Part One: Mirror Neurons

Mirror neurons were initially discovered in premotor and inferior parietal regions of rhesus monkey brains (Di Pellegrino et al. Reference di Pellegrino, Luciano, Leonardo, Vittorio and Rizzolatti1992). Evidence of human mirror neurons was discovered soon afterward in homologous regions. Most neurons in these brain regions perform a single function: they communicate sensory information to the brain or transmit motor commands from the brain. Mirror neurons combine these functions: an individual observes goal directed behavior being performed by a second individual; the motor activation pattern in the observer's brain mirrors (matches) the pattern in the performer's brain; the process is not conscious and the motor behavior is not performed. Mirror neurons are multimodal: they respond to visual stimuli, auditory stimuli (e.g. the sound of a peanut shell being cracked open), imagined events (e.g. an athletic performance), and texts containing action verbs (Fadiga et al. Reference Fadiga, Pavesi and Rizzolatti1996; Fogassi et al. Reference Fogassi, Gallese and Rizzolatti2002 and Reference Fogassi, Ferrari, Gesierich, Rozzi, Chersi and Rizzolatti2005; Kohler et al. Reference Kohler, Keysers, Umlitá, Fogassi, Gallese and Rizzolatti2002; Rizzolatti and Arbib Reference Rizzolatti and Arbib1998; Rizzolatti and Craighero Reference Rizzolatti and Craighero2004; and Tettamanti et al. Reference Tettamanti, Buccino, Succaman, Gallese, Danna, Scifo, Fazio, Rizzolatti, Cappa and Perani2005).

Mirror neuron research on monkey brains is based on an invasive procedure: microelectrodes are inserted into the brain at selected points. For ethical reasons, this technique was not permissible for human studies. Consequently, human research was based on non-invasive techniques, fMRI and PET. These technologies can image the activation of populations of neurons but not single neurons. The problem with this technology is that the existence of mirror neurons can be inferred but not demonstrated. The original conception was that every mirror neuron is “strictly congruent” with an identifiable goal-oriented action. The current conception divides mirror neurons into four sub-populations: strictly congruent neurons; broadly congruent neurons (the correspondence is not precise); canonical neurons (these fire when someone observes himself performing a relevant action or observes an object to which this action might be directed, but not when he observes someone else performing the same action); and anti-mirror neurons (which inhibit the performance of mirrored action patterns) (Keysers et al. Reference Keysers, Kaas and Gazzola2010).

Evidence of human mirror neurons is necessarily indirect and contested. Critics have argued that the “mirror effect” that is observed via neuroimaging is the product of three sub-populations (sensory-motor, sensory only, motor only), no one of which corresponds to the original idea of mirror neurons (Dinstein Reference Dinstein2008; Dinstein et al. Reference Dinstein, Uri Hasson and Heeger2007; Dinstein et al. Reference Dinstein, Thomas, Behrmann and Heeger2008; Jacob Reference Jacob2008; Jacob and Jeannerod Reference Jacob and Jeannerod2005; Singer Reference Singer2006; Turella et al. Reference Turella, Pierno, Tubaldi and Castiello2009). The very existence of human sensory-motor neurons has been questioned. But a recent and unprecedented single-neuron investigation (Mukamel et al. Reference Mukamel, Eckstrom, Kaplan, Iacoboni and Fried2010), in which microelectrodes were inserted into the brains of surgical patients diagnosed with intractable epilepsy, is said to “finally provide direct electrophysiological evidence that humans have mirror neurons” (Keysers and Gazzola Reference Keysers and Gazzola2010, R353).

The mirror phenomenon is a brain-to-brain product, unmediated by mental states, and different from situations where a match between brains (observer and performer) occurs in parallel and is routed through consciousness. Here is an example: Participants in this experiment were inserted into an fMRI scanner, where they viewed photos of hands and feet in painful situations. They were next asked to assess the level of pain being experienced by the anonymous individuals. Participants’ assessments and brain images were compared. High assessments correlated with high activations in brain regions known (from previous research) to play a significant role in pain processing (Jackson et al. Reference Jackson, Meltzoff and Decety2005, 771). In other words, the distinctiveness of the mirror phenomenon depends on more than demonstrating neural matching: it is necessary to likewise demonstrate the absence of mental processing. To understand the efforts of researchers to connect mirror neurons to empathy, it is important to carefully distinguish between mirror states and parallel states. There are social neuroscientists who insist on this distinction (e.g. Singer et al. Reference Singer, Seymour, O'Dougherty, Kaube, Dolan and Frith2004), but others who do not. For convenience, I will call the former the “maximalists.”

According to the maximalists, since the frontal and parietal regions that are commonly associated with mirror neurons are connected to multiple brain regions, it can be presumed that mirror neurons extend to these regions as well (Keysers and Gazzola Reference Keysers and Gazzola2010, R353). The relevant research has focused on the role of mirror neurons in the social emotions, notably disgust. In the next paragraphs, I summarize the work published on this subject by maximalists (Vittoriao Gallese, Marco Iacoboni, Giacomo Rizzolatti, Christian Keysers and colleagues) during the period of most intense interest, between 2002 and Reference Gallese, Eagle and Migone2007.

