MOLeCS is a new Classifier System which addresses its learning as a multiobjective optimization of two goals: classifier accuracy and generality. These objectives are both emphasized in the fitness evaluation stage, driving the GA search towards the formation of accurate and general rules. We consider severa! existing multiobjec­tive optimization strategies which establish a compro­mise between generality and accuracy in different ways. The system introduces two new proposals which balance both objectives with a bias towards accuracy, resulting in better classification performance. The system also considers a third major objective: covering. It is achieved using sorne niching mechanisms that favour the maintenance of a set of cooperative rules.