participants were themselves touched and when they observed someone else getting touched by objects.
Finally, a mirroring or resonance system has been established for pain.
Mirror neurons for pain were first discovered serendipitously by Hutchison, Davis, Lozano, Tasker, and Dostrovsky (1999), while preparing a neuro-logical patient for cingulotomy. They found that a single cell responded both to painful stimulation (a pinprick) and to the observation of the same stim-ulation applied to another person (the medical examiner). More recently, three fMRI studies—one by Singer et al. (2004), one by Jackson, Meltzoff, and Decety (2004), and one by Morrison, Lloyd, de Pellegrino, and Roberts (2004)—report findings of pain resonance or mirroring. Singer et al. (2004) assessed brain activity while volunteers experienced a painful stimulus and compared it with the activity elicited when a signal indicated that a partner, present in the same room, was receiving a similar pain stimulus. Several brain regions—bilateral anterior insula (AI), rostral anterior cingulate cortex (ACC), brain stem, and cerebellum—were activated both when subjects re-ceived pain and when they rere-ceived a signal that a loved one experienced pain. These common areas do not exhaust the entire ‘‘pain matrix’’ but are restricted to that part of the pain network associated with its affective (rather than sensory) qualities. Extremely similar results were obtained by Jackson et al. Participants were shown photographs of hands and feet in situations that are likely to cause pain. Perceiving these third-person painful situations was associated with significant bilateral changes in activity in the ACC, the AI, the cerebellum, and the thalamus, regions known to play a significant role in one’s own pain processing.
recent theoretical statement, Gallese, Keysers, and Rizzolatti (2004) charac-terize mirror systems as providing a distinctive kind of social ‘‘understanding’’
that doesn’t require ‘‘conceptual’’ elements. In that article, they restrict them-selves to the question of how the meaning of an action is understood, not how an agent’s intention is captured (2004: 397).
Is there evidence, then, of mindreading based on mirroring? As indicated earlier, the definition of mirroring (as I construe it) does not imply mind-reading. This leaves open the possibility, however, that some mindreading is based on mirroring.
Very few of the studies that establish mirroring phenomena outside the emotion domain have included tests that provide evidence for or against mindreading. As an example chosen at random, the vision-of-touch study by Keysers et al. (2004) was a straightforward fMRI study that explored the brain areas activated when one is being touched and when one watches others being touched (more precisely, movies of other people or objects being touched). The study did not ask participants to judge or attribute any mental states to others, or seek to determine whether they spontaneously formed mental-state beliefs about the people observed. There are good reasons for this omission, no doubt. Questions of belief and ascription, especially mental-state belief and ascription, are not trivial matters to test for experimentally, and this research project had different fish to fry. Do other studies provide good evidence for mirroring-based mindreading?
We needn’t restrict the inquiry to evidence embedded in a single study; of equal interest is evidence that might be assembled from multiple studies. The Wicker et al. (2003) study of disgust plus lesion studies of selective im-pairment of disgust recognition jointly provide this type of evidence. The Wicker et al. fMRI study (along with similar ones that preceded it) clearly establishes a mirroring relationship between observing someone else exhibit disgust and experiencing disgust oneself. That study did not employ any tasks concerning mental-state judgments or attributions. But the lesion studies we reviewed clearly indicate that there is an association between damage to substrates for the experience of disgust and impaired interpersonal judgments of disgust (through faces and sounds). This association speaks strongly to a counterfactual relationship between the integrity of one’s own disgust sub-strate and a facility for interpersonal disgust attribution: ‘‘If one didn’t have an intact disgust substrate, one wouldn’t make disgust attributions normally.’’
This counterfactual, in turn, lends support to a causal connection between experiencing disgust while observing a disgust expression and attributing disgust to the person who makes that expression. Taken together, the evi-dence points toward the use of one’s disgust experience as the causal basis for third-person disgust attributions.17
For additional evidence, let us examine two of the pain studies reported previously. Singer et al. (2004) used a conventional fMRI technique and did
not employ tasks requesting attributions or ascriptions to the ‘‘other.’’
However, one important feature of their study provides evidence that the mirroring relationship is not, or need not be, a process of pure emotional contagion that entirely bypasses cognitive-level involvement. In their ex-perimental setup, subjects being scanned did not see the face of their partner or perceive any other emotional cue (though they saw the partner’s hand that would receive painful stimulations). The subject was shown an arbitrary cue on a large screen that signaled the partner’s feeling state. Thus, the subject had to infer the partner’s feeling state from the cue, a cognitive-level act that produced a resonant, or empathic, feeling of her own. Obviously, this doesn’t speak to the question of whether a mindreading attribution might result from a resonant experience. But it does show an interaction between mirroring-type phenomena and cognitive-level judgments. The Singer et al. manipu-lation is also relevant to a different question, namely, which simumanipu-lationist model of emotion attribution is most likely in FaBER tasks? The Singer et al.
study shows that mirroring of feelings can occur without the sorts of intermediaries postulated by Models 1 and 2, such as cues from facial musculature. A scanned pain subject did not see the partner’s face, so there was no stimulus that would prompt automatic facial mimicry that might launch a Model 2-like process.
More pertinent evidence for mindreading comes from the Jackson et al.