Mirror neurons and disgust

“[T]he fundamental mechanism at the basis of the experiential understanding of others’ actions is the activation of the mirror neuron system. A similar mechanism, but involving the activation of viscero-motor centers, underlies the experiential understanding of the emotions of others.” Evidence obtained from multiple sources (electrical stimulation of monkey brains, task-driven fMRI research, clinical observations) indicates that these neural populations intersect in the insula. The insula “contains neural populations active both when the participants directly experienced disgust and when they understood it through the facial expression of others.” It is a “center of viscero-motor integration,” and a relay between action representations (mirror neurons) and emotion (Gallese et al. Reference Gallese, Keysers and Rizzolatti2004, 396, 398–400).

This “large-scale network,” composed of the mirror neuron system, the insula, and limbic structures including the amygdala, would permit an observer “to empathize with others through the representation and ‘inner imitation’ of the actions (facial expressions, body postures) of others” (Iacoboni and Dapretto Reference Iacoboni and Dapretto2006, 942; also Carr et al. Reference Carr, Iacoboni, Dubeau, Mazziotta and Lenzi2003). (This “inner imitation” or “action representation” is synonymous with the “activation pattern” that I have already mentioned.)

The process is demonstrated in “an fMRI study in which participants inhaled odorants producing a strong feeling of disgust. The same participants observed video clips showing the emotional facial expression of disgust. Observing such faces and feeling of disgust activated the same sites in the anterior insula. . . . [Just] as observing hand actions activates the observer's motor representation of that action, observing an emotion activates the neural representation of that emotion.” Activation was more intense when participants imitated the facial emotion than when they observed it. An analogous effect is observed in monkey and mirror neuron studies of goal directed behavior (Wicker et al. Reference Wicker, Keysers, Plailly, Royet, Gallese and Rizzolatti2003, 655; also van der Gaag et al. Reference van der Gaag, Minderaa and Keysers2007). In an interview in the New York Times, Christian Keysers explained that this neural network underpins additional empathic social emotions – his list includes guilt, shame, pride, embarrassment, humiliation, rejection, and lust – and also empathic responses to pain (Blakeslee Reference Blakeslee2006).

These researchers (Keysers, Kass, and Gazzola Reference Keysers, Kaas and Gazzola2010) welcomed the publication of Mukamel et al.'s single-neuron research, since it confirmed the existence of human mirror neurons and, more specifically, located them in regions beyond the frontal (premotor) and parietal regions of the brain, i.e. “supplementary motor area, and hippocampus and environs.” While these findings are consonant with the maximalist vision of mirror neurons, there is only indirect evidence connecting mirror neurons to the insula and, through this interface, connecting mirror neurons into a “large-scale network” and social emotions. The reasoning is illustrated in an fMRI study (Carr et al. Reference Carr, Iacoboni, Dubeau, Mazziotta and Lenzi2003) in which participants are asked to both imitate (a motor activity) and observe (a visuo-sensory experience) emotional displays. These are the findings:

1. 1. Imitating and observing the same emotion are shown to activate the same premotor areas. This finding is necessary, but not sufficient, for showing the presence of sensory-motor neurons and, more specifically, mirror neurons.

2. 2. Imitating an emotion is shown to activate additional brain areas, insula and amygdala, that are “relevant” to (unconscious) action representations. This finding is necessary, but not sufficient, for showing “the modulation of the action representation circuit onto limbic activity.”

3. 3. In other words: sensory stimulus → activation of mirror neurons → action representation → insula interface → limbic activation (includes the amygdala).

The lynchpin in this operation is the presence of “action representations” (Fadiga and Craighero Reference Fadiga and Craighero2004; Shmuelof and Zohary Reference Shmuelof and Zohary2007). I will continue with this point in the following section.

Mirror neurons and empathy

Although there is no standard definition of empathy among social neuroscientists, there is a general understanding that “empathy” refers to a state that is shared by an observer and another individual, present or imagined. Empathic states include, singly or in combination, cognitive empathy, emotional empathy, and somatic empathy (mainly pain). Human mirror neurons are routinely depicted as being intrinsically empathic (e.g. Iacoboni and Dapretto Reference Iacoboni and Dapretto2006; Iacoboni Reference Iacoboni2009; Rizzolatti and Craighero Reference Rizzolatti and Craighero2004; Nummenmaa et al. Reference Nummenmaa, Hirvonen, Parkkola and Hietanen2008).

Motor theories of cognition date back to the nineteenth-century (James Reference James1890, on ideomotor theory); mirror neurons provide the latest chapter in this history (Jeannerod Reference Jeannerod1994; Hickcok Reference Hickcok2009). Until recently, interest in empathy has concentrated in the fields of developmental psychology, social psychology, counseling psychology, and in clinical investigations of certain psychiatric disorders, notably childhood autism and personality disorders. Biological research on empathy focused on limbic system and neuroendocrine responses to social signals such as expressed emotion (see Brothers Reference Brothers1989 for a review). The discovery of the human mirror neuron system has expanded the neurological frontiers of empathy by identifying a biological mechanism that appears to illuminate, if not answer, one version of an enduring philosophical question, “the problem of other minds.” How do minds infer intentions embedded in other minds (Custers and Aarts Reference Custers and Aarts2010, 49)?

