Developing Methodological Tools for Ecologically-valid Infant Eye-Tracking Research

Conducting robust and relevant research is uniquely challenging when investigating infant participants given their limited linguistic and behavioural repertoires. However, the location of an infant's gaze is a reliable index of their attention, and this was exploited in key methodological innovations during the 60s which allowed researchers to answer foundational questions about infants' perceptual abilities within an experimental laboratory setting. Since then, technological improvements such as remote eye-tracking have enabled infants' eye movements to be recorded with greater accuracy and precision. Yet fundamental questions remain concerning the validity of recording the duration of infant looking toward highly artificial stimuli, particularly when investigating complex, higher-order social and cognitive abilities which naturally occur within the context of dynamic interactions. Nevertheless, using naturalistic (e.g. dynamic, interactive) stimuli presents substantial data processing challenges. This thesis introduces the 'Gaze Contingent Social Interaction paradigm', as a flexible methodological template for conducting lab-based eye-tracking experiments with greater ecological validity. This paradigm is applied within several empirical 'worked examples'; research topics of current interest to infant socio-cognitive research that predominantly use static stimulus presentations (e.g. facial race and affect processing). Within these experiments, several methodological tools (both novel and adapted for infant research) are also demonstrated, (e.g. 'Dynamic AOIs', heatmap and temporal analyses) that will hopefully aid developmental researchers to reliably analyse the rich data generated within this paradigm. Overall, the empirical chapters of this thesis highlight key differences between infants' eye movements for naturalistic stimuli compared to static images, which encourages us to consider if the methods currently used in lab-based studies reliably generalise to the 'real-world'.