The behaviors and stability of a new CMAC control scheme are studied, whose comparisons with PID controllers are also investigated. Unlike the traditional role of modeling inverse system dynamics, the CMAC controller is proposed to emulate the PD plus gravity-compensation control for robotics. In this new control scheme, the CMAC produces enormous integration action when the input vector moves slowly in the space so that it can respond quickly to large loads. On the other hand, it can also forget efficiently when the input vector moves fast in the space so that it is not burdened by previous integration. It is shown that the CMAC control system can converge into the target position. The systems employed in this study are one-link and two-link robotics.