The simplest form of mirror neuron activation (and action representation) is referred to as “resonance” and it is unconscious. The human mirror neuron system is capable of “de-coupling” an action representation so that it can be projected (attributed) back to its source, the performer, thereby differentiating between self and other. Recent research suggests that sometimes de-coupling is spontaneously reversed during memory reconsolidation (retrieval). This occurs in instances where an observer remembers someone else's performance (via an action representation) as having been his own experience – a phenomenon that is conventionally called “source amnesia” and “false memory” (Lindner et al. Reference Lindner, Echterhoff, Davidson and Brand2010).

The mirror neurons and action representation are sufficient for inferring the performer's intention. I put it this way – failing to indicate who or what is doing the inferring – to make a point. Many theorists presume that mirror neurons possess this capacity: that is, they can interpret action representations and detect intentions without cognitive processing by other parts of the brain. It is possible that certain non-human species are capable of doing this, but there is much disagreement on this point. It is clear that only humans are capable of the final stage, in which de-coupled representations can be objectified, i.e. detached from their source (performers and original performances) and stored for future use – as templates and prototypes for imitation, emulation, analogical reasoning, and sophisticated forms of mind-reading. While language obviously facilitates objectification, researchers are divided as to whether it was essential for the emergence of this capacity, since pre-linguistic babies are demonstrably capable of performing these functions.

The final process, the basis of cognitive empathy, operates with “different levels of action representation, from the motor intention that drives a given chain of motor acts to the propositional attitudes (beliefs, desires and so on) that . . . explain the observed behaviour in terms of its plausible psychological reasons.” The end-product is an over-arching “mentalizing network” that connects the mirror neuron system – whose action representations provide “a subpersonally instantiated common space” between observer and performer and a scaffold for “bootstrapping mutual intelligibility” (Gallese and Goldman Reference Gallese and Goldman1998; Gallese Reference Gallese2001 and Reference Gallese2003) – to high-level operations (reasoning in cortical areas without demonstrable mirror properties (Corrado 2010, 264; also Kilner and Frith Reference Kilner and Frith2008). Evidence connecting mirror neurons to the putative network and other brain regions is, once again, “indirect,” e.g. bolstered by efforts to discover mirror neuron abnormalities in people diagnosed with autism spectrum disorders, a diagnosis associated with poor mind-reading abilities and other deficits in cognitive empathy (Hadjikhani et al. Reference Hadjikhani, Joseph, Snyder and Tager-Flusberg2006; Oberman et al. Reference Oberman, Hubbard, McCleery, Altschuler, Ramachandran and Pineda2005; Rizzolatti and Fabbri-Destro Reference Rizzolatti and Fabbri-Destro2010).

Controversy

There are obvious similarities between human mirror neurons, as I have described them, and the mental modules delineated by Jerry Fodor (Reference Fodor1983) and later elaborated by evolutionary psychologists (Barkow et al. Reference Barkow, Cosmides and Tooby1992). The modules and neurons respond spontaneously to domain-specific sensory inputs that require no cognitive processing. Their operations are depicted as automatic (obligatory), rapid (unimpeded by consciousness), and encapsulated (invisible to reflective consciousness). Each is associated with a dedicated neural architecture and a characteristic pattern of decomposition (pathology). And they are both products of evolution and adaptation. Commonality is exemplified by the so-called “intentionality detector” (a function required for mind reading) that evolutionary psychologists conceive as a mental module and neuroscientists attribute to human mirror neurons. The intentionality detector also highlights an important difference, affecting the way in which the two perspectives are connecting the mind to the brain: mental modules are connected to the brain metaphorically (e.g. via “inference engines” and “mental organs”); mirror neurons are biological mechanisms that uniquely bridge the gap between mind and brain (via action representations).

While the discovery of a visible bridge connecting mind to brain and the possibility, perhaps inevitability, of reducing mind to brain, has stimulated intense popular interest in mirror neurons, these claims have likewise drawn the attention of skeptical researchers. Skeptics have three concerns. Do mirror neurons exist? (Mukamel et al. Reference Mukamel, Eckstrom, Kaplan, Iacoboni and Fried2010 has resolved this concern for the time being). If they do, do they possess this list of module-like features? Is there convincing evidence that mirror neurons can independently infer intentions?

To answer these questions, one should begin with the lynchpin of mirror neuron operations, the notion of “action representations.” The cognitive science view of the central nervous system is that it is a dynamic apparatus comprising parts and processes that respond adaptively to a dynamic external world, a process facilitated by a sensory-motor loop. Adaptation is constant and requires the nervous system to construct and continually update its representations of the external world (Boden Reference Boden2006, 1178–1179). Mirror neurons are presumed, by advocates, to be pre-adapted components, and their action representations are static, unaffected by experience or memory. During interactions with the external environment – observing behavior – mirror neurons form “logically related” chains of representations that, in turn, make it possible to represent performers’ intentions (Iacoboni et al. Reference Iacoboni, Molnar-Szakacs, Gallese, Buccino, Mazziotta and Rizzolatti2005). The process takes place outside of consciousness, without the interference of mentalizing structures. This “logic” is unlike propositional logic, and Gallese identifies it with the brain's “forward model architecture.” This model is a “theoretical” system that describes how motor output and bodily movement is regulated in vertebrate species. Thus:

The job of the system represented in the forward model is to compare its established predictions/representations with internal sensory feedback from the external environment (“inverse model representations”). When a mismatch between predictions and real-time feedback exceeds a predetermined limit, the forward model readjusts the motor command. Following sufficient cycles, a revised prediction/representation is established. The forward model logic – thus logical relations among mirror neuron representations – has been appropriated from the field of control engineering, where the model is employed to manage dynamic constellations of internal forces, resistances (intrinsic to all mechanical systems), and feedback from the external world (Schwartz Reference Schwartz1999; Kosslyn Reference Kosslyn2005; Boden 2008, 1184; de Vignemont and Haggard Reference de Vignemont and Haggard2008) systematically decomposes the non-propositional logic attributed to mirror neurons.

