We will apply the theory to see the exact way of walking, jumping and tree climbing at humans. In accordance to MDT, an action on the external reality (e.g. walking) implies the existence of a long-range action model (ZAM). This model generates the approximate plan of the action. ZAM will build and activate a number of local models (ZAM and AZM) to reach its goals. A local ZAM will simulate the movement of the leg for the first step. If the simulated step is successful, then ZAM will activate the action on the external reality. The leg will move in the same way as the leg 'moved' during the simulation. It is not possible to do any movement, if it had not been successfully simulated before. Let's see the case of jumping. In front of an obstacle, which has to be jumped over, the brain will 'execute' a simulated jump. If the simulated jump succeeds, it can be done in the external reality as well, activating the model, which did the simulated successful jump. If the simulation does not succeed, there will be no model to activate the muscles of the body, and the being will be blocked to act. Any attempt to go against the internal decisison will fail. The conclusion from the previous analysis is that a more or less elaborate simulation precedes any action on the external reality. The result is that an extremely complicated activity, like e.g. walking, is executed with remarkable precision and elegance. At first sight, walking seems to be a relatively simple activity. At a closer analysis, one can see extreme complexity. The first problem is keeping the equilibrium during walking. The stability of humans and animals during walking is a dynamical stability. This means that, if we "froze" the body in an intermediate position, the body would not be stable and would fall. During walking, the models anticipate the movements of the body through simulation and send suitable commands in advance. If there was no anticipation of the evolution and we counted only on the stability and position sensors, the information would get delayed to the device taking the decision and such, the system would have a reduced stability. This is how all the electronic stabilizer systems work: they wait for something to happen to make a correction. In the case of the brain, the information from the stability and position sensors is used to anticipate the possible future problems and act before the problem arised. This is the dynamical stability and, I think, this problem cannot be solved in real time by any existing computer due to the low power of the present computers. From here we can see the huge capacity of information processing of any brain, starting with mammals. The most primitive mammals, with brains of a few grams or tens of grams, are able of higher performance than humans, in running and jumping.