Parent-infant social interactions start early in development, with infants showing active communicative expressions by just two months. A key question is how this social capacity develops. Maternal mirroring of infant expressions is considered an important, intuitive, parenting response, but evidence is sparse in the first two months concerning the conditions under which mirroring occurs and its developmental sequelae, including in clinical samples where the infant's social expressiveness may be affected. We investigated these questions by comparing the development of mother-infant interactions between a sample where the infant had cleft lip and a normal, unaffected, comparison sample. We videotaped dyads in their homes five times from one to ten weeks and used a microanalytic coding scheme for maternal and infant behaviours, including infant social expressions, and maternal mirroring and marking responses. We also recorded maternal gaze to the infant, using eye-tracking glasses. Although infants with cleft lip did show communicative behaviours, the rate of their development was slower than in comparison infants. This group difference was mediated by a lower rate of mirroring of infant expressions by mothers of infants with cleft lip; this effect was, in turn, partly accounted for by reduced gaze to the infant's mouth, although the clarity of infant social expressions (indexed by cleft severity) and maternal self-blame regarding the cleft were also influential. Results indicate the robustness of parent-infant interactions but also their sensitivity to specific variations in interactants' appearance and behaviour. Parental mirroring appears critical in infant social development, likely supported by the mirror neuron system and underlying clinical and, possibly, cultural differences in infant behaviour. These findings suggest new avenues for clinical intervention.