There are prominent neuroscientists who reject this account, on the grounds that mirror neurons cannot independently infer intentions because they do not possess the modular properties that make this feat conceivable. Thus according to Heyes and colleagues, humans are not born with mirror neurons; while human mirror neurons do emerge, they are not pre-adapted to anything in particular. Evolution provides us with “motor neurons that become mirror neurons, but it has not selectively established links between visual and motor neurons coding the same action” (Heyes Reference Heyes2010, 790). These neurons form during each individual's development via a familiar Hebbian process, the idea that “neurons that fire together wire together” (Heyes Reference Heyes2010, 789; Hickok Reference Hickcok2009). Similarly, it is mistaken to suppose that these neurons are capable of representing a movement or object without a context. Hume correctly argued that meaningful perceptions presuppose an act of interpretation and contextualization, and this act relies on memory and associations acquired through experience (Brass et al. Reference Brass, Schmitt, Spengler and Gergely2007; Csibra and Gergely Reference Gergely and Gergely2007). Consider an experiment conducted by Gergely et al. (Reference Gergely, Bekkering and Kiraly2002). Preverbal babies observed a goal-directed action: a seated adult activated a light-box by leaning forward and touching the box with her forehead. At one session (context), her hands were encumbered and unavailable; at a second session, her hands were free. Two-thirds of the babies imitated the unusual head-down behavior after watching the hands-free demonstration; just one-fifth imitated this behavior after observing the hands-occupied demonstration. In the first context, babies “inferred that the head action offered some advantage,” since the demonstrator could have turned on the light in the familiar way, using her hands, if she had wished. In the second context, most babies chose to emulate motor actions in their repertoire (using their hands), after “presumably concluding that that the head action was not the most rational.”

Part Two: A neo-Darwinian back-story

A narrative's “back-story” is a fictive or notional history that precedes the events described in the narrative. An effective back-story makes fictional characters “recognizable,” and creates expectations in the reader that the author can manipulate as the narrative develops. A prologue is an overt back-story; the “introduction” section in science publications performs a similar function. The invocation can be didactic, as when a fictional character overtly reflects on his past, or subtle, as when back-stories unfold in a series of clues and strings of bibliographic citations. The effectiveness of these clues for creating a sub-text depends partly on the reader's background knowledge.

This neo-Darwinian back-story is a plausible genealogy of the social brain. By the end of the Paleolithic era, the course of social and biological evolution has produced the “psychologically modern” human and the cognitively demanding kinds of reciprocity that are both source and product of the social brain. This story has been (and continues to be) erected piece-meal: the work of multiple authors, each tackling a discrete puzzle, with no collective goal in mind. The story is compelling because it is consonant with the available empirical evidence and because the constituent puzzles and solutions are logically connected. The story is also implicit, in that it is part of the collective consciousness of social brain researchers, but there is (as yet) no occasion when it is retold in its entirety. Individual investigators are not necessarily familiar with, or subscribe to, all of the episodes. While individual episodes can be (and are) operationalized through empirical research, the entirety – extending perhaps seven million years – can only be inferred, and a researcher can justifiably claim that her work is confined to the topic that currently attracts her attention, e.g. the intersection of affective empathy and perceived fairness. One might compare this back-story with the idea of myth described in Paul Veyne's book, Did the Greeks Believe in their Myths? An Essay on the Constitutive Imagination:

Before proceeding to the back-story, I want to clarify what “neo-Darwinian” means in this context of the social brain. Beginning in the 1950s, John Maynard Smith, an evolutionary biologist, pioneered an empiricist approach to natural selection that combined the methods and perspectives of population genetics, game theory, and cost-benefit analysis. The approach was the basis for a dialectical understanding of human biological evolution and, we shall see, a back-story for the neuroscience revolution (Maynard Smith Reference Maynard Smith1979). In an article, “Reconciling Marx and Darwin,” published shortly before his death in 2004, Maynard Smith recalls Marx's thesis on human consciousness – thoughts, inferences, perceptions, and desires – the view that consciousness is determined by the material conditions of social life. Marx might reasonably claim to be a materialist, Maynard Smith writes, but he was not a reductionist, someone interested in investigating the biological evolution of social life and the brain that made human consciousness possible in the first place. Marx's vision of human nature as malleable – a view shared by many cultural anthropologists during the 1950s – was unscientific and proved to be “tragically flawed” and a “manifest failure” when put into action by communist regimes.

