This review focuses on three different processes: action priming, action prediction, and outcome evaluation. Together, these processes form a foundation for social perception early in life. Priming and prediction is argued to be separable processes with different degrees of plasticity, based in part on unique neural structures. These two future-oriented processes are assumed to operate in a sequential manner. A third set of processes, outcome evaluations, follows the completion of observed events and compare the actual events with the assumptions postulated by the preceding future-oriented processes. Together, these processes are argued to provide good grounds for learning via internal models that detect error signals that arise from the potential mismatch between priming and prediction and actual events as they unfold in the external world and use this information to update the accuracy of future-oriented processes.