Abstract
Recent eye-tracking studies have demonstrated significant differences in face scanning strategies between Western Caucasians (WC) and East Asians (EA), challenging the notion of universality in face perception. However, previous studies have been restricted to screen-based paradigms, which lack the visual complexity and social presence of real-world conditions. The current study therefore used head-mounted eye-tracking techniques to investigate cultural differences in naturalistic face scanning behaviour. Thirty British and 27 Japanese adults introduced themselves and played a story-telling game with a local confederate (in the UK or Japan) while their eye movements were recorded. We developed semi-automatic MatLab tools that can dynamically track regions of interest (upper/lower face) and classify gaze points accordingly. Results showed that both EA and WC groups looked significantly more at the face when listening compared to speaking. Cultural differences were observed for speaking periods, with WC individuals showing more face gaze at the listening partner. A tendency for increased gaze scanning (proportional to face looking time) was found for the EA group, challenging reports of gaze avoidance in EA observers (Argyle et al., 1986). To employ a more spatially-sensitive and data-driven approach, we mapped face regions and gaze points into a normalised space to generate z-scored difference maps of gaze density. Initial results showed that EA participants exhibited more localised face scanning, with increased gaze at the nose and between the eyes of the conversational partner. WC observers, meanwhile, showed greater gaze distribution and looked more at the mouth region. Overall, this study introduced semi-automatic and data-driven approaches for analysing data from head-mounted eye-trackers, with findings demonstrating cultural differences in face scanning under naturalistic conditions for the first time.
Meeting abstract presented at VSS 2018