12 human agency
(1925, 34, 118­22, 1965, 39­40; cf. Arbman 1939, 27). I have elsewhere tried
to show in detail that here Durkheim errs: all people all over the world have
intuitive expectations about the ordinary course of events, together with ideas
that violate these intuitive expectations (Geary 2005; Pyysiäinen 2001b, 55­74,
2004d, 39­52, 81­89, 2005d). Moreover, the "regeneration" of the group is
a cognitive act, consisting of persons' mental representations of other persons'
mental representations; what he called "collective consciousness" thus is
shared knowledge and can be studied from the cognitive psychological perspective
(see Pyysiäinen 2005f).
Belief in supernatural agents forms an obvious recurrent pattern in various
cultural traditions. As some of these beliefs help people express, process,
and justify their central values and norms, we can try to combine the Tylorian
and Durkheimian legacies in explaining how beliefs about supernatural
agents are used in organizing a society (see Boyer 2000b, 2002; Pyysiäinen
2005f). Typical examples of supernatural agents are gods, spirits, ghosts, angels,
demons, and so forth. The category of supernatural agents may seem so heterogeneous
that one may well ask whether it is a genuine category at all. Some
might want to claim that it is a pseudocategory just like "mysticism," "ritual,"
or even "religion" itself (see, e.g., Fitzgerald 1996, 1997; Penner 1983). I think
this view is overly pessimistic. If we can find some theoretical depth to the ideas
of agency and counterintuitiveness, the idea of supernatural agents might be
operationalized for research. We need a theoretical frame of reference in which
the various emic distinctions between different kinds of supernatural agents are
seen as subdivisions within the general, etic category of supernatural agency.
Using concepts at a mediating level between cultural particulars and the general
category of supernatural agents, we may then conceptualize the various
recurrent patterns within that category.
1.3 There Must Be Somebody Out There
I use quite consciously the expression "supernatural agents" instead of "supernatural
beings." By agents, psychologists and philosophers mean organisms
whose behavior can be successfully predicted by postulating conscious beliefs
and desires (Dennett 1993, 15­17) or entities whose behavior is caused by their
mental states (Bechtel 2008, xii). I use the concept in the sense of an organism
to which animacy (liveliness, self-propelledness) and mentality (beliefs and
desires) are (correctly or incorrectly) attributed.
I follow Barrett (2008) in distinguishing between two components of
agency: animacy and mentality. The domain of animacy is characterized
by such things as self-propelledness and goal-orientation: organisms tacitly
postulate animacy for other organisms when they feel that those organisms
move by themselves and seem to be moving toward some goal (see Boyer and
religion: her ideas of his ideas of their ideas . . . 13
Barrett 2005). When one also attributes to the organism a conscious intention
and begins to simulate its mental states or to try to theorize about its beliefs
and desires, one moves on to "mentalizing" or "mind reading" (Nichols and
Stich 2003). This ability is often called a "theory of mind" (ToM), in the sense of
folk-psychological theories about other minds (Carruthers and Smith 1996).7
I distinguish three overlapping cognitive mechanisms that contribute
to agentive reasoning. The first is hyperactive agent detection (HAD, Barrett
2000): the tendency to postulate animacy--this mechanism is triggered by
cues that are so minimal that it often produces false positives, for example, we
see faces in the clouds, mistake shadows for persons, and so forth. Second is
hyperactive understanding of intentionality (HUI): the tendency to postulate
mentality and to see events as intentionally caused even in the absence of a visible
agent. Third is hyperactive teleofunctional reasoning (HTR): the tendency
to see objects as existing for a purpose.
1.3.1 Agency
Barrett (2000) coined the acronym HADD (hyperactive agent detection device)
to refer to the cognitive processes that help us recognize agents (actually, animacy)
and distinguish them from nonagents. This mental "device" is hyperactive
or hypersensitive, in that under certain conditions, it is triggered by very minimal
cues.8
We see faces in the clouds and detect predators in rustling bushes
because such ambiguous perceptions easily trigger the postulation of agency.
According to Barrett, a normally functioning HADD is hyperactive by its very
nature--hyperactivity is not something exceptional. From an evolutionary point
of view, this is plausible, insofar as the costs of false positives that an overreacting
detector produces are lower than the benefits it brings (Atran 2006).