The neo-Darwinian back-story for the social brain begins with two puzzles. It is propelled forward, from the emergence earliest hominids (perhaps five million years ago) to the advent and dispersion of psychologically modern humans, by a dialectical logic. I will consider each puzzle in turn. The first puzzle concerns altruism. Population geneticists define altruism as behavior that transfers some or all of the altruist's reproductive potential (fitness) to a beneficiary. In the most extreme case, an altruist dies so that a beneficiary might live and reproduce. Altruistic behavior is often observed in animal populations, and presumed to be genetically determined. It is a puzzle because it gives the non-altruist recipients (lacking altruism genes) a reproductive advantage. They will eventually outbreed altruists, and altruists and altruism genes will eventually disappear from the population. This does not happen however and this is the puzzle. W.D. Hamilton solved the puzzle in the 1960s with the theory of kin selection (inclusive fitness). The theory says that altruistic behavior, including self-sacrifice, has favorable cost-benefits if the beneficiaries are closely related to the altruists. When this happens, altruism preserves a homogeneous gene pool and a continuing supply of altruism genes for future generations.

Hamilton's solution leads to a new puzzle, once biologists discover populations in which altruistic behavior is common but the altruists and beneficiaries are often unrelated or too distantly related for Hamilton's solution. Robert Trivers, an anthropologist, solved this puzzle in Reference Trivers1971 by demonstrating that the arrangement works (benefits are greater than costs) if altruism is reciprocal. The recipient repays the altruist. Reciprocal behavior is common among non-human primates, but limited to individuals living in close proximity. It is typically restricted to grooming and, less often, food-sharing, and the interval between gift-giving and repayment is brief, a few seconds in the case of capuchin monkeys and just minutes with chimpanzees. More complex forms of reciprocity would involve dispersed partners, and long intervals between gifts and repayment. In time, multiple local groups would be connected, and extensive social networks and coalitions would develop. But these developments depend on the ability of the individuals to keep track of their exchanges and the reputations of potential exchange partners. Thus memory-based reciprocity represented a great leap forward, but was cognitively demanding, and presupposed various developments in the nervous system, including the capacity to inhibit the archaic impulse for immediate gratification.

When altruism and reciprocity are limited to a genetically homogeneous population, cheating is impossible. Of course there will be non-reciprocators and their reproductive fitness will benefit from the costs absorbed by the altruists. But the costs and benefits circulate within a closed system. The term “cheater” is reserved for genetically heterogeneous populations, in which costs and benefits can accumulate within different lineages – circumstances that emerge together with memory-based reciprocity. Once again, there is a solution (reciprocity) that creates a problem (cheaters) and a puzzle. Cheaters are inevitable because all organisms are driven to maximize their reproductive interests and the interests of close kin – a neo-Darwinian premise. The cost-benefits of cheating always favor cheaters. All things being equal, cheaters (non-reciprocators) will outbreed the reciprocators; the nascent social networks will collapse; human social evolution will progress no farther (Nowak and Sigmund Reference Nowak and Sigmund2005; Rosas Reference Rosas2008). Of course the collapse did not occur. This is a puzzle. There is a neo-Darwinian solution, and it leads to a further contradiction. Puzzle ⇒ solution ⇒ contradiction ⇒ puzzle ⇒ etc.: this is the dialectic that glues the back-story together.

In this case, the solution is in two parts. The social fabric is preserved by the evolution of positive sentiments that include friendship, gratitude, and sympathy. Fast forward to the present time: evidence from game-playing experiments, mathematical modeling and simulation, and ethnographies of small-scale, pre-industrial communities indicate that positive sentiments are insufficient to prevent the collapse. Something stronger is needed and the evidence points to a gene-driven impulse to punish cheaters. The solution is efficient except in one respect: it cannot pay for itself. The fitness costs of being an enforcer exceed the benefits. Punishment consumes resources (e.g. energy) and can be terminally expensive if cheaters retaliate violently. For this reason, the enforcer's behavior is labeled “altruistic punishment.” There is a further problem or contradiction, in that punishment creates a new class of cheaters, namely friends and neighbors who are good reciprocators but unwilling to be enforcers. These are “second order cheaters”; they get the cost-free benefits of punishment and eventually they should outbreed gene-driven enforcers (Boyd et al. Reference Boyd, Gintis, Bowles and Richerson2003).

Human nature and empathic cruelty

Punishment can prevent entropy, but why would a rational individual – someone innately self-interested and able to estimate cost-benefits – become an enforcer? Benefits are often hypothetical (scheduled to arrive in the distant future) and indirect (dissuading potential cheaters), and costs are unpredictable (perhaps bringing retaliation by the cheater or his kin). Even when an enforcer gets his fair share, he cannot know whether this would happen without his intervention. Thus material rewards can provide only a weak motive for altruistic punishment.

