To what degree encode action representations only low level visual features or more semantic aspects of an action? In the literature both of these possibilities are mentioned. However the two options have never been directly compared. This project examines whether action recognition processes are sensitive to the meaningfulness of an action, to low level kinematic features of an action or both. To change the meaningfulness of an action stimulis, we chose known (hug, and push) and unknown actions (RA3, SA3 - both are movements in which an actor angles both of their arms in different positions). Moreover low level kinematic feature can be examined by scrambling these actions (remapping of arm movements on the feet and vice versa): this generates biological impossible actions that are likely to carry no action meaning. These actions have similar kinematic features to the non-scrambled versions of the actions. We will use a behavioral adaptation effect to measure the sensitivity of action representation to these different manipulations. The presence of adaptation effects will indicate the presence of a neural representation of the action. The results have implications for the understanding of the origin of action adaptation effects and the fundamental principles of action recognition. E.g. if we find action adaptation effect for biological impossible movements that are similar in magnitude to their non-scrambled counterparts then action representations encode low level visual features that cannot necessarily be simulated within the motor system. Hence, in this case it would be unlikely that the motor system contributes to action recognition.