyBabies2). Given the recent number of unsuccessful replications in this field, the focus of this thesis lies in investigating the impact of methodological considerations regarding data quality on the obtained results. Specifically, this thesis addresses differences in eye-tracking data quality (accuracy, precision, and data loss) between toddlers and adults. Further, it examines how differences in data quality (precision and data loss) affect the dependent variables in an anticipatory looking (AL) task to measure ToM, and whether the results of the AL task change by adjusting the parameters of a standard fixation classification algorithm (i.e., the Tobii I-VT filter) to account for low data quality. A total of 16 adults and 16 toddlers (N = 32) participated in the ManyBabies2 study at the Leiden University lab. The AL task included two test trials with an...Show moreThis thesis is part of a multi-lab eye-tracking project investigating implicit Theory of Mind (ToM) in toddlers and adults (ManyBabies2). Given the recent number of unsuccessful replications in this field, the focus of this thesis lies in investigating the impact of methodological considerations regarding data quality on the obtained results. Specifically, this thesis addresses differences in eye-tracking data quality (accuracy, precision, and data loss) between toddlers and adults. Further, it examines how differences in data quality (precision and data loss) affect the dependent variables in an anticipatory looking (AL) task to measure ToM, and whether the results of the AL task change by adjusting the parameters of a standard fixation classification algorithm (i.e., the Tobii I-VT filter) to account for low data quality. A total of 16 adults and 16 toddlers (N = 32) participated in the ManyBabies2 study at the Leiden University lab. The AL task included two test trials with an ignorance and knowledge condition, during which participants’ looking behavior (first look and proportion differential looking score [DLS]) was measured. The results showed a significant difference in average data quality between toddlers and adults, with toddlers having lower accuracy and precision and higher data loss than adults. Moreover, the results indicated that data loss significantly affects proportion DLS in the ignorance and knowledge condition. However, no differences in AL looking behavior were found between the Tobii I-VT default and adjusted parameters. The findings of this thesis emphasize the need for consistent reporting of data quality. Future research should conduct larger-scale studies to validate and expand upon the findings in this thesis, and focus on refining the fixation classification algorithm parameters of the Tobii I-VT filter to adjust for differences in data quality.Show less

Author

Lange, Heleen de

Faculty

Faculty of Social & Behavioural Sciences

Specialisation

Applied Neuroscience in Human Development

Supervisors

Biro, S.
Van Rijn, S.

ECTS Credits

25

Language

en