Mirror neurons may strike people as mysterious. Where do they come from?
Are they innate? If so, how or why did they evolve? If not, what in ontogeny could account for them? Hardwiring initially seems like the most plausible hypothesis; it is widely regarded as the default. However, it is not firmly established. No experiments on infant monkeys have been conducted, for example, because it isn’t feasible. Moreover, a principled explanation in simple physiological terms has now been offered of how mirror properties could arise in ontogeny, specifically through Hebbian learning. Such an ex-planation is offered by Christian Keysers and David Perrett (2004). Although I do not endorse this explanation, it must be included as a serious possibility.
In his book The Organization of Behavior, D. O. Hebb (1949) offered the following postulate: ‘‘When an axon of cell A is near enough to excite cell B or repeatedly or consistently takes part in firing it, some growth or metabolic change takes place. . . such that A’s efficiency, as one of the cells firing B, is increased.’’ Put in a simple phrase, ‘‘neurons that fire together, wire together.’’
Modern neuroscience now strongly supports this speculation (Markram and Lubke, 1997).21Keysers and Perrett show how a circuit involving three areas in the brain of the macaque monkey can interact over time to acquire mir-roring properties; a similar circuit appears to exist in humans.
The relevant anatomical circuit in the macaque involves the superior tem-poral sulcus (STS), area PF of the inferior parietal lobule, and the premotor cortex (F5). STS, F5, and PF all provide high-level visual descriptions of observed actions. F5 and PF also have motor properties (as we have seen earlier); their visual and motor properties are usually congruent, which is why they are called ‘‘mirror neurons.’’ Here, according to Keysers and Perrett, is why the mirroring properties arise in this circuit.
Monkey (and human) infants observe their own actions carefully. During this process, some movements lead them to grasp objects in particular ways.
Action potentials in STS neurons responding to the sight of this type of grasping overlap in time with activity in the PF and F5 neurons that cause the infant to grasp in that way, for example, precision grip. This creates the prerequisites for Hebbian associations: Neurons that fire together, wire to-gether. For example, at first the STS? F5 connections are weak and unse-lective. But because the monkey is currently performing, say, a precision grip, the synaptic signal from the STS finds only the precision grip neurons to be active in F5. Hebbian learning will hence reinforce only the connections between precision grip in STS and precision grip in F5. Over time, neural associations strengthen to the point where neurons in F5 will respond to the visually driven synaptic input from STS. Many STS cells show viewpoint invariance, so the same STS neurons will respond both to the monkey’s own actions and to those of others. After ‘‘learning’’ the association between F5 motor commands and visual STS descriptions of the monkey’s own actions, the observation of someone else performing a similar action will also activate the neurons in F5, and mirror properties will have emerged.
This explanation is tendered for the emergence of motor mirroring prop-erties, but the main ideas could apply to other cases where brain areas are involved in both observing and experiencing an event. Keysers and Perrett sketch such accounts for other mirroring examples reviewed in this chapter, such as touch and disgust. The social properties of these systems—for ex-ample, our ability to empathize with other people without confusing our-selves with others—arise from basic anatomical facts and physiological principles.
I have argued that low-level simulational mindreading rests substantially on mental mirroring, and mirroring may be underpinned by Hebbian learning.
But isn’t Hebbian learning a kind of ‘‘theorizing,’’ at least a species of ‘‘in-ference’’? If so, isn’t there a new threat that simulation will collapse into theory? No. First, it is a dubious proposition that Hebbian association is a form of theorizing. Second, even if it is, we have already rejected the notion
(chapter 2) that any grounding of simulation in theorizing is ipso facto a threat to simulation. Here we have another case in point, where grounding poses no threat. The mere fact (if it is a fact) that mirroring emerges as a result of Hebbian learning doesn’t eliminate mirroring, doesn’t show that it’s a chimera. Moreover, mirroring is a clear form of mental simulation, under our definition, and it doesn’t cease to be so because of its etiology.
Notes
1. The first three sections of this chapter are adapted from Goldman and Sripada (2005). Thanks to Chandra Sripada for substantial contributions to this material, both scientific and philosophical.
2. This claim needs qualification for one case to be discussed in section 6.2.
3. Additional evidence that the amygdala is intimately linked to fear experience is the finding that when the amygdala is stimulated during surgery, fear is commonly evoked (Halgren, 1992).
4. For further details, see section 8.9.
5. As the discussion continues, I sometimes focus on the evidence involving FaBER tasks and ignore the guilt example. I don’t mean to discount the evidence in the guilt case, but the FaBER data facilitate a uniform treatment of the theoretical issues.
6. In the case of the psychopaths, it is less clear that they have a full grip on the concept of guilt. Their failure to attribute guilt correctly in story cases may indicate a deficient grasp of the concept.
7. In calling experience/observation matches ‘‘simulations,’’ I have in mind the
‘‘success’’ sense of simulation explained in chapter 2, where ‘‘success’’ merely means actual (rough) duplication, whether or not there is voluntariness or intent.
