If we mentalize about each other by imagining ourselves experiencing an event ‘as another person’ and then predicting our own mental states in that situation, how do we make reasonable predictions about our hypothetical thoughts and feelings in the first place? To use ourselves as a proxy for others' mental states, we must not only be able to imagine ourselves as another person but also be able to simulate richly enough to provoke in us a concomitant set of feelings and thoughts. That is, we must first conjure up the actual feeling states that accompany a particular experience or think the thoughts that might arise in a given scenario before we can proceed to extend those simulated feelings and thoughts to another person.
Surprisingly little cognitive work has addressed the question of how humans predict their own mental states, despite reasonable arguments that this skill represents an important line separating human cognition from the mental systems of other primates (Gilbert 2007
). Do perceivers simply liken such experiences to similar situations from their past, thereby drawing on episodic memory to generate a prediction of the kinds of mental states they might encounter? Do they apply a set of rules—somewhat similar to those postulated by theory–theorists for mentalizing about others—that output a prediction about what a perceiver herself would experience in a situation? Or do perceivers engage in something richer and more constructive than either of these two cognitive strategies?
Although we almost certainly make occasional use of both memory and some kind of rule-based inferences for predicting our own mental states, we must also possess a system for predicting our mental states in truly novel situations, where we cannot avail ourselves of memory for our past experiences or global rules about people in general. When asked to answer highly unusual questions, such as whether they would rather spend a year alone as an astronaut on Mars or a year in a submarine stationed under the polar ice cap, respondents can generate an answer that feels as if it accurately reflects their preference. Such questions do not leave us dumbfounded, despite most of us having neither lived on Mars nor under the North Pole (nor knowing anyone who has); moreover, if asked again tomorrow, we would probably provide the same answer. How does our cognitive system predict our mental states under such radically novel and unusual situations?
Across a number of studies, our group has attempted to address this question by examining the neural systems that support stable predictions about a particular form of one's future mental states: one's preferences about the kinds of things one would like or dislike (Ames et al. 2008
; Jenkins et al. 2008
; Mitchell et al. 2006b
). In these studies, participants considered a series of questions that asked them about their opinions and preferences across a range of topics. Questions were designed such that respondents were unlikely to be able to answer them on the basis of ‘precompiled’ semantic representations. Intriguingly, this task consistently engaged one of the regions involved in projective simulation of other minds and other times and places: the ventral MPFC. These observations are suggestive that, when asked to calculate their own preferences, respondents may begin by (either consciously or unconsciously) simulating themselves enmeshed in the relevant situation and reading off the kinds of feelings they expect to have about it. However, additional work is needed to provide additional empirical support for this suggestion more fully.
Consistent with the observation that the ventral MPFC subserves predictions about one's own likes and dislikes, neuropsychological patients with damage to this region show considerable instability in their reported preferences. Fellows & Farah (2007)
asked patients with damage to the ventral MPFC, patients with damage to the dorsolateral PFC and healthy controls to indicate how much they liked a series of actors. Actors were presented in pairs, and participants were instructed to report which of the two they preferred. Both healthy controls and patients with dorsolateral PFC damage reported highly stable preferences: if a participant preferred Ben Affleck over Matthew Broderick, and Broderick over Tom Cruise, then he almost always also preferred Affleck over Cruise (i.e. A>B and B>C, therefore A>C). By contrast, patients with ventral MPFC damage showed much more inconsistent preferences; for example, an individual might indicate that he preferred Affleck over Broderick, Broderick over Cruise, but choose Cruise in a head-to-head comparison with Affleck (i.e. A>B, B>C, but C>A). Although they do not directly address the question of how exactly individuals come to an understanding of their own preferences, these data support the view that doing so draws on the ventral MPFC, and may rely on projective simulations of one's potential experience.
Recently, we have extended this suggestion by demonstrating that individual differences in the tendency to engage the ventral MPFC during judgements of future preferences correlate with rational economic decisions (Mitchell et al. submitted
). Research in behavioural economics has repeatedly demonstrated that individuals make decisions that maximize happiness in the present at the expense of one's future enjoyment. For example, when given the choice between $10 now and $12 in one week, people have a tendency to give up the larger, later reward in favour of the immediate payment. Likewise, we have a tendency to commit ourselves to future actions that we will regret when the actual time comes to carry them out. We may anticipate it being fun to travel to a conference or write a chapter in some months' time, but, upon actually having to carry through on our obligation, regret having consigned ourselves to spending time that we now see that we could use in other ways.
Such inaccurate predictions about our distant preferences may result from failure to project ourselves appropriately into the future. That is, we may fail to engage in a rich simulation of the concrete details involved in actually engaging in certain future activities (Trope & Liberman 2003
). To the extent both that the ventral MPFC subserves these kinds of simulations and that simulating the distant future will be more difficult than thinking about the here and now, we might expect to observe less activation in this area when perceivers report their preferences for the future compared with the present. We tested these predictions by scanning participants while they predicted how much they would enjoy a series of everyday activities (e.g. ‘browsing in a bookshop for 30
min’) at one of two time horizons: either ‘in the next 24 hours’ or ‘this time next year’. As predicted, significantly less ventral MPFC activity accompanied predictions about the far future, consistent with the possibility that participants failed to project themselves fully into another time and place when considering distant events.
More intriguingly, the extent to which this ventral MPFC activity differentiated between present and future preferences correlated significantly with participants' unwillingness to wait for a larger payment instead of taking a smaller, immediate one. Specifically, participants were offered the chance to receive a $10 gift certificate immediately or to wait one month for a larger amount of money. Although some participants were willing to wait a month for any larger reward (even $11), most would only wait when the later reward was substantially larger than the immediate pay-off. Critically, the minimum amount necessary for which a participant would wait (a measure of how impatient that participant was for the immediate reward) was directly related to the degree to which the ventral MPFC differentiated between judgements in the present and the future, suggesting that impatient participants may have failed to adequately simulate the experience of receiving the larger reward.