IntroductionInfants with asymmetric brain lesions are at high risk of developing congenital hemiplegia. Action–observation training (AOT) has been shown to effectively improve upper limb motor function in adults with chronic stroke. AOT is based on action observation, whereby new motor skills can be learnt by observing motor actions. This process is facilitated by the Mirror Neuron System, which matches observed and performed motor actions. This study aims to determine the efficacy of AOT in: (1) influencing the early development of reaching and grasping of typically developing infants and (2) improving the upper limb activity of infants with asymmetric brain lesions.Methods and analysisThis study design comprises two parallel randomised sham-controlled trials (RCTs) in: (1) typically developing infants (cohort I) and (2) infants with asymmetric brain lesions (eg, arterial stroke, venous infarction, intraventricular haemorrhage or periventricular leukomalacia; cohort II). Cohort II will be identified through a neonatal ultrasound or neonatal MRI. A sham control will be used for both RCTs, taking into consideration that it would be unethical to give no intervention to an at-risk population. Based on a two-tailed t test of two independent means, with a significance (α) level of 0.05, 80% power, predicted effect size of 0.8 and a 90% retention rate, we require 20 participants in each group (total sample of 40) for cohort I. The sample size for cohort II was based on the assumption that the effect size of the proposed training would be similar to that found by Heathcock et al in preterm born infants (n=26) with a mean effect size of 2.4. Given the high effect size, the calculation returned a sample of only four participants per group, on a two-tailed t test, with a significance (α) level of 0.05 and 80% power. As cohort II will consist of two subgroups of lesion type (ie, arterial stroke and venous infarction), we have quadrupled the sample to include 16 participants in each group (total sample of 32). Infants will be randomised to receive either AOT or standard Toy Observation Training (TOT). Both interventions will be of 4 weeks’ duration, from the infant's 9th–13th post-term week of age. Three sessions of 5 min each will be performed each day for 6 days/week (total of 6 h over 28 days). Parents of the AOT group will repeatedly show the infant a grasping action on a set of three toys, presented in random order. Parents of the TOT group will show the infant the same set of three toys, in random order, without demonstrating the grasping action. At 14, 16 and 18 weeks, the quantity and quality of reaching and grasping will be measured using the Grasping and Reaching Assessment of Brisbane; symmetry of reaching and grasping will be measured using the Hand Assessment of Infants (HAI) and pressure of grasping for each hand with a customised pressure sensor. At 6 months’ corrected age, the primary outcome measures will be the HAI and Bayley Scales of Infant and Toddler Development (third edition; BSID III), to measure cog...