It has long been held that our capacity to understand mental representations, referred to as Theory of Mind (ToM), emerges around 4 years, when children begin to understand that beliefs may differ from reality. This view, however, was questioned by evidence that already preverbal infants predict how others will act in line with their mental states. When does Theory of Mind develop and what explains this discrepancy? With structural MRI in children aged 3 to 4 years, we showed that different neural networks support verbal belief reasoning at 4 years and non-verbal success in the infant tasks. Verbal false belief reasoning is supported by the maturation of regions and connections of the Default Network, also involved in Theory of Mind in adults. In contrast, infant Theory of Mind tasks are associated with the maturation of the Salience Network involved in social attention processes. Using frequency tagging with EEG, we show that infants’ (and adults’) neural visual processing of objects is enhanced if other agents see these objects. I propose that this offers a cognitively efficient way of considering the things that others see, available from infancy, as an alternative to more effortful and later-developing verbal ToM.