Evolution could be assumed as a natural reinforced learning. We tried simulations of Mutual-association with varying population size to investigate evolution and learning. Mutual associative memory is our extension from hetero-association or temporal-association of the Associative Memory by J.J.Hopfield[1]. Mutual Associative Memory is used as memory of organism for the tool to investigate evolution and learning. Genetic Algorithms are used to evolve the weights of mutual associative memory. We got the result that evolution of learning can be observed when organisms change rule itself during their lifetime.