Studies of infants’ and adults’ social cognition frequently use geometric-shape agents such as coloured squares and circles, but the influence of agent visual-form on social cognition has been little investigated. Here, although adults gave accurate explicit descriptions of interactions between geometric-shape aggressors and victims, implicit association tests for dominance and valence did not detect tendencies to encode the shapes’ social attributes on an implicit level. With regard to valence, the lack of any systematic implicit associations precludes conclusive interpretations. With regard to dominance, participants implicitly associated a yellow square as more dominant than a blue circle, even when the true relationship was the reverse of this and was correctly explicitly described by participants. Therefore, although explicit dominance judgements were strongly influenced by observed behaviour, implicit dominance associations were more clearly influenced by preconceived associations between visual form and social characteristics. This study represents a cautionary tale for those conducting experiments using geometric-shape agents.