Abstract: Animal learning theory has been enormously influential in setting up laws of how individuals gradually learn associations and instrumentation by reinforcement. Yet, the theory rests on data collected from socially isolated laboratory animals, exposed to artificial cause–effect relations without visible agents. We review the primate vocal learning literature and find that animal learning theory performs poorly in accounting for real-life learning and evolutionarily relevant problem-solving. Instead, learning occurs when conspecifics act as event-causing agents, often without direct consequences for learners. We illustrate this with recent field studies, which suggest that the default mode of learning may not be through reinforcement and repeated trials but by acquiring scripts — mental representations of how events typically unfold. Becoming communicatively competent may be more about learning how events unfold than becoming conditioned to stimuli and responses.