The neural evidence for simulation is weaker than I think you think it is

Abstract Simulation theory accounts of mind-reading propose that the observer generates a mental state that matches the state of the target and then uses this state as the basis for an attribution of a similar state to the target. The key proposal is thus that mechanisms that are primarily used online, when a person experiences a kind of mental state, are then co-opted to run Simulations of similar states in another person. Here I consider the neuroscientific evidence for this view. I argue that there is substantial evidence for co-opted mechanisms, leading from one individual’s mental state to a matching state in an observer, but there is no evidence that the output of these co-opted mechanisms serve as the basis for mental state attributions. There is also substantial evidence for attribution mechanisms that serve as the basis for mental state attributions, but there is no evidence that these mechanisms receive their input from co-opted mechanisms. Keywords Theory of mind Simulation theory Mirror neurons Mentalizing Temporo-parietal junction Medial prefrontal cortex A man stops at the gym on his way to work early on Monday morning, but the gym is closed, so he heads to work. Now consider this extra information: when he sees the gym is closed, he remembers that it’s a national holiday; earlier, he had forgotten about the holiday (Apperly 2008). The extra information describes mental states—the man’s (later) true and (earlier) false beliefs. These features of the scenario can’t be perceived directly by external observers. Even for the man himself, the later realization may be conscious, but the earlier forgetting did not have any observable phenomenal experience. There is nothing it feels like to be forgetting that it’s a holiday. Nevertheless, mental states like forgetting are