There is disagreement about the effectiveness of observational practice for the acquisition of novel coordination skills and the type of processes involved in observation of novel movements. In this study, we examined learning of a bimanual 90 degrees phase offset through comparisons of three groups; physical practice, observational practice and no practice (n = 12/group). Groups were compared before and after practice on perception and production scans of the practised pattern. The observation group was yoked to the physical group such that observers watched repeated demonstrations of a learning model. Although there were no positive effects of observational practice for physical performance measures, the observation group did not differ from the physical practice group and was more accurate than controls on perceptual discrimination measures after practice. We concluded that observation of a novel bimanual movement can aid perception but that physical practice is necessary for immediate physical performance benefits. These results are discussed in terms of cognitive mediation models of motor skill learning.