8. This did not permanently rehabilitate SM’s impaired recognition of fear, how-ever. When subsequently shown face stimuli under unconstrained viewing conditions, she failed to fixate the eye region spontaneously and reverted to impaired fear rec-ognition.
9. The ‘‘reverse simulation’’ idea is invoked by Blakemore and Decety (2001), in connection with their ‘‘forward-model’’ approach to intention attribution.
10. Robert Gordon (1995a) has discussed facial feedback as a mechanism of emotion contagion, citing early descriptions of the phenomenon by Hume.
11. However, section 6.7 sketches a possible story of the acquisition of mirroring properties, and this may contain hints about how mirror processes work.
12. The topic of classifying one’s own mental states is explored in chapters 9 and 10.
13. The study of ‘‘point-light walkers’’ was introduced by Johansson (1973), who attached small lights to the major joints of actors and filmed them walking or running in a dark room. These moving lights are immediately recognizable as human motion (often called ‘‘biological’’ motion).
14. If the reader isn’t prepared to concede that a process might consist in only one event, bear in mind that ‘‘simulation’’ didn’t have to be defined in terms of processes.
We might have defined simulation from the outset as a relation between either single events or sequences of events. Under this definition, the current matter wouldn’t be an issue. So, if the reader prefers, let this be our original definition.
15. Goldman and Sripada (2005) took no stance on which of the four models is correct.
16. A related tradition that also studies automatic mappings between self and other goes under the heading of ‘‘common-coding theory’’ (Prinz, 1990, 1997; Viviani, 2002; Preston and de Waal, 2002; Knoblich and Flach; 2001; Barsalou, Niedenthal, Barbey, and Ruppert, 2003). Its core assumption is that actions are coded in terms of the perceivable effects they should generate, and that perception of the behavior of another individual automatically activates one’s own representations of that behavior.
17. The existence of such a causal basis for attribution was already argued for, in effect, in sections 6.2 and 6.3. What is being added here is that the causal basis of attribution is associated with mirroring.
18. To be sure, the correlation doesn’t establish the causal status of the mirrored feeling.
19. Skepticism about primate mindreading was dramatically sparked by findings of Povinelli and Eddy (1996) that chimpanzees were unable to take into account what different experimenters could or could not see when choosing whom to ask for food.
Without training, they failed to distinguish between an experimenter whose head orientation was directed at them versus one directed away from them. Even after training, the apes continued to beg for food from experimenters unable to see them.
This led the authors to conclude that chimpanzees possess little understanding of the nature of visual attention, let alone mental states such as beliefs or intentions. More recent work by Hare, Call, and Tomasello (2001; Hare, Call, Agnetta, and Tomasello, 2000) and by Santos and colleagues (Flombaum and Santos, 2005; Santos, Flombaum, and Webb, in press) has turned the field around by obtaining new results that run in the opposite direction. Flombaum and Santos (2005) investigated whether macaque monkeys spontaneously took into account the direction of an experimenter’s gaze when attempting to steal food. By this technique, they probed monkeys’ abilities in a more naturalistic setting, where there is an element of competition. In one study, monkeys were shown two experimenters with a grape nearby, one who was facing the grape and the other with his back turned on the grape. The monkeys approached the experimenter who had his back turned. They also were able to use subtler cues, for example, selectively retrieving the grape from someone whose head and eyes were oriented away, whose eyes alone were oriented away, or whose gaze was blocked by a small barrier. Flombaum and Santos also found evidence for monkeys’ understanding that seeing leads to knowing (Santos et al., in press). Evidence for mindreading in birds has been obtained by Emery and Clayton (2001), who studied scrub jays. Scrub jays not only cache their food and remember where they cached it but also remember where conspecifics have cached and pilfer them when given an opportunity. Might they adjust their own caching strategies to minimize potential pilfering? Emery and Clayton found that jays with prior experience of pilfering another bird’s caches subsequently recached food in new cache sites, but only when they themselves had previously been observed caching. Jays without pilfering experience did not adjust their own caching strategies, even though they had observed other jays caching. These results suggest that jays relate information about their previous experience as a pilferer to the possibility of future stealing by another bird. This supports the notion not only that scrub jays engage in mindreading but also that they employ a simulation heuristic, because they project their own experience onto intentions of conspecifics. (Thanks to Randy Gallistel for pointing me to this last wrinkle.)
20. Onishi and Baillargeon’s (2005) work considerably advances the period of false-belief understanding. If they are right, false-belief understanding does not begin at the canonical age of 4, or even 3. So to speak of high-level mindreading as ‘‘late’’
may require some revision.
21. There are some skeptics about Hebbian learning. C. R. Gallistel, for example, questions the existence of this putative kind of learning, as well as associative learning in general. However, Gallistel has no quarrel with the neuroscientific evidence of changes in synaptic conductance more or less along the lines Hebb imagined; he questions only whether this constitutes learning (personal communication). However, for understanding the origins of mirroring properties, it doesn’t matter whether the hypothesized synaptic-change story is classified as an instance of ‘‘learning.’’ It still contrasts importantly with an innateness explanation.