Robot-assisted treadmill training improves motor function and walking ability in neurologically impaired patients. However, despite attention having been shown to play a role in training success, psychological responsiveness to task difficulty and motivational levels at task onset have not been measured. Seven healthy subjects participated in a robot-assist treadmill training task. Subjects engaged in a virtual task with varying difficulty levels that was shown to induce a feeling of being bored, excited and over-stressed. The participants\' mental engagement was measured using the ECG-based heart rate variability in real time, during gait training as a proxy for EEG and psychological test batteries. Heart rate variability (HRV), which has been shown to reflect cortical engagement for both cognitive and physical tasks, was measured using nonlinear measures obtained from the Poincare plot. We show that the cortical response to the task measured with HRV varies in relation to the level of mental engagement in response to the difficulty level of the virtual task. From these results we propose that nonlinear measures quantify cortical response / motivational level to robot-assist motor learning tasks and that the adaptation to the task is dependent on the level of motivation.