A solution is described in “The neural basis of altruistic punishment,” published in Science (de Qervain et al. Reference de Quervain, Fischbacher, Treyer, Schellhammer, Schnyder, Buck and Fehr2004). The experiment is bit of ontological theater, based on the “trust game.” The game comprises interaction between two anonymous participants, A and B. Individual A plays the game seven times, facing seven different Bs. Each participant is given 10 “money units,” convertible into Swiss francs. The rule is that A can transfer all or none of his units to B. If he gives 10 units to B, they are quadrupled, so that B now has 50 units (40 plus his original 10). B can now send 25 units to A (“fair” behavior), or he can keep the entire 50 for himself (“unfair”). A is now given an additional 20 units. If A now wants to punish B, he can do this in three ways: 1. he can buy “penalty points” (each point costs A one unit, and subtracts two units from B); 2. he can reduce B's sum cost-free (A does not have to pay for the penalty points); or 3. he can reduce B's sum symbolically (B is informed of A's decision but loses no units). Each game is limited to just one formula (1, 2, or 3), designated by the researchers. There is an additional variable: in some games, a “random device” determines whether B's choice will be fair or unfair. A knows the rules obtaining during each game: that is, whether his option is 1, 2, or 3, and whether B's choice was decided by the random device. Once A learns B's choice, he has one minute to make a decision – whether to punish, how much to punish. During this one-minute interval, his brain is scanned via PET. A's brain images reflect his deliberations and his anticipation (imagining) of B's response.

Thus the experiment reproduces moments in the evolutionary narrative –the punishment of free-loaders by enforcers is simulated when B is unfair and A penalizes him; altruistic punishment is simulated by option 1 (A buys B's penalty points). The PET images of enforcers’ brains show activation of the caudate nucleus of the dorsal striatum, a region is associated with dopamine excretion and part of the brain's “reward center.” The images also indicate that the intensity of the activation correlates positively with the severity of the punishment – the number of penalty points that A assigns to B. Thus the images show that when A is paying out units to punish B, A's brain is paying itself back with pleasure. In which case, (altruistic) punishment is its own reward.

The Concise Oxford Dictionary defines “cruelty” as “having pleasure in another's suffering.” If so, the back-story permits the conclusion that cruelty entered human nature along with our felicitous pro-social disposition to altruistic punishment. (This is an ethnographic observation, not a moral judgment.) The enforcers in the experiment actively imagined the disappointment that will be experienced by the defectors; they vicariously participated in the defectors’ state of mind. This is not simple cruelty but, more precisely, empathic cruelty.

Research by Takahashi et al. (Reference Takahashi, Kato, Matsuura, Mobbs, Suhara and Okubo2009) on Schadenfreude is even more explicit. According to Takahashi, Schadenfreude (pleasurable response to news that a misfortune has fallen to a person who is envied or resented) and envy (a painful feeling of inferiority and resentment that results from awareness of someone else's superior quality, achievement, or possessions) are two sides of one coin. Students in the experiment read descriptions of three fictive students – A, B, and C – and were instructed to see B and C from A's perspective. Student A had average abilities, achievements, possessions, social endowments and prospects. Student B was superior and successful in each respect and also in life domains important to A. Student C was superior and successful but in domains not important to A. Participants silently read additional texts pertaining to A, B, and C while their brains were scanned (fMRI). In phase one, the texts described the successes of B and C, and participants reported how envious the descriptions made them feel. In phase two, texts described misfortunes that spoiled events and prospects for A, B, and C. Participants reported the intensity of their pleasure (Schadenfreude) regarding each of the events. Brain images and self-reports were compared.

Empathic cruelty has three elements: the target's present or anticipated distress, the perpetrator's pleasure (reward), and empathy (shared pain, distress, etc.). In Takahashi's experiment, brain images demonstrated the pleasure part directly, via the activation of the participant's reward center (dorsal and ventral striatum and medial orbitofrontal cortex). Evidence of empathy is more complicated. Prior research shows that cognitive conflicts and social pain are processed in the dorsal anterior cingulate cortex (dACC), part of the brain's “pain matrix.” Brain images of the participants in Takahashi's experiment showed activation of the dACC after reading the texts. This activation had two sources: the participant's envy (phase one) and his internal representations (imagination) of the target's distress (phase two). Thus the brain images are evidence of shared affect identified with empathy.

To the extent that envy is an expression of frustrated entitlement, the envied person (target) in Takahashi's experiment resembles the unfair person in de Quervain's experiment. No data relating to empathic affect was collected by de Quervain et al. Participants were simply asked to describe their own feelings while deciding to punish unfair participants (for additional research relating to empathic cruelty, see Fehr and Camerer Reference Fehr and Camerer2007; Fliessbach et al. Reference Fliessbach, Weber, Trautner, Dohmen, Sunde, Elger and Falk2007; Knoch et al. Reference Knoch, Pascual-Leone, Meyer, Treyer and Fehr2006; Lanzetta and Englis Reference Lanzetta and Englis1989; Shamay-Tsoory et al. Reference Shamay-Tsoory, Fischer, Dvash, Harari, Perach-Bloom and Levkovitz2009; Singer et al. Reference Singer, Seymour, O'Dougherty, Stephan, Dolan and Frith2006).

Empathy and evolution

The neo-Darwinian back-story of the social brain starts with two puzzles: one puzzle concerns altruism, the second concerns the brain. The size of the human brain is an evolutionary puzzle. Our ancestors split from the great apes six million years ago. During this period, the ancestral human brain quadrupled in volume. The metabolic cost of the human brain is enormous: it constitutes 2 per cent of total body weight and consumes 15 per cent of cardiac output and 20 per cent of body oxygen, and these demands are ceaseless and inflexible. We can assume that the bigger brain paid for itself, yielding favorable cost-benefits. However efforts to model these developments indicate that increasing metabolic cost would eventually exceed benefits. Why did the brain continue to grow (in size and power) when further growth was no longer adaptive? This is the puzzle, and its solution is a story about how brains adapted to other brains.

