One measure of ToM is called intentionality.13 An organism that is capable of reflecting about its own state of mind, about its own beliefs and desires, as in I want a bite of my mother’s pot roast—is called “first-order intentional.” Most mammals fit in that category. But knowing about yourself is a far different skill from knowing about someone else. A second-order intentional organism is one that can form a belief about someone else’s state of mind, as in I believe my son wants a bite of my pot roast. Second-order intentionality is defined as the most rudimentary level of ToM, and all healthy humans have it, at least after their morning coffee. If you have third-order intentionality you can go a step further, reasoning about what a person thinks a second person thinks, as in I believe my mom thinks that my son wants a bite of her pot roast. And if you are capable of going a level beyond that, of thinking I believe my friend Sanford thinks that my daughter Olivia thinks that his son Johnny thinks she is cute or I believe my boss, Ruth, knows that our CFO, Richard, thinks that my colleague John doesn’t believe her budgets and revenue projections can be trusted, then you’re engaging in fourth-order intentionality, and so on. Fourth-order thinking makes for a pretty complicated sentence, but if you ponder these for a minute, you’ll probably realize you engage in it quite frequently, for it is typical of what is involved in human social relationships.
Fourth-order intentionality is required to create literature, for writers must make judgments based on their own experiences of fourth-order intentionality, such as I think that the cues in this scene will signal to the reader that Horace thinks that Mary intends to dump him. It is also necessary for politicians and business executives, who could easily be outmaneuvered without that skill. For example, I knew a newly hired executive at a computer games company—call her Alice—who used her highly developed ToM to get out of a touchy situation. Alice felt certain that an outside company that had a long-term contract for programming services with her new employer was guilty of certain financial improprieties. Alice had no proof, and the outside company had an airtight long-term contract that required a $500,000 payment for early termination. But: Alice knew that Bob (the CEO of the outside company) knew that Alice, being new on the job, was afraid to make a misstep. That’s third-order intentionality. Also: Alice knew that Bob knew that she knew that Bob was not afraid of a fight. That’s fourth-order thinking. Understanding this, Alice considered a ploy: What if she made a bluff that she had proof of the impropriety and used that to force Bob to let them out of the contract? How would Bob react? She used her ToM analysis to look at the situation from Bob’s point of view. Bob saw her as someone who was hesitant to take chances and who knew that he was a fighter. Would such a person make a grand claim she couldn’t back up? Bob must have thought not, for he agreed to let Alice’s employer out of the contract for a small fraction of the contractually obligated sum.
The evidence on nonhuman primates seems to show that they fall somewhere between first- and second-order thinking. A chimp may think to itself, I want a banana or even I believe George wants my banana, but it wouldn’t go as far as thinking, I believe George thinks that I want his banana. Humans, on the other hand, commonly engage in third- and fourth-order intentionality and are said to be capable of sixth-order. Tackling those higher-order ToM sentences taxes the mind in a way that, to me, feels analogous to the thinking required when doing research in theoretical physics, in which one must be able to reason about long chains of interrelated concepts.
If ToM both enables social connection and requires extraordinary brain power, that may explain why scientists have discovered a curious connection between brain size and social group size among mammals. To be precise, the size of a species’ neocortex—the most recently evolved part of the brain—as a percentage of that species’ whole brain seems to be related to the size of the social group in which members of that species hang out.14 Gorillas form groups of under ten, spider monkeys closer to twenty, and macaques more like forty—and these numbers accurately reflect the neocortex-to-whole-brain ratio of each of these species.
Suppose we use the mathematical relationship that describes the connection between group size and relative neocortex size in nonhuman primates to predict the size of human social networks. Does it work? Does the ratio of neocortex to overall brain size apply to calculating the size of human networks, too?
To answer that question, we first have to come up with a way of defining group size among humans. Group size in nonhuman primates is defined by the typical number of animals in what are called grooming cliques. These are social alliances like the cliques our kids form in school or those adults have been known to form at the PTA. In primates, clique members regularly clean each other, removing dirt, dead skin, insects, and other objects by stroking, scratching, and massaging. Individuals are particular about both whom they groom and whom they are groomed by, because these alliances act as coalitions to minimize harassment from others of their kind. Group size in humans is harder to define in any precise way because humans relate to one another in many different types of groups, with different sizes, different levels of mutual understanding, and different degrees of bonding. In addition, we have developed technologies designed specifically to aid large-scale social communication, and we have to be careful to exclude from group size measurements people such as e-mail contacts we hardly know. In the end, when scientists look at groups that seem to be the cognitive equivalent of nonhuman grooming cliques—the clans among Australian aboriginals, the hair-care networks of female bushmen, or the number of individuals to whom people send Christmas cards—the human group size comes out to about 150, just about what the neocortex size model predicts.15
Why should there be a connection between brain power and the number of members in a social network? Think about human social circles, circles consisting of friends and relatives and work associates. If these are to remain meaningful, they can’t get too big for your cognitive capacities, or you won’t be able to keep track of who is who, what they all want, how they relate to one another, who can be trusted, who can be counted on to help out with a favor, and so on.16
To explore just how connected we humans are, in the 1960s the psychologist Stanley Milgram selected about 300 people at random in Nebraska and Boston and asked each of them to start a chain letter.17 The volunteers were sent a packet of materials with a description of the study, including the name of a “target person”—a randomly chosen man in Sharon, Massachusetts, who worked as a stockbroker in Boston. They were instructed to forward the packet to the target person if they knew him or, if they didn’t, to send it to whichever of their acquaintances they deemed most likely to know him. The intention was that the acquaintance, upon receiving the packet, would also follow the instructions and send it along, until eventually someone would be found who did know the target person and would send it directly to him.
Many people along the way didn’t bother, and broke the chain. But out of the initial 300 or so individuals, 64 did generate chains that ultimately found the man in Sharon, Massachusetts. How many intermediaries did it take until someone knew someone who knew someone who knew someone … who knew the target? The median number was only about 5. The study led to the coining of the term “six degrees of separation,” based on the idea that six links of acquaintanceship are enough to connect any two people in the world. The same experiment, made much easier by the advent of e-mail, was repeated in 2003.18 This time the researchers started with 24,000 e-mail users in more than 100 countries, and 18 different target people spread far and wide. Of the 24,000 e-mail chains those subjects started, only about 400 reached their target. But the result was similar: the target was contacted in a median of five to seven steps.