The following most common direct cues for agency have been suggested
(Blakemoore et al. 2003; Boyer and Barrett 2005; see Heider and Simmel 1944):
1. Animate motion that has as its input such things as nonlinear changes
in direction, sudden acceleration without collision, and change of
physical shape that accompanies motion (e.g., caterpillar-like crawling)
2. An object reacting at a distance
3. Trajectories that only make sense if the moving entity is trying to reach
or avoid something, which leads to goal-ascription
4. An entity appearing to be moving by conscious intention to an
apparent end result (intention-ascription)
5. The experience of joint attention (for which we develop a capacity
between nine and twelve months of age)
These cues trigger the feeling or intuition of agency spontaneously and automatically,
in the sense that this intuition can neither be rationally controlled
nor initiated or terminated at will (see Bargh 1994; Pyysiäinen 2004c; see
14 human agency
the appendix). However, I suggest that the first three cues only trigger animacy
assumptions, while the latter two trigger HUI. It may also be misleading
to call HADD a (single) "device," as several systems contribute to the perception
that one is facing an agent (see Boyer and Barrett 2005). Once people
start reasoning about the beliefs and desires of a postulated agent, they use
mind reading: the capacity to make inferences about the beliefs and desires of
others and to explain their behavior on that basis (see Carruthers and Smith
1996; Frith and Frith 2005; Nichols and Stich 2003; Premack and Woodruff
1978; Saxe and Baron-Cohen 2006; Tomasello et al. 2005; Tremlin 2006, 75).
Mind reading is often understood to be innate--not a learned ability but rather
something triggered in the course of normal development (Leslie 1996; Wellman
and Miller 2006, 28). It has been seen as an innate modular algorithm,
(e.g. Leslie), as an innate body of knowledge (e.g. Pinker and Sperber), and as
modular in all three senses of the concept of a "module" (Baron-Cohen) discussed
in the appendix (see Gerrans 2002, 308). It can also be understood to
be based on a nonmodular conceptual competence (Wellman et al. 2001; see
Yazdi et al. 2006).
The standard psychological test for ToM is the so-called false belief task
(Wimmer and Perner 1983). In this test, children are shown a sketch in which
a boy called Maxi first puts a chocolate bar in container A. Then, without Maxi's
knowing, his mother moves it to container B. The subjects are then asked
where Maxi would look for the chocolate bar when he returns. Children around
three years of age tend to say (incorrectly) that Maxi would look in container B
(where the bar was hidden). Only in the fourth year does the tendency appear
to say that Maxi would look in container A. It thus seems that younger children
cannot understand that Maxi remains ignorant of his mother hiding the
bar in container B. These results have been obtained in many replications of
the experiment and across many variations of the original design. Scholarly
opinions differ on what these results really tell about the ways children think
(see Baron-Cohen 2000; Baron-Cohen et al. 1985; Fodor 1992; Perner 1995;
Wimmer and Perner 1983).9
Simon Baron-Cohen, Alan Leslie, and Uta Frith (1985) used the same task
to test autistic children' ability to impute mental states to others. They tested
twenty autistic children, fourteen children with Down syndrome, and twentyseven
clinically normal preschool children. Two dolls were used, Sally and
Anne. First, Sally placed a marble in her basket, and then left the scene. Anne
then transferred the marble into her own box. When Sally returned, the experimenter
asked: "Where will Sally look for her marble?" Children with Down
syndrome as well as normal preschool children answered correctly by pointing
to the basket where Sally had put the marble. The autistic group consistently
answered by pointing to the box where the marble really was; the autistic
children did not seem to get the difference between their own and the doll's
knowledge. In the authors' words, they "failed to employ a ToM"--a failure
religion: her ideas of his ideas of their ideas . . . 15
consisting of an "inability to represent mental states" (Baron-Cohen et al. 1985,
43). Autism thus might seem to involve a lack of a ToM. The inability to make
inferences about what other people believe to be the case in a given situation
prevents one from predicting what they will do (Baron-Cohen 2000; BaronCohen
et al. 1985, 39; see Carruthers and Smith 1996, 223­73; Frith 2001).
