["Developmental Science, Volume 29, Issue 3, May 2026. ", "\n\nABSTRACT\nThe origin of co‐representation during joint action poses a puzzle: It apparently only emerges around the age of four in humans, suggesting it is cognitively demanding, but has also been demonstrated in several nonhuman primate species whose cognitive skills do not match human four‐year‐olds. We therefore reassessed co‐representation in 2–4‐year‐old human children (n = 38, 11 females) with a directly comparable, nonverbal task previously applied to nonhuman primates. Co‐representation was already present and strongest in the youngest children, not constrained by Theory of Mind and inhibitory control skills, and weaker in children than in nonhuman primates. Together, this suggests co‐representation may be an early default mode of processing joint action. However, species differed in the flexibility to adjust when to merge perspectives by co‐representing, and when not. Children and the cooperatively breeding marmosets were most flexible and relied on coordination smoothers to achieve this (marmosets: mutual gaze; children: mutual gaze and communication).\n\n\nSummary\n\nCo‐representation, that is the merging of perspectives, is present in the joint Simon task in 2‐year‐olds and decreases with age\nCo‐representation is weaker in children than in nonhuman primates\nCooperation success requires flexibly switching between merging or not\nChildren and highly cooperative nonhuman primate species rely on mutual gaze as coordination smoothers for switching\nCo‐representation is likely an early default mode of processing joint actions\n\n\n"]