It means that the increments of the independent variable correspond to the difference in magnitude and direction of the dependent variable change. The curvilinear (nonlinear) regression function can be, for example, square or exponential function.