However, failure in the test does not necessarily indicate a lack of a ToM,
because having a ToM does not necessarily require the ability to reason about
false beliefs (Bloom and German 2000; see Wellman et al. 2001). Reasoning
about false beliefs is just one way of using ToM (see Stone 2005). Philip Gerrans
and Valerie Stone also argue that there is no hardwired ToM as a dedicated
module at all and that belief attribution is not supported by a domain-specific,
modular mechanism (see appendix). Instead there is interaction between such
low-level, domain-specific mechanisms as tracking gaze and bodily movement,
joint attention, and so forth and higher-level domain-general mechanisms for
metarepresentation, recursion, and executive function. The output of the lowlevel
mechanisms serve as input for the higher-level, domain-general processes
(Gerrans 2002, 306, 311; Stone and Gerrans 2006a,b).
Perner et al. (2006), as well as Saxe et al. (2006) present neuroscientific
evidence for the view that in the right and left temporo-parietal junction of the
brain there is a specialized, domain-specific neural mechanism for reasoning
about beliefs. Inhibitory control of mental contents, together with response
selection, is mediated by domain-general mechanisms, while the domainspecific
mechanism of the temporo-parietal junction is only recruited for
processing beliefs. And reasoning about beliefs is faster than following domaingeneral
rules in reasoning about other things than beliefs (Saxe et al. 2006; see
Saxe and Baron-Cohen 2006).
Leslie et al. (2004, 2005) argue that ToM is a partly modular learning
"mechanism" that only "kick-starts" the attribution of beliefs and desires. Just
as color vision provides us with color concepts, ToM introduces belief and desire
concepts. Reasoning about the contents of beliefs takes place through a selection
process (SP) with inhibition. Developing slowly from the preschool period
onward, SP acts by selecting a content for an agent's belief and an action for the
agent's desire. In everyday thinking, taking a belief to be true is the default, so
success in the false belief task requires an ability to inhibit the default attribution
and to select the erroneous belief for the character looking for the hidden
object.
The mentality of organisms that ToM processes is very early on understood
as separate from the physical body. Kuhlmeier et al. (2004), for example,
show that five-month-old infants apply the constraint of continuous motion to
inanimate blocks but not to persons. They thus do not seem to view human
agents as material objects. In the experiment, twenty infants watched a film in
which a woman was standing on a stage that contained two large red screens
separated by 1.21 meters. The woman then went behind the first screen; her
16 human agency
identical twin sister in identical clothing had been hiding behind the second
screen and now emerged from behind it. What the infants saw was a woman
moving behind a screen and then suddenly emerging from behind the other
screen without being visible in the space between the two screens. The infants
revealed that they were not surprised by this apparently miraculous event,
although they were surprised when they watched mere physical objects (as distinct
from humans) behaving in the same way. (They exhibited surprise by
looking longer at the action they were witnessing.) This suggests that they do
not consider agency to be constrained by the physical body.
Even as adults, people do not feel themselves to be bodies but instead feel
that they occupy bodies (Bloom 2005, 191, 2007; Merleau-Ponty 1992; see
Bronkhorst 2001, 402). Routinely, people spontaneously attribute their own
agentive properties to the homunculus called the "self," which is not merely
the sum of a set of lower-level, "dumb[er]" homunculi (Nichols and Stich 2006,
9­10). To the extent that the agentive properties are detached from the physical
body, it is perfectly natural for people to have intuitive beliefs about disembodied
agents such as spirits. However, disembodiment does not mean a complete
lack of bodily form; mentality may instead be attributed to various forms of
"subtle" or otherwise nonstandard bodily forms.
1.3.2 Intentionality
There are also indirect cues that lead people to interpret an event as intentionally
caused even when they do not perceive an agent. Hyperactive understanding
of intentionality produces an automatic conclusion that an event or
structure must have been caused or designed by an intelligent agent, even
when no trace of an agent is evident (Barrett 2004b, 34). For example, normally,
when one sees an artifact, one knows that it must have been designed
and made by an agent, even though this agent is not present and one does not
know her or his identity (see Czachesz 2007, 86).
This kind of reasoning contributes to beliefs about supernatural agency
when it is extended to such cases as, for instance, seeing the image of Jesus
falling off the wall right at the moment when someone says something blasphemous.
