This paper examines how different structures of social networks affect linguistic selection type of change or functionally biased change, and game type of change or socially biased change, based on simulation and historical data from English. It shows that the weaker the functional or social bias, the greater the effects of different network structures on diffusion processes of change. It also shows that the weaker the functional or social bias, the more probabilistic the learner becomes, and the stronger the functional or social bias, the more categorical the learner becomes. Furthermore, it is discussed that there is little increase in diffusion time with the increase in population size in a small-world and scale-free networks.