The apparatus of the present invention corrects a control target value of ignition timing using a multipoint learned value AGdp(n) for compensating for a change amount of the ignition timing caused by time-dependent change of the engine and a basic learned value AG(i) for compensating for a change amount of the ignition timing caused by a factor other than the aforementioned time-dependent change of the engine. In a multipoint learning range n in which the time-dependent change of the engine influences the ignition timing to a great extent, the control target is corrected using the multipoint learned value AGdp(n) and the basic learned value AG(i). In ranges other than the multipoint learning range n, the control target is corrected using only the basic learned value AG(i). Normally, only the learning of the multipoint learned value AGdp(n) is permitted in the multipoint learning range n, and only the learning of the basic learned value AG(i) is permitted in the ranges other than the multipoint learning range n. When it is determined that there is a possibility that fuel has been fed to a fuel tank, the learning of the multipoint learned value AGdp(n) is prohibited for a subsequent predetermined period (if NO in step S103 or step S104).