There is a lack of reliable and valid measurement tools to assess neck function in infants with congenital muscular torticollis, and most physiotherapists use visual estimation, which has not been adequately tested for reliability in this population. We examined the reliability of visual estimation of head tilt and active neck rotation in the upright position, on infants with congenital muscular torticollis by physiotherapists. We recruited 31 infants and 26 physiotherapists. Therapists rated videos of infants’ head position in the frontal plane (tilt) and transverse plane (active rotation) using visual estimation, on two occasions at least one week apart. Overall, inter-rater reliability was good (mean ICC, 0.68 ± 0.20; mean SEM, 5.1° ± 2.1°). Rotation videos had better reliability (mean ICC, 0.79 ± 0.14) than head tilt videos (mean ICC, 0.58 ± 0.20). Intra-rater reliability was excellent (mean ICC, 0.85 ± 0.08). Both head tilt and rotation had excellent reliability (mean ICC, 0.84 ± 0.08 for head tilt and 0.85 ± 0.09 for rotation). There was no correlation between intra-rater reliability and clinical experience.