This final section of the chapter considers some contrasts between the version of simulation theory advocated here and the one proposed by Robert Gordon (1995b; 1996). One purpose of this discussion is to foreshadow and defend the upcoming treatments of other mentalizing topics in later chapters, specifically, first-person introspective attribution (chapter 9) and mental concepts (chap-ter 10).
Gordon (1995b) construes the standard version of ST (which he opposes) as involving three elements:
1. an analogical inference from oneself to others;
2. premised on introspectively based ascriptions of mental states to oneself;
3. requiring prior possession of the concepts of the mental states as-cribed (1995b: 53).
Gordon thinks ST should distance itself from all these elements. The ana-logical inference element, he fears, threatens to make ST collapse into TT, and the doctrine of introspection burdens ST with a philosophically controversial position. His rationale for opposing the concept-possession element is elusive, but he clearly resists it.
Gordon proposes an alternative version of ST not hobbled by these com-mitments. On the usual version of ST, when I set out to predict your decision, I imagine myself in your situation. That leads to a pretend decision to do A. I introspect this pretend decision state and ‘‘transfer’’ it to you, assuming you are like me. This formulation, says Gordon, allows ST to drift into prob-lematic waters.
When I simulate you, I do not imagine myself in your situation. Instead, what I imagine requires an ‘‘egocentric shift’’ or a ‘‘recentering’’ of the egocentric map. In imagination, the referent of the pronoun I becomes you.
Such recentering is a ‘‘transformation’’ of myself into you, much as actors become the characters they play. ‘‘Once a personal transformation has been accomplished, there is no remaining task of mentally transferring a state from one person to another, no question of comparing [the target] to myself’’
(1995b: 56). According to Gordon, when I recenter my egocentric map on you, I do not consider what I, AIG, would think, want, or decide. I consider what you would do, which frees me from the task of making an analogical inference from me to you.
In evaluating Gordon’s proposal, we must distinguish two questions: (a) Who is the subject of the imagining states and (b) what are the contents of the imagining states—including the labels or tags associated with them? The imagining states are, of course, actual states: The mentalizer actually engages in imagination or pretense. Moreover, the mentalizer is the subject of those states. She cannot literally transform herself, metaphysically, into the target of her attribution. She remains the same individual throughout the simulation exercise. On the other hand, she can label, or tag, her pretend states as belonging to somebody else. Indeed, a minimally competent simulator must do something like this (as noted originally in figure 2.3), especially if she attempts to track two or more targets concurrently. Pretend states must be tagged to keep track of their intended targets—and perhaps to keep them mentally distinguished from one’s own nonimaginary states. If we speak of the ‘‘content’’ of an imagining state in a broad sense, which includes the tag or label, then that content can be said to refer to the simulated target. When Gordon speaks of simulation as a ‘‘transformation,’’ one must be careful about its meaning. Gordon is right that within the content of an imagining state (the world of imagination), the pronoun I refers to the target. But the
real-world identity of the subject of an imagining state does not undergo any change in virtue of the subject’s imaginative act. The language of ‘‘trans-formation’’ is misleading on this point.
After presenting his ‘‘recentering’’ construal of the simulation routine, Gordon focuses on the alleged analogical inference from self to target, which he rejects. On this point we may agree. As discussed in section 2.3, it is questionable whether the movement from a pretend decision to do m to a belief that the target will decide to do m requires the use of an analogical premise, ‘‘The target is like me.’’ On the other hand, we reviewed empirical evidence in section 7.6 that people’s use of simulation does depend on judgments of similarity to the self. So it is unwise to insist on a version of ST that categorically denies this.
The second element of the ‘‘standard’’ ST view holds that attributors make an introspective identification of their own (final) pretend state and ‘‘transfer’’
it, or project it, onto the target. Gordon rejects this element as well. By contrast, I wish to retain the second element, including introspective identi-fication. Notice an important difference between the (pretend) state of de-ciding to do m and the final (genuine) state of believing that the target will decide to do m. Unlike the pretend decision, the belief state is a meta-representation. It represents another mental state, a future state of the target.
Moreover, it characterizes that state in terms of content and attitude type.
Where does this characterization come from? It isn’t provided by any prior element in the causal chain. This is clearly true of the attitude type: being a decision. The immediately preceding state, of course, is a decision (in pretend mode). But a decision does not characterize itself as a decision; it doesn’t metarepresent itself. So how does the system ‘‘know’’ from this decision that the target’s future state will be a decision? To select ‘‘decision’’ as an item of metarepresentational content, some classification process is required, perhaps introspection or self-monitoring. Thus, as I interpret ST, it naturally invites an introspective approach to first-person attribution, which will be defended in chapter 9. As we saw in section 7.6, there is empirical evidence that third-person mindreading involves use of a brain region responsible for self-reference or self-reflection. And research reported in sections 7.7 and 7.8 shows that failure to inhibit one’s own current states produces inaccurate, egocentric third-person mindreading. Normal inhibition of such states is presumably executed via the monitoring of those states.