(2004) pain study. Here, subjects watched depictions of hands and feet in painful or neutral conditions and were asked to rate the intensity of pain they thought the person shown in the picture would feel. This intensity rating is a kind of third-person attribution task. One result of this experiment was that watching others in pain-inducing situations triggered a part of the neural network known to be involved in self-pain processing. This confirms the mirroring process for pain. More important for the present point, there was a strong correlation between the ratings (attributions) of pain intensity and the level of activity within the posterior ACC, a crucial part of the network for self-pain processing. This definitely confirms the idea that a mirror-induced, or resonant, feeling can serve as the causal basis of a third-person mindreading.18 A more recent article by Iacoboni et al. (2005), including members of the Parma group, provides evidence that intention ascription sometimes involves the motor mirror system. Iacoboni et al. begin by acknowledging that the basic properties of motor mirror neurons could be interpreted as a mechanism for recognizing merely the motor acts of a target, such as grasping, holding, or bringing to the mouth, not as recognizing or imputing to the target a mental state like an intention. True, mirror theorists often define ‘‘action’’ as im-plying an embedded goal, but it isn’t clear whether such goal-talk implies a mentalistic state. If this is implied by the definition, it is debatable whether mirror neurons recognize ‘‘actions’’ so defined. Iacoboni et al. therefore proceed as follows: If the mirror neuron system codes a goal beyond the
action itself—something that explains why the agent is doing the action—that should support the idea that the mirror neuron system ‘‘codes’’ a global intention. Is Mary grasping an apple in order to eat it, to give it to her brother, or to throw it away?
Iacoboni et al. noted that the same action in two different contexts may reflect two different intentions. So they investigated whether observation of the same grasping action, embedded in different contexts, elicited the same or different activity in an observer’s mirror neuron areas for grasping. If the mirror neuron system simply codes the type of observed action and its em-bedded goal, the neural activity in mirror neuron areas should not be influ-enced by the presence or absence of context. By contrast, if the mirror neuron system codes the global intentions associated with the observed action, then the presence of a context that cues the observer should modulate activity in mirror neuron areas. Thus, they had subjects watch three kinds of movie-clip stimuli: (1) grasping hand actions without any nearby objects (‘‘Action’’
condition); (2) context only, that is, scenes containing still objects like tea-cups and saucers (‘‘Context’’ condition); and (3) grasping actions performed in two different contexts, a during-tea scene and an after-tea scene (‘‘Inten-tion’’ condition). The third condition is called ‘‘Intention condi(‘‘Inten-tion’’ because each context suggests a further intention associated with the hand grasping—
either grasping a cup to drink or grasping it to clean up. In comparison with the other two conditions, the Intention condition yielded a significant signal increase in premotor areas already known to be mirror neuron areas. This increase suggests that this cortical area does not simply provide an action-recognition mechanism (‘‘that’s a grasp’’) but is critical for understanding the intentions behind others’ actions.
A further manipulation in this study speaks even more directly to intention attribution being mediated by the mirror system. Participants in the study received two different kinds of instructions. Half were told simply to watch the movie clips (Implicit task). The other half were told to infer the intention of the grasping action according to the context in which it occurred in the Intention clips (Explicit task). After the imaging experiment, participants were debriefed. Notably, all of them had clearly attended to the stimuli and could answer appropriately. They all associated the intention of drinking to the grasping action in the ‘‘during tea’’ Intention clip and the intention of cleaning up to the grasping action in the ‘‘after tea’’ Intention clip. This was true regardless of the type of instruction received. Moreover, the two groups of participants receiving different instructions had similar patterns of in-creased neural signal as compared with rest. Critically, the right inferior frontal cortex—the grasping mirror neuron area that showed increased signal for Intention as compared with Action and Context—showed no differences between participants receiving Explicit instructions and those receiving Im-plicit instructions. Thus, the mirror neuron areas made the same contribution
to interpreting intentions behind the grasping actions even when the partic-ipants were not explicitly instructed to infer intentions.
The foregoing discussion of intention attribution might make it sound as if the motor mirror system contributes to high-level, ‘‘propositionalized’’ in-tention states. Doesn’t this belie our claim that mirror systems contribute to a more primitive, lower level of mindreading? Further details from the Iacoboni et al. study support our original low-level interpretation rather than a high-level one. Iacoboni et al. postulate that mirror systems learn which motor acts commonly follow other acts to achieve a characteristic goal. One chain of acts might include grasping followed by drinking (bringing the cup to the mouth), and another chain might include grasping followed by cleaning. Intention attribution, at the level of mirror systems, may consist of coding an action as part of a probable sequence of motor acts with a final upshot (goal). This isn’t a terribly ‘‘high’’ level of cognitive functioning. When subjects are verbally debriefed, of course, higher cognitive centers are also brought into play, which characterize the behavior in fully propositionalized terms.
The Iacoboni et al. study of human intention understanding has been (more or less) duplicated by Fogassi et al. (2005) in the case of monkeys. This lends added support to the relative ‘‘primitiveness’’ of the cognitive level. It also introduces a topic we have thus far sidestepped: the question of mindreading in animals. Because of the massive literature on the subject, and the rapid changes it is undergoing, I have carefully avoided this terrain. But now, after a period of considerable skepticism about primate mindreading, there is new evidence—unrelated to mirror systems—that substantially supports nonhu-man primate mindreading. Indeed, there is evidence of mindreading even among birds, some of which might point in a simulation theoretic direction.19 To summarize, the case for mirroring-based attribution of mental states is quite strong, at least in three categories: emotions, feelings, and intentions. Be-cause mirroring is one species of interpersonal simulation, this lends weighty support to the thesis that low-level mindreading typically proceeds by simu-lation. I do not claim that whenever there is mirroring, there is also low-level mindreading. But wherever there is mirroring, the potential for simulation-based mindreading is there, and creatures with the requisite conceptual re-sources, especially humans, seem to exploit this potential extensively.