The process is a cognitive arms race (Byrne and Whiten Reference Byrne and Whiten1988; Barton and Dunbar Reference Barton, Dunbar, Whiten and Byrne1997; Dunbar Reference Dunbar2003). It begins with the emergence of a “mind-reading” capacity: the ability to detect the intentions and predict the behavior of other individuals. Mind-reading facilitates more complex social relations, but it also facilitates “cheaters,” individuals who use mind-reading to manipulate and deceive other individuals. All things being equal, cheating is cost-effective: it gives cheaters a reproductive advantage. As cheaters increase as a proportion of a population, social life grows more unpredictable, thus undermining the stability of social relations. Entropy looms. This might have been the evolutionary fate of our hominid ancestors, but it was not. Entropy was avoided through a further improvement to the neural hardware: the emergence of a “cheater detector.” This could be no more than a temporary solution however. The next generation of opportunists used their improved brains to subvert the function (or adaptation) of the cheater-detector. Entropy was avoided by the evolution of an improved version of the cheater-detector that would, of course, facilitate the emergence of a cohort of improved cheaters. And so on over millions of years, following the dialectical logic of the neo-Darwinian back-story.

This part of the story begins with mind-reading. Opinion in neuroscience is divided on the biological basis of this ability. There are rival explanatory accounts of mind-reading: a version based on mirror neurons and a version based on cognitive mechanisms, including analogical reasoning, as mentioned by Hume. In both versions, the evolution of empathy is understood to be a pro-social development: “the ‘glue’ of the social world, drawing us to help others and stopping us from hurting others” (Baron-Cohen and Wheelwright Reference Baron-Cohen and Wheelwright2004, 163; see also Lawson et al. Reference Lawson, Baron-Cohen and Wheelwright2004; Baron-Cohen et al. Reference Baron-Cohen, Knickmeyer and Belmonte2005; Wheelright et al. Reference Wheelwright, Baron-Cohen, Goldenfeld, Delaney, Fine, Smith and Wakabayashi2006; Williams et al. Reference Williams, Whitten, Suddendorf and Perrett2001; Iacoboni and Dapretto Reference Iacoboni and Dapretto2006). According to Simon Baron-Cohen, an authority on autism, human evolution produced two (polar) kinds of brains: a female brain with highly developed empathic capacities and a male brain adapted to manipulating objects and creating systems. Empathy originated as a pro-social adaptation allowing females to detect the wants of pre-verbal children and the moods of the potentially dangerous males with whom they lived (see also Hrdy Reference Hrdy2009). Autistic individuals are characteristically poor empathizers, and autism's epidemiology is biased towards males: 10 to 1, in high functioning autistic disorder.

In both versions, there is the further assumption that empathy is intrinsically a morally positive disposition. According to Baron-Cohen, we respond to suffering in three ways: the response mirrors the sufferer's distress (we experience it); the response is culturally appropriate (e.g. pity) but does not mirror the suffering; or the observer takes pleasure in the sufferer's condition. Baron-Cohen equates “empathy” with the first two responses, and explicitly excludes the third. De Quervain's research vindicates a further possibility – empathic cruelty – in which the observer's brain mirrors the sufferer's distress and also takes pleasure in the sufferer's condition.

This is an unexpected twist in the evolutionary back-story: evidence of an evolved empathic disposition that is simultaneously pro-social and cruel. Human nature in the age of neuroscience grows morally complex: a theme that recurs in the story of the cognitive arms race, where an uncomplicated kind of cheating, the refusal to reciprocate, evolves, via mind-reading, into deception, the effort to represent the current situation as something different from the reality. Deception puts great demands on the brain, which is now required to do two things simultaneously. An individual must construct a lie and also withhold (inhibit) the truth. Experimental and clinical evidence suggests that telling the truth is the brain's default response, a legacy of a pro-social evolutionary adaptation. “Responding with a lie demands something ‘extra’, and . . . will engage executive prefrontal systems [responsible for planning, decision-making, and monitoring] more than does telling the truth” (Spence Reference Spence2004, 8; see also Spence et al. Reference Spence2004). Thus a deceiver is in constant danger of signaling the truth and betraying himself, for instance, by involuntarily and detectable hesitation preceding a lie. To succeed, deception requires a capacity for self-deception, the ability to conceal one's true intentions and facts from oneself (Trivers Reference Trivers1971). At this point, two unprecedented objects are created – there is the other (created via projection) and the self (refracted in a doppelganger, the fraudulent self).

Empathic time travel

The evolutionary origins of the self are brought to life in research on “mental time travel” (Suddendorf et al. Reference Suddendorf, Rose Addis and Corballis2009). Mental time travel is the capacity to project one's self into situations in the past, future, and subjunctive (an alternative scenario to the actual past or present). Travel to the past evolved first and provided a prototype (an episodic memory in which the thinker is spectator or protagonist) for constructing mental representations of possible futures and also alternative presents and pasts.Footnote ¹ The emergence of language – notably pronouns and verb forms – and empathy were prerequisite for the evolution of mental time travel (Corballis Reference Corballis2009). The idea that we simply “project” ourselves wholesale into the past or future oversimplifies time travel. It is a first-person experience that requires the splitting of the self: one must be both here-and-now and there-and-then at the same time. The bond between the split-selves is empathic, but not necessarily positive. (In this sense, time travel parallels Trivers’ account of self-deception.)

