Learning classifier systems (LCS) traditionally use a binary string rule representation with wildcards added to allow for generalizations over the problem encoding. We have presented a neural network-based representation to aid their use in complex problem domains. Here each rule's condition and action are represented by a small neural network, evolved through the actions of the genetic algorithm. In this paper, we present results from the use of backpropagation to provide local search in conjunction with the global search of the genetic algorithm within XCS creating a memetic neural LCS. Significant decreases in the time taken to reach optimal behaviour are obtained from the incorporation of this local learning algorithm.