Experimental psychologists frequently present participants with social stimuli (videos or pictures) and measure behavioural responses. Such designs are problematic in that they remove the potential for social interaction and inadvertently restrict our eyes multifaceted nature as a tool to both perceive and communicate with others. The aim of this study was to develop a new paradigm within which we can easily and reliably measure the influence of top-down processes (belief), social activity (talking and listening), and possible clinical traits (gaze anxiety, and social interaction difficulties) onto gaze behaviours. Participants were engaged in a “real” or pre-recorded Skype conversation. Findings suggest that participants who believed they were engaging in a real conversation spent less time looking at the speaker’s eyes, but no differences were found for dwell time onto the whole face. Within our non-clinical sample, higher levels of gaze anxiety resulted in reduced dwell time onto the whole face but not eyes, whereas social interaction difficulties produced reduced dwell time onto the eyes only. Finally, talking consistently produced reduced dwell time onto the whole face and eyes regardless of any other conditions.