Is the classification of one’s decision a classification of it as one’s own state, as a state of the self? This, of course, would lend support to the self-other ‘‘transference,’’ which Gordon seeks to avoid. The state being classified is certainly a state of the self. It is less clear whether an attributor ascribes it to the self. While in simulation mode, we have said, each state is ‘‘tagged’’ as belonging to some target other than the current self: either another person entirely or the future self. From this point of view, it seems difficult to hold
that the attributor ‘‘transfers’’ the state from self to other. Perhaps the best thing to say is that, while in simulation mode, pretend states are coded as belonging to something other than the current self, but when the simulation mode is exited, any (retrospective) representation of those states represents them as belonging to the (actual) self.
Finally, what about the third element in Gordon’s list of no-no’s: the prior possession of mental-state concepts? If mindreading includes mental-state attribution, if mental-state attribution is having a belief about a mental state, and if a belief about a mental state features mental-state concepts, then simulation-based mindreading (like all mindreading) requires mental-state concept possession. Because such concepts presumably aren’t created on the fly, prior possession of them is necessary. Because all these assumptions are entirely plausible, Gordon’s version of ST, which ducks the question of mental-state concept possession, is a tough row to hoe. On the approach favored here, ST ought to deliver a congenial account of mental-state con-cepts. I shall make a stab at such an account in chapter 10.
Notes
1. Actually, it is debatable whether even ordinary thinking about imagination re-quires it always to be under voluntary guidance. Budd (1989), White (1990), and Currie and Ravenscroft (2002) point out that a great deal of imagery—for example, dream imagery—isn’t subject to voluntary control. Because all imagery is the product of imagination, not all imagination involves voluntary control. This is undoubtedly correct, but I am not certain that it is recognized in ordinary thought. Indeed, I am not even sure it’s part of folk wisdom that dreaming is imagining.
2. The case of this patient might pose a challenge to the view that primary visual cortex is required for visual imagery, because the patient’s infarctions involved an almost complete destruction of primary visual cortex. However, destruction of pri-mary visual cortex was not complete. Islands of intact cortex were preserved.
3. In response to Chatterjee and Southwood (1995), Butter, Kosslyn, Mijovic-Prelec, and Riffle (1997) point out that their report of imagery in the face of cortical blindness did not use very sensitive measures of imagery, and the patients did have some intact medial occipital cortex.
4. Preserved imagery in the face of cortical blindness is not the only dissociation that poses prima facie difficulties for a vision–visual imagery correspondence thesis.
Cerebral achromatopsia (failure to see colors) can be dissociated from the capacity to have colorful images, hemispatial neglect can be manifested independently in vision and imagery, and visual agnosia can occur with intact imagery ability (see Bartolo-meo, 2002; Pylyshyn, 2003b: 405). But our discussion of the cortical blindness cases, including congenital cortical blindness, indicates that these various findings may not be insuperable.
5. Admittedly, Pylyshyn proceeds to speculate about the possibility that the re-ported finding doesn’t even involve the visual system, let alone a cortical screen (2003b:
397). However, this is just speculation and not terribly plausible in my judgment.
6. Elsewhere he says that images differ from vision in at least the following three respects: (1) unlike percepts, images fade rapidly; (2) unlike percepts, images are very malleable; and (3) images are created from stored information (Kosslyn, 1994: 74).
7. It goes without saying that they do not share all of their relational properties, specifically, their ways of being causally generated. On my account, having different types of causes simply follows from being images versus percepts.
8. M. Wilson (2003) discusses some other interesting details concerning Schwoebel et al’s patient CW, but these need not be pursued.
9. Currie and Ravenscroft (2002) argue that motor imagery is imagined percep-tion of movement rather than imagined acpercep-tion. ‘‘The reason we spontaneously de-scribe certain mental events as motor images is because those events seem, in various ways, like perceptions of movement, not because they seem like the movements themselves’’ (2002: 85). I would say that motor imagery is not the reproduction of perception of a movement but the reproduction of an inner, central event that plans, orders, or launches a movement. This explains the case of patient CW. The central event, when produced by imagination, is normally followed by inhibition, so motor production doesn’t occur. But intrinsically it is still very similar to a counterpart event that actually launches movement. In other words, an imagination-produced central event is a good surrogate of a normal, action-producing central event. One simulates the other.
10. Barsalou and Gallese, among others, cite these kinds of experiments as sup-portive of an ‘‘embodied’’ view of the mind (Barsalou et al., 2003; Gallese, 2003).
