Tamir et al. do some interesting MRI studies that suggest that understanding the mental states of others starts with self perception as an anchor from which serial adjustments of the perceptions of others are made:
Recent studies have suggested that the medial prefrontal cortex (MPFC) contributes both to understanding the mental states of others and to introspecting about one's own mind. This finding has suggested that perceivers might use their own thoughts and feelings as a starting point for making inferences about others, consistent with “simulation” or “self-projection” views of social cognition. However, perceivers cannot simply assume that others think and feel exactly as they do; social cognition also must include processes that adjust for perceived differences between self and other. Recent cognitive work has suggested that such correction occurs through a process of “anchoring-and-adjustment” by which perceivers serially tune their inferences from an initial starting point based on their own introspections. Here, we used functional MRI to test two predictions derived from this anchoring-and-adjustment view. Participants (n = 64) used a Likert scale to judge the preferences of another person and to indicate their own preferences on the same items, allowing us to calculate the discrepancy between the participant's answers for self and other. Whole-brain parametric analyses identified a region in the MPFC in which activity was related linearly to this self–other discrepancy when inferring the mental states of others. These findings suggest both that the self serves as an important starting point from which to understand others and that perceivers customize such inferences by serially adjusting away from this anchor.
A bit more on the actual experimental design:
Figure - The relation between BOLD response and self–other discrepancy during Other trials was calculated separately for subregions of the MPFC. Although the response of dorsal MPFC (A) increased linearly with increasing self–other discrepancy, the response of ventral MPFC (B) distinguished only between trials on which self–other discrepancy was zero (overlap between self and other) versus greater than zero (discrepancy between self and other). Error bars indicate the SEM.
Although the specific design of the four experiments differed slightly, each required participants to answer a series of questions about their opinions and preferences and to judge how other individuals would answer the same questions. On each trial, participants saw a cue that indicated the target of the judgment (self or another person) and a brief phrase (e.g., “enjoy winter sports such as skiing or snowboarding”; “fear speaking in public”). Participants used either a four- or five-point scale either to report how well the statement described themselves or to judge how well it described the other person. Within each experiment, participants considered the same set of statements for self and other.
Before scanning, participants were told that the purpose of the experiment was to examine how people make inferences about target individuals on the basis of minimal or no information. In all studies, targets were college-aged individuals depicted by a photograph downloaded from an internet dating website, although the specific identity of individuals varied across studies.