Motor imagery (MI) refers to the mental rehearsal of a movement without actual motor output. MI training has positive effects on upper limb recovery after stroke. However, until now it is unclear whether this effect is specific to the trained task or a more general motor skill improvement. This study was set up to advance our insights into the efficacy of MI training and the specificity of its effects. We investigated whether MI training affected the trained hand exclusively, or both hands. Four stroke participants received a 15-min MI training four times a week for 3 weeks. Hand function was measured before and after the training using three measurements of increasing complexity. Hand function improved after MI training, thus confirming the earlier studies. Second, we found specific effects of the MI training for two of the three measurements. These results suggest that MI specificity is dependent on the complexity of the hand function task.