The way animals perceive partners, including humans, is yet relatively unknown. Research has shown that humans can provide social support or act as social substitute for domestic animals. Nonetheless, studies investigating the perception of humans by domestic animals in their social environment, i.e. alongside conspecifics, have been scarce. We investigated if humans could modulate lambs’ separation distress in a social environment paradigm. Artificially suckled lambs were familiarized to humans using positive tactile contacts from birth. A “familiar” and a “known” human had extensive and minimal contact with the lambs respectively. At four weeks of age, lambs were tested either individually or in a group of penmates. The test was a paradigm of “union-separation-reunion” (1½ min each) by visually isolating the individual lamb from the group during separation. We used a factorial design by constituting groups of two sizes, ‘two’ vs. ‘three’, with or without a human included. Therefore, the group was either: “1 lamb + 1 human” or “2 lambs” (for group sizes of two), “2 lambs + 1 human” or “3 lambs” (for group sizes of three). Lambs showed the strongest distress behaviours when isolated and, to a lesser extent, when they had only one partner (P < 0.001). On the other hand, group sizes of three partners and above (during reunion phases) nearly suppressed their vocalisations during social separation (P < 0.001). Our results confirmed the high distress lambs experienced when alone, as well as the importance of being in a group of three, possibly the minimum group size to reduce separation distress. Furthermore, during separation, the groups of three partners did not differ in the amount of vocalisations, whether this group was composed of two lambs with a human or three lambs together (P = 0.68), both vocalising much less than a group of only two lambs (P < 0.001). This supports the idea that, in our husbandry conditions, the human provided social support and could have played a role of social substitute during social separation. However, a lamb placed alone with a human vocalised much more than two lambs together (P < 0.001). Hence, the effectiveness of the human at providing social support was modulated by the social environment, i.e. by requiring a minimal group size in which the human was included. During social separation, a human who previously interacted positively with the lamb can act as a social substitute in presence of a few other conspecifics.