Since the seminal work of Fritz Heider and Marienne Simmel (1944) the study of animacy perception, or the perception and attribution of life from the motion of simple geometrical shapes has intrigued researchers. The intrigue for psychologists and vision scientists then and today centered on the stark disconnect between the simplicity of the visual input and the universal richness of the resulting percept. Infant research in this domain has become critical in examining the ontological processes behind the formation of animated percepts. To date, little is known about how infants process these kinds of stimuli. While numerous habituation studies have shown sensitivity to animate motion in general, none to date has examined whether infants actually perceive animate displays as social interactions. The overarching goal of the present thesis is to answer this question and further augment knowledge about the mechanisms behind the formation of animated percepts in infancy. I, along with my collaborators, do so in three ways, in three separate studies. First, we examined visual attention during online observation of randomly moving geometrical shapes in adults and infants (Study I, using eye tracking). Second, we examine distribution of visual attention in infancy during online observation of non-contact causal interactions, focusing on the most ubiquitous, fitness relevant of interactions – chasing (Study II, using eye tracking). Third, we answer the question whether infants perceive social content in chasing displays by measuring the neural correlates in response to chasing (Study III, using EEG). The collective contribution of the present work is also three fold. First, it demonstrates that starting at the end of the first year of life, human visual system is sensitive to cues that efficiently predict an interaction. Second, at 5-months infants begins allocating attention differently across agents within interactions. Finally, attention to specific objects is not due to low-level saliency but the social nature of the interaction. Subsequently, I present the case that perception of social agents is fast, direct, and reflects the workings of a specialized learning mechanisms whose function is the detection of heat-seeking animates in motion.