How is mindreading really like reading?

I suggest an alternative basis for Heyes' analogy between cultural learning of mindreading and text reading. Unlike text reading, mindreading does not entail decoding of observable stimuli. Like text reading, mindreading requires relevant inferences. Identification of relevant inferences is a deeply challenging problem, and the most important contribution of cultural learning to mindreading may be an apprenticeship in thinking like a mindreader.