In order to grasp characteristics of pilot maneuver, this paper analyzes nonlinearity of human internal models corresponding to feedforward and feedback roles. Manual control experiments were carried out and feedforward and feedback human internal models are identified from the experimental data by using neural networks. The obtained internal models are analyzed through time records and spectral analysis. In the results, it is found that human operations include nonlinear behaviors. Conventionally, there are a large number of researches that describe human operation as quasi-linear models and analyze them to estimate the difficulty of operation. However, this research suggests a possibility that actual human operations include nonlinearity and the nonlinear characteristics of human internal models enable us to explain contextual human operation. It is also suggested that the level of controlling skill correlate well with nonlinearity of human internal models.