I am cautious of the embodiment theme, at least as the central construct for the explanation of mindreading. More precisely, embodiment seems to play a role in some but not all kinds of mindreading. In this context, it bears repetition that my use of the term enactment does not carry the connotation of motoric activity (as it does in the embodiment literature). As far as terminology goes, one may E-imagine desiring that p, for example, without engaging motoric systems.
11. The researchers were interested in perceived similarity judgments because, drawing on Heal (1986), they assumed that people use the simulation heuristic, or self-reflection, only when they judge their target to be similar to self.
12. Mitchell et al. (2005) do not claim that this wholly arbitrates between ST and TT, because some versions of TT, they say, leave open the possibility that knowledge about the self could be a useful basis for theorizing about others. But they don’t specify which version of TT they have in mind or discuss its viability. I raise questions about such self-oriented versions of TT in section 7.8.
13. However, there is independent evidence supporting this suggestion. For ex-ample, Ames (2004) found that people more readily project their own goals and pre-dilections onto similar targets (e.g., people sharing the same hobbies) than dissimilar ones.
14. One might wonder whether the Mitchell et al. (2005) study really tapped a high-level mindreading process. In the mentalizing task of their study, subjects viewed facial photographs and assessed how pleased the targets looked. Isn’t this a FaBER task, which we have treated as a paradigm of low-level mindreading? Assuming that
‘‘being pleased’’ qualifies as an emotion, it is true that this was a FaBER task.
However, we shouldn’t assume that the execution of all FaBER tasks has the earmarks of low-level mindreading. There may be no mirroring system associated with this task.
Indeed, the brain-scanning evidence points to the recruitment of the regions typically
involved in high-level mindreading (e.g., dorsal MPFC). This is the rationale for treating the study under the heading of high-level mindreading.
15. Thus, I am proposing that first-person future mindreading is just as open to simulation as third-person mindreading. That invites the question: What about first-person past mindreading? Could ST be extended to that domain? Karen Shanton (unpublished) argues that it can. She argues, first, that a large slice of first-person past mindreading uses autonoetic memory. Second, autonoetic memory is a matter of mental simulation, in the sense we have given it. Third, Shanton adduces striking evidence that projection effects of the kind reported in the remainder of this section also occur in autonoetic memory.
16. Notice, moreover, that whereas Saxe, Carey, and Kanwisher (2004) suggest that the processing of different mental states (e.g., knowledge, beliefs, desires) relies on distinct functional and neural mechanisms, Samson et al.’s (2005) data indicate that there may be common processes for the different mental states. That fits better with ST than TT, where the latter would tend to invoke different naı¨ve theories for different mental states.
17. This thesis has much in common with Jane Heal’s thesis that simulation in-volves ‘‘co-cognition’’ (Heal, 1998). However, Heal appears to equate co-cognition with simulation, which I think is a mistake. Although content-simulation is a funda-mental part of mindreading, it doesn’t exhaust mindreading. Ascribing attitudes and other mental-state types is also a core part of mindreading. In responding to Heal’s co-cognition thesis, Nichols and Stich (1998) complain that no actual theory-theorist denies the co-cognition thesis, or imputes to attributors massive content theoretization.
This is correct, but only because theory-theorists haven’t pursued the logic of TT to its natural conclusion. Just as full-tilt TT includes the thesis that human mentalizers represent attitude types in theorizing terms, so it should feature the thesis that human mentalizers represent attitude contents in wholly theorizing terms. That is a relevant thesis to dispute, as it is disputed here.
18. An example of theorizing about concepts in this fashion is Peacocke (1992).
19. Content holists will firmly deny that one person’s mental contents bear any close resemblance to another person’s mental contents, simply because each content in a person’s head is fixed by the entire web of other contents in her head. Because webs of content differ substantially across individuals, their respective contents will not be very similar (despite agreements in verbal utterances, for example). Two responses to the content holist are in order. First, it is a very questionable doctrine about the ‘‘real’’
nature of mental content (Fodor and Lepore, 1992). Second, even if holism is cor-rect about the real nature of mental content, naı¨ve cognizers, unaware of this philo-sophicoscientific truth, might proceed with a content-resemblance aim that guides simulation.
20. In a rather loose or weak sense of ‘‘theorizing,’’ in that no laws or general-izations are invoked by a definite description of the kind in question.
21. As used in this section, ‘‘projection’’ is not intended to connote inappropri-ateness. It is used in the more generic sense introduced in chapter 2.
22. In a paper published after this book was in press, Vann McGee (2005) presents a philosophy of language–based line of argument for the simulation, or empathy, approach (though his concerns are somewhat orthogonal to the present ones). McGee proposes a reconfiguration and solution to Quine’s problem of inscrutability of reference that prominently includes the assumption that third-person interpretation