The coincidence of the words uttered and the picture falling creates
the feeling that the falling was intentionally caused, despite the fact that no one
is present and one has no idea of the mechanism by which an agent could have
caused the picture to fall (see Atran 2002a, 59­63; Bering 2006; Bering and
Parker 2006). Bering (2003) suggests that there may even be a specific cognitive
mechanism devoted to processing apparently intentional events ("existential
meaning") in the absence of a physical agent.
Such inferences are fast and automatic intuitions (see the appendix).
The low-level intuitions about agency are dedicated to combining perceived
movement with the perceiver's reflective ideas of agency and free will. When
religion: her ideas of his ideas of their ideas . . . 17
action consistently follows prior thought and when apparent causes of action
other than somebody's thought are excluded, we have the experience of volitional
action. Volition, in turn, presupposes an agent. Torsten Nielsen (1963)
showed in the 1960s that when subjects were asked to put one hand in a box
and draw a straight line with it while looking into the box through a tube, they
could be fooled into believing that the experimenter's hand seen in the box
through a mirror was their own (which they did not see). When both hands drew
a straight line, the subjects perceived the alien hand as their own. When the
experimenter's hand drew a curve to the right, the subjects still experienced the
hand as their own, and now perceived it as making involuntary movements.
Similarly, Ramachandran and colleagues arranged an experiment in which
a subject with an amputated arm was asked to put his real arm and what he
experienced as his phantom limb in a box with a vertical mirror in the middle;
the reflection of the real arm then appeared in the mirror where the phantom
would have been if it were real. When the subject moved his real arm, he felt
he was moving his phantom limb. Then the experimenter put his arm inside
the box so that it appeared in the place of the phantom. Observing his real arm
and the experimenter's arm in the box made the subject feel as though his
phantom arm was moving (Ramachandran and Blakeslee 1999, 46­48; see
Wegner 2002, 40­44).
Wegner (2002, 44) thus argues that one can think movement is intentional
by watching any body move where one thinks one's own body is. But there are
also cases in which one perceives a movement as guided by an intentional will
not one's own. When the picture of Jesus Christ suddenly falls off the wall, the
default explanation is not necessarily that one did it oneself (although this is
possible). One can come to think that the movement is intentional by combining
one's perception of it with a representation of an invisible agent.
Wegner (2002, 44) suggests in passing that the experience of another's
movement as willful is mediated by the so-called mirror neuron system in the
brain. This system mediates such intuitive reasoning as, for example, feeling
disgust when witnessing someone drink a glass of milk with a face contracting
in an expression of disgust. It cannot mediate such reflective ToM reasoning as
in, for instance, trying to figure out what gift would please a foreign colleague
(Bargh 1994, 3; Keysers and Gazzola 2007). Mirror neurons were first found
in monkey brains in premotor area F5 and parietal area 7b. They are called
mirror neurons because they are activated during both the execution of purposeful,
goal-related hand actions (grasping etc.) and the observation of similar
actions performed by conspecifics (see Gallese et al. 1996; Rizzolatti et al.
1988; Rizzolatti et al. 2000). A mirror neuron system has also been found in
the human brain (although no special type of neuron seems to be involved).
Implicit, automatic, and unconscious simulation processes establish a link
between an observed agent and the observing agent, thus making imitation
possible (Bremmer et al. 2001; Gallese and Metzinger 2003; Hari et al. 1998).
18 human agency
It has been suggested that social cognition is based on the mirror neuron
system, which mediates observations about other agents' apparently intentional
behavior (see Hurley 2008a). When two agents interact socially, the mirror
neuron system is activated and creates a shared neural representation (Becchio
et al. 2006, 66­67). It thus helps people understand the intentions of others
and to relate them to their own (see Gallese and Goldman 1998; Gallese and
Metzinger 2003; Gallese et al. 2004). In being restricted to motor processes
and intuitions, though, the mirror neuron theory alone may not be sufficient
to account for social cognition (see Bargh 1994; Hurley 2008b; Keysers and
Gazzola 2007). Social cognition also involves complicated reflective processes
related to understanding of intentionality.