Neuroscience makes it possible to see time travel: pain provides an efficient modality. Jean Decety and his collaborators conducted fMRI experiments in which they asked participants to imagine themselves and others in painful situations (Decety and Grèzes Reference Decety and Grèze2006). In other words, participants traveled to the subjunctive (an alternative present-time). One expects that, in some participants, the targeted situations stimulated spontaneous time travel to other places as well – notably travel to intensely empathic memories involving loved ones in pain. The imagined situations activated – one might say “mirrored” – brain regions reliably associated with experiencing the emotional content of pain in present-time. There was no phenomenological confusion between the mental act and the experience however. Neuroimages in the self vs. other scenarios were similar but, as one might expect, there were discernible differences – which is what one would expect, given that “a minimal distinction between self and other is essential for social interaction in general and for empathy in particular.”

In the neo-Darwinian back-story, human evolutionary history begins with a great leap forward. Up to this point, social exchange is based on altruism and simple kinds of reciprocity between genetically similar individuals. Afterwards, relations take the form of networks of exchange (reciprocity) among individuals genetically unrelated or only distantly related. Many problems were encountered along the dialectical road to modern times. The “future” was one of these problems. Where reciprocity entails long delays between giving and repayment, exchange partners must share some awareness of “time,” as a continuum that connects past-time to future-time and is the sine qua non for “debt.” It is assumed that the concept of time would emerge from incessant travels between memories of the past and the future.

The benefits of having or knowing “time” are obvious. Time travel promotes behavioral flexibility in novel situations and it is the basis for long-term strategic planning targeted to pre-selected goals. With the emergence of language, transient memories of individuals could be transformed into reproducible narratives that could be accumulated and circulated within groups, creating a powerful collective memory. But human nature can be uncooperative, since there is a demonstrable tendency for people to treat present-time and future-time unequally when they calculate costs and benefits: they discount future benefits while inflating costs incurred in the present. Thus self-interest is intrinsically impulsive and opportunistic, going for immediate gain. If unrestrained, it limits reciprocity and would have curtailed the dialectical developments described in the back-story. We have already seen this work in the genesis of punishment: the cost-benefits of being an enforcer and the benefits (rewards) provided by empathic cruelty and time travel to the future. A suitable “countermotivation device” was needed and there was one available:

Clinical psychiatry was acquainted with mental time travel avant la lettre, as early as the 1880s. Today, posttraumatic stress disorder (PTSD) is the most widely known time travel syndrome (Young Reference Young and Rosen2004). PTSD comprises an etiological event, a distressful and intrusive memory of this event, and a behavioral syndrome that represents an adaptation to the memory. Traumatic memories are a pathological expression of the “phenomenon of rumination” mentioned by Boyer (Berntsen and Hall Reference Berntsen and Hall2004). In the language of psychiatry, traumatic memories are “re-experiences” and their exemplar is the “flashback.” Schreckneurose or “fright neurosis” is especially interesting in this regard. The disorder, the German variation of shellshock during World War One, was characterized by the victim's terrifying dreams of a traumatizing experience. In the view of influential German doctors, the syndrome could be caused by re-experiences (memories, nightmares) of the future in combination with the past. The theory is that the soldiers were fixated on visions of their deaths. The man is overcome by an empathic tenderness for himself – the subject in the nightmares and intrusive images. The remembered event is a composite of two events: a real past event and an imagined future event. The past is re-enacted in the future but with a significant change. The past event was harmless; the future event is fatal. He experiences two events as a single, etiological event, in the past. His abnormality is a weakness of will and an excess of self-empathy. His symptoms, which often include psychogenic paralysis, are an unconscious effort to hide from the future. His true defect is moral, not medical. A real man (a soldier) lives in the present moment. The doctor's job is to terminate to the patient's pathogenic time travel with the most effective means, including electrical torture (Lerner Reference Lerner2003; Young Reference Young1995).Footnote ²

Post-Paleolithic developments

Human nature, as portrayed in the back-story, was fully formed during the upper Paleolithic Period. The story tells us that we are innately empathic, but that “empathy” may include unexpected and undesirable attitudes: dissimulation, self-deceit, spitefulness, Schadenfreude, and cruelty. Western normative institutions – religion, secular ethics, clinical psychology – regard these attitudes as anti-social and self-destructive. The dialectical history in this chapter views these attitudes from a different perspective, as the causes and consequences of human social evolution and self-awareness. Leaving aside the professional cynics, no one is claiming that these attitudes form the core or essence of human nature. Mind-reading, perspective-taking, hormonal responsiveness, and mental time travel were likewise responsible for “psychological altruism,” the propensity to adjust one's desires and intentions to the perceived needs or wishes of others. (In contrast, the starting point for the neo-Darwinian back-story is “biological altruism,” which is defined by fitness costs and unconcerned with altruists’ perceptions and intentions.) Human nature, as described in the back-story, is morally complex, even contradictory, certainly inclined to “read and share the concern of others” (Hrdy Reference Hrdy2009), but likewise prepared for cruel pleasures.