Asocial learning is a mechanism by which innovations develop, and social learning is a mechanism by which innovations spread. Penetration of an innovative behavior through a population is measured by the proportion of the population that possesses the innovation. Via agent-based modeling, we examine innovation diffusion with agents learning and interacting in space. Simulations show that innovation spread systematically deviates from differential equations of the proportion of the population that has the innovation. Mediation analysis confirms that boundary surface length of groups having the innovation accounts for these spatial effects. Proportion of asocial innovative learners increases surface length which, in turn, increases social imitative learning.