Recognizing consciously goal-directed action involves recognizing intentionality
in three senses: intention recognition, attribution of intention to its
author (HUI), and understanding motivation behind the intention (HUI, HTR)
(see Becchio et al. 2006). Intentionality as the mark of the mental is usually
understood as "directedness" or "aboutness" (Brentano 1924; Dennett 1993, 67,
1997, 66; Husserl 1950a,b). Mental states are beliefs about the world, desires
for things, and so forth (Wellman and Miller 2006, 34­35). Following Jaakko
Hintikka (1975), I argue that it might be better understood as intensionality,
however (see the appendix, section 2.1).
Intensionality relates to such things as meanings, properties, propertybased
relations, and propositions. When we describe a set intensionally, we
list the properties the members of the set have. We may, for instance, try to list
the various gods of the religions of world using the following formula: {x:x is
a god}. An extensional listing, in contrast, is made by just listing the members:
{Allah, Quetzalcoatl . . . Zeus}. These two lists are extensionally equivalent
but intensionally distinct. This distinction is based on the fact that terms and
expressions have an intension or meaning (Sinn) and an extension or reference
(Bedeutung) (Frege 1966). For example, someone's beliefs about the Evening
Star are not necessarily beliefs about the Morning Star, because the person in
question may not know that these two are the same planet.10
The terms have
different intentions but the same reference. Mere extensional concepts are not
enough for conceptualizing agency because agency is defined by the mental act
of directedness (instead of mere reactivity, Leslie 1994).
The idea of intentionality as intensionality brings to the fore what seems to
be crucial in intentionality in a psychological sense. According to Hintikka, "a
concept is intentional if and only if it involves the [logically] simultaneous consideration
of several possible states of affairs or courses of events." Intentionality
as intensionality thus means that for an act to be intentional, it must involve
a conscious choice made between several alternatives (1975, 195; see also 212­13,
1982). Intentionality as intensionality means that intentional systems involve
the consideration of possible worlds as logically alternative scenarios (2006a,
21­24, 2006b, 557­58; see Dennett 1993, 122, 174­75). Thus, ToM depends
religion: her ideas of his ideas of their ideas . . . 19
on the ability for counterfactual reasoning about contrastive states of affairs
or courses of events (see also Buller 2005a, 194). Directedness is involved, but
in an intensional sense: a mental state is directed toward something, but with
the awareness that it might just as well be directed toward something else. In
intentionality, as distinct from mere reactions, a course of action is selected
from among multiple mentally represented alternatives.
If mind reading or ToM operates through a selection process with inhibition,
understanding intentionality as intensionality fits it much better than
the idea of intentionality as simply directedness. For example, understanding
another's beliefs and desires in the false belief task involves precisely consideration
of possible worlds as logically alternative scenarios. This entails understanding
that the mind is a representational device and that there is no reason
that all represented propositions should be true (Perner 1993). One has to be
able to metarepresent another's beliefs, that is, to embed them in one's own
beliefs, as in "I believe that `she thinks that he is wrong.'" Valerie Stone (2005)
argues that it is precisely the ability of metarepresentation that enables children
to succeed at explicit false belief tasks: the child must be able to understand
that various agents represent the location of the chocolate bar in various ways.
According to Bering, it is around six or seven years of age that children become
capable of understanding such third-order intentionality in which "he thinks
that she believes that he wants," for example. At that point in their development,
they are also able to regard random events as symbolic and declarative of
a supernatural agent's mental states (Bering and Johnson 2005, 134­36).11
Degrees or orders of intentionality is an idea of Dennett (1993, 243­46). It
has been argued that normal adult humans are capable of fourth- or fifth-order
intentionality (when no external memory stores are used) (Dunbar 2003, 170,
2006, 171­72). "I know12
that John wants Mary to understand that Bill believes that
Linda loves him" is an example of fourth-order intentionality. Such metarepresentation
of intentional attitudes requires a capacity to understand recursive13
structures (Stone 2005; see Hintikka 1975). To the extent that we are interested
in how ToM works in social cognition, only embedded mental states count, not
mere embedded observed facts. For example, the sentence "John says Mary to
claim that Bill wrote her that Linda sent him a love letter" is not a case of degrees
of intentionality. Even an autistic person incapable of mind reading might be
able to understand this sentence as reporting four interrelated facts.
