Do babies already track what other people believe?

Possibly, in a quiet, implicit way, but the evidence is disputed. Theory of mind is the capacity to understand other individuals by ascribing mental states to them. It is called a theory because no one has direct access to another's mind, so its existence and nature must be inferred.

The classic milestone is the false-belief task: most typically developing children pass from around age four. Infants cannot be asked questions, so researchers use the violation-of-expectation procedure, which exploits infants' tendency to look longer at unexpected and surprising events. In Onishi and Baillargeon's 2005 study, an actor hid a toy in a green box; then, without her seeing it but in sight of the baby, the toy moved to a yellow box. The 15-month-olds looked reliably longer when she searched the yellow box, as though surprised that she acted against her false belief.

The prevailing interpretation is that infants possess implicit mentalization, which is fast, non-conscious and non-verbal, rather than the explicit kind that emerges around age four. But infants do not consistently perform well on tasks requiring them to predict others' behavior from false beliefs, as reflected in the difficulty of replicating some findings. One longitudinal study found that children anticipated correctly only by age 4, performing at chance at 2 and 3. Why children fail traditional false-belief tasks until four, if infants already understand false beliefs, has become one of the most controversial debates in developmental psychology.

Language seems to matter for the explicit version: findings from deaf children with delayed exposure to sign language support the idea that language scaffolds explicit mental-state reasoning.