This paper presents an iterative learning control method with which the error trajectory can be pre-specified. The method does not require that the initial condition should remain to be a fixed value for each iteration, whereas this requirement is usually assumed in conventional methods. The proposed strategy is to make the tracking error trajectory converge to the pre-specified one over the entire interval. The constant parametrization, time-varying parametrization, and a combined situation are respectively examined. By the Lyapunov-like approach, accordingly, learning laws are given and the learning systems are analyzed in detail. With the utilization of unsaturated/saturated learning laws, the system error coincides with the pre-specified error trajectory over the entire interval, and all the signals in the closed-up system are guaranteed to be bounded. Effectiveness of the proposed method is verified with the numerical results presented.