First, on the test data, this section shows the anticipation performance for the MLP and for the abstract RNN and an analysis of predictions from outside the training domain. Second, on the robot, the MLP chain is applied to goal-directed action-selection and mental transformation.