The recursive embedding of metarepresentations seems to be an exclusively
human capacity; apes, for example, do not understand recursive structures,
though the evidence here is somewhat ambiguous (Dennett 2006, 111;
Dunbar 2003, 170). Premack and Premack (2003, 149­53), however, claim
that chimpanzees are only capable of first-order intention ascription and that
even this requires much conscious effort from the chimp, whereas a ten- to
twelve-month-old human infant does this spontaneously. Human social cognition
starts where the chimpanzee's cognitive ability ends. As Tomasello and
20 human agency
Carpenter (2007, 122) observe, "from a very early age human infants are motivated
to simply share interest and attention with others in a way that our nearest
primate relatives are not." Joint attention, which is one of the low-level inputs
of ToM, starts to develop in human infants as early as nine months, when the
learning child begins to look at an object and at the parent, trying to draw the
adult's attention to a shared object (see Moore and D'Entremot 2001; Tomasello
and Carpenter 2007; Tomasello et al. 2005).
Sperber (1997) argues that recursively metarepresenting a belief within
another belief can provide a validating context for the embedded belief. It is
then accepted as true because of certain second-order beliefs about it (Boyer
1994b, 120; Sperber 1996, 69­70, 89­97). To the extent that these secondorder
beliefs seem compelling, the first-order belief embedded in them is also
plausible.14
In, for example, "`Jesus is our redeemer' is true because it says so
in the Bible," "Jesus is our redeemer" derives its plausibility from the validating
embedding "It says so in the Bible" (see Pyysiäinen 2003c).
Metarepresentation thus makes it possible to make inferences about propositions
with an undecided truth-value--as in, for instance, "`If it was John
who stole my wallet,' then he should be arrested." Here the judgment of John
being guilty must be temporally suspended in order to make a provisional
inference (see Cosmides and Tooby 2000, 59­60; Pyysiäinen 2003c; Sperber
1996, 71­73). It is important to distinguish between what Sperber calls "halfunderstood"
propositions and propositions with an undecided truth-value. It is
possible to generate conditional inferences from premises with an undecided
truth-value, but it is not possible to do so from incomprehensible premises.
These can only be used as quotations that receive a meaning from the validating
context in which they are metarepresented.
Failure to understand the nature of metarepresentation leads to accepting
all beliefs as one's own (see Sperber 1994, 2000a; Pyysiäinen 2003c). A psychiatrist
may, for example, entertain the metarepresentation "Joe believes `I am
Jesus.'" If she is not able to use the metarepresentation "Joe believes," then
"I am Jesus" becomes her own belief (see Corcoran et al. 1995; Frith 1995). Metarepresented
beliefs are decoupled from reality, and their semantic relations are
suspended, in the sense that one cannot directly reason from "Ann believes
`John is a spy'" that John really is a spy (Cosmides and Tooby 2000). This would
only follow in the case that the metarepresentational context is automatically
validating. For a Christian believer, the metarepresentation "It says so in the
Bible" often is such a validating context, for example (Pyysiäinen 2003c). It is
beliefs about beliefs that validate beliefs.
1.3.3 Teleofunctional Reasoning
Humans are prone to see things as existing for a reason or purpose. This tendency
to view entities as existing for a purpose may derive from children's early
religion: her ideas of his ideas of their ideas . . . 21
emerging ability to think that there are hidden intentions behind everything
and to interpret agents' behavior as goal-directed (Johnson 2000, 188, 208;
Kelemen and DiYanni 2005, 6). Children's understanding of how agents use
objects as means to achieve goals may provoke a rudimentary teleofunctional
view of entities: agents' intentions are understood as being intrinsic properties
of the objects themselves (Kelemen 1999a,b; Kelemen and DiYanni 2005, 6).
There is experimental evidence for a "promiscuous teleology": British and
American elementary schoolchildren are prone to generating teleofunctional
explanations of the origins and nature of living and nonliving natural entities
and endorsing intelligent design as the source of both animals and artifacts
(Evans 2000, 2001; Kelemen 2004; Kelemen and DiYanni 2005).
