AimUnderstanding how haptic interaction supports interpersonal coordination during locomotion is important to develop assistive technologies when necessary. While significant work has been done on haptic interactions during adult locomotion, little is known about how children interact between each other or with an adult during walking. Here, we studied haptic-guided locomotion in children and adults.MethodsWe examined 11 pairs (adult-adult, child-child [6–8 years old], and adult-child) walking side by side with hand contact toward targets, with one participant leading and one blindfolded follower. The walking path was either straight or curved. We recorded and analysed upper limb muscle electromyography, kinematics, and haptic interaction forces.Results and conclusionAll dyads (adult-adult, child-child, adult-child) showed relatively small interaction forces (around 3 N), which presumably function primarily as communicative cues rather than as direct mechanical drivers of movement of the partner. Gait initiation involved compliant interaction in all dyads, with frequent anterior deltoid shortening reactions aiding arm elevation and movement onset, particularly prominent in adult-child pairs. During curved locomotion, small direction-specific adjustments in force (2–3 N) and arm elevation (3–4) conveyed effective haptic cues across ages. In addition, we found clear age-related features in the haptic interaction. Adults reduced upper-limb compliance when guiding children as compared with guiding another adult. However, children were systematically more compliant when interacting with adults, irrespective of their role, leader or follower. We interpret this difference as indicating that adults emphasize precise control and interaction stability, whereas children display more variable and reactive motor behaviour. The latter behaviour may reflect the need of children to learn and explore while walking in tandem. However, it may also reflect a compliance control that is different from that of adults.LimitationsThe sample size and children age range were limited. Moreover, we only included female adults.