In recent years, there has been growing interest in how we perceive dyadic interactions between people. It has been proposed that pairs of individuals shown upright and face-to-face recruit a form of configural processing, similar to that engaged by upright faces. This processing is thought to aid the detection and interpretation of social interactions. Dyadic arrangements shown back-to-back or upside-down are not thought to engage configural dyad processing. One of the key advantages conveyed by configural face processing is greater sensitivity to the spatial relationships between facial features when faces are viewed upright, than when viewed upside-down. If upright dyads arranged face-to-face engage similar configural processing that is not engaged by non-facing or inverted dyads, participants should therefore exhibit disproportionate sensitivity to the spatial relations between the constituent actors under these conditions. In four well-powered experiments, we find no evidence for this prediction: participants exhibited similar levels of sensitivity to changes in interpersonal distance regardless of whether dyads were shown upright or inverted, face-to-face or back-to-back. In contrast, we observe clear evidence that upright presentation affords greater sensitivity to inter-feature spatial relationships (interocular distance) when viewing faces. These results suggest that any configural processing engaged by upright facing dyads likely differs qualitatively from that engaged by upright faces.