Post-stroke rehabilitation of the upper extremity is a challenge in neurorehabilitation. Selective training of different degrees of freedom (training of specific monoarticular movements and in one plane of movement of the upper extremity) to achieve true recovery could be a good approach. However, it is not known how the training should be structured to avoid interference between the different movements trained, hindering the motor learning process during neurorehabilitation. This research aims to determine the effect on performance and kinematic control of a selective movement control task during one- or three-days training of 2 different degrees of freedom, close or distant in cortical representation, of the upper extremity in people with early subacute and chronic phase of stroke. The hypothesis is that in the upper extremity of subacute or chronic phase stroke survivors, learning 1 target DoF is interfered by training another DoF of the same joint. In contrast, learning 1 DoF from a different, distant joint does not interfere with learning the target DoF in a training session. Likewise, this interference is diminished in multiple training sessions. On the other hand, in early post-stroke phase, within the first 5 weeks, the gain in kinematic control of each controlled DoF is greater than in later post-stroke phases (between 9 to 12 weeks of evolution or chronic phase). An experimental design of 4 days of training, of approximately 60 minutes per day, will be carried out in people who are within the first 3 months after a stroke or in a chronic stage. There will be 3 intervention groups that will train the shoulder flexion movement of the paretic upper extremity and another movement, which can be of the same affected upper extremity or another of the contralateral lower extremity.