According to the standard interpretation, Samson and colleagues (2010) well-known paradigm provides evidence of automatic level-1 visual perspective taking. In this study, we explore the alternative interpretation that the effect found in this paradigm is at least partially driven by a domain-general attentional process, i.e. spatial cueing. We are conducting a series of experiments designed to adjudicate between these two theoretical interpretations, and have found preliminary evidence supporting the latter interpretation. However, we also argue that agents (i.e. bodies, faces, eyes) are likely to function as special kinds of cue with unique characteristics  which can, for example, be evaluated for trustworthiness (Bayliss and Tipper 2006). In this talk, I will sketch a paradigm aiming to test a battery of hypotheses based upon the idea that agents  in contrast to arrows  generate an implicit commitment to be informative by making eye contact.