Kelemen and DiYanni (2005, 7) point out that this view about purpose does
not necessarily derive from agentive intuitions. Teleo-functional reasoning may
also be an independent characteristic of causal reasoning. However, Asher and
Kemler Nelson (2008) provide evidence for the interpretation that three­ and
four­year-old children can adopt the intentional stance (Dennett 1993) and do
understand the true functions of artifacts to be the designed functions. It thus
is also possible that intuitions about purpose and design and intuitions about
agency derive from a common source.
Lewis Wolpert (2007, 27­33, 67­82) argues that all causal reasoning in
humans derives from the manufacturing and using of tools that typify the species,
not from social interaction (cf. Dunbar 1993, 2002, 2003, 2006). The
making of first tools created a new kind of selection pressure, to which causal
thinking is an adaptation: the making of tools required an ability to understand
the ideas of means and ends and of causality and intentionality. Although nonhuman
primates can distinguish the animate from the inanimate, they do not
view the world in terms of intermediate and often hidden underlying causes,
reasons, intentions, and explanations. (They understand animacy but not
mentality.)
Wolpert neither develops this idea carefully nor contrasts it in detail with
competing accounts. It does not have to be incompatible with, for example,
Dunbar's view of human sociality as having provided the selection pressure for
the large neocortex of our species. Tool use and social interaction are not two
mutually exclusive phenomena but seem to have developed together. Thus, we
can see Wolpert's idea as an important addition or qualification to other arguments
that may be able to help explain beliefs about supernatural agents.
The hyperactivity in HADD, HUI, and HTR means that they produce many
false positives: people perceive agents where there are none, ascribe intentionality
to events and structures that are purely mechanical, and use teleofunctional
reasoning to explain the natural world. As natural-born tool users, humans see
intelligent agency and design everywhere. As humans are the prototype of an
agent, humans attribute at least some humanlike features also to the agents
HADD postulates (Boyer 1996c; see Guthrie 1993; Richert and Barrett 2005).
22 human agency
However, supernatural agent concepts are not derived from these false
positives. First, ambiguous perception and overextended attribution of intentionality
and design in themselves do not contain enough information for
forming a persisting agent concept. Second, the supernatural agent concepts of
religious traditions are abstractions from a large number of individual mental
representations that are communicated among persons. Therefore, we cannot
explain these concepts only by referring to ambiguous individual perception;
we have to explain why they have become widespread in populations (see
Barrett 1998, 617, 2004b, 41, 43; Boyer 1994a,b).
The notions of HADD, HUI, and HTR can help explain why certain kinds
of supernatural agent concepts are easier to adopt than others; this ease then
explains, partly at least, why these concepts are found all over the world both
in ancient and modern times. Because there is a natural place for these concepts
in the human mind and in human practices, they are contagious and
have become widespread (Barrett 2004b, 33­39). Mind and culture thus are
not so much two different levels as the endpoints on a scale from individual
to public and "shared" (see Sperber 2006). The folklore scholar Lauri Honko
(1962, 93­99) tried to explain the actualization of traditional beliefs in casual
encounters with the spirits by referring to the psychology of perception; we can
now try to do much the same with the help of cognitive and developmental psychology
and evolutionary theory. That effort may turn out to have far-reaching
consequences for the study of religion and culture. The next thing for me to do
with this purpose in mind is to provide some criteria for calling some concepts
counterintuitive.
1.4 Out of the Ordinary
1.4.1 Intuitive Ontology and Counterintuitiveness
Animacy, mentality, and teleology are, of course, not all that the human mind
processes intuitively. Some of the basic domains of intuition are enumerated
in the list that follows (see Barrett 2000, 2008; Boyer 1994b; Keil 1979, 1996;
Sommers 1959). We intuitively understand that things exist in space and time;
some things also have physicality and solidity, and some of them belong to
the folk-biological domain of living kinds (plants and animals). Animacy is
a mediating category between folk biology and mentality, in the sense that it is a
subcategory within folk biology and may be the basis on which mind reading
has developed: the idea of beliefs and desires in a way completes the understanding
that something is goal-directed. Social position, for its part, is the domain
of such social relationships as "being someone's father" or "being the king,"
and so forth. Such phenomena as cooperation and reciprocal altruism require
one to have intuitions about the kind of social relationship they entail. Finally,
we must add the category of time--though temporality is a neglected theme