them when we are not actuallyinteracting with people. All inference
systems can run in a decoupled mode, that is, disengaged from actual
external inputs from the environment or external output in behavior.
A crucial human capacity is to imagine counterfactuals—What would
happen if I had less meat than I actually have? What would happen if
I chose this path rather than that one?—and this applies to interac-
tion too. Before we make a particular move in any social interaction,
we automatically consider several scenarios. This capacity allows us,
for instance, to choose this rather than that course of action because
we can imagine other people's reactions to what we would do.
In fact, we can run such decoupledinferences not only about persons [149]
who are not around but also about purely imaginary characters. It is
striking that this capacity seems to appear very early in children's
development. From an early age (between three and ten years) many
children engage in durable and complex relationships with "imaginary
companions." Psychologist Marjorie Taylor, who has studied this phe-
nomenon extensively, estimates that about half of the children she has
worked with had some such companions. These imagined persons or
person-like animals, sometimes but not always derived from stories or
cartoons or other cultural folklore, follow the child around, play with
her, converse with her, etc. One girl describes her companions Nutsy
and Nutsy as a couple of birds, one male and one female, who accom-
pany her as she goes for a walk, goes to school or gets in the car.
Taylor's studies show that having long-term relationships with
nonexistent characters is not a sign of confusion between fantasy and
reality. Developmental psychologists now use precise tests to determine
how children mark off the real from the fantastic. Those with compan-
ions pass such tests from the age of three and are often better than other
children at differentiating between the real and the imagined. They
know perfectly well that their friends the invisible lizard, the awkward
monkey, or the amazing magician, are not there in the same sense as real
friends and other people. Also, children with companions are often bet-
ter than others at tasks that require a subtle use of intuitive psychology.
They seem to have a firmer grasp of the difference between their own
and other people's perspectives on a given situation and are better at
construing other people's mental states and emotions.
All this led Taylor to the intriguing hypothesis that imaginary com-
panions may well provide a very useful form of training for the social
mind. The relationship with such a companion is a stable one, which
means that the child computes the companion's reactions by taking
WHYGODS AND SPIRITS?