This
article appeared in Philosophical Topics, 28,
171-199.
The Mind’s “I” and the Theory of Mind’s “I”:
Introspection and Two
Concepts of Self[1]
Shaun
Nichols
College of
Charleston
I.
INTRODUCTION
Introspection plays a
crucial role in Modern philosophy in two different ways. From the beginnings of Modern
philosophy, introspection has been used a tool for philosophical exploration in
a variety of thought experiments.
But Modern philosophers (e.g., Locke and Hume) also tried to characterize
the nature of introspection as a psychological phenomenon. In contemporary philosophy,
introspection is still frequently used in thought experiments. And in the analytic tradition,
philosophers have tried to characterize conceptually necessary features of
introspection.[2] But over the last several decades,
philosophers have devoted relatively little attention to the cognitive
characteristics of introspection.
This has begun to change, impelled largely by a fascinating body of work
on how children and autistic individuals understand the mind.[3] In a pair of recent papers, Stephen
Stich and I have drawn on this empirical work to develop an account of
introspection or self-awareness.[4]
In this paper, I will elaborate and defend this cognitive theory of
introspection further and argue that if the account is right, it may have
important ramifications for psychological and philosophical debates over the
self.
Since the paper will cover a
rather diverse set of issues, let me begin by mapping out the structure of what
follows. In section II, I will set
out the most prominent account of introspection in the recent literature, the
Theory Theory of self-awareness, according to which the capacity to detect one’s
own mental states depends on the capacity to detect other people’s mental
states. I’ll then sketch the
alternative “Monitoring Mechanism” account that Stich and I have defended. I will go on to offer a couple of new
arguments for the Monitoring Mechanism account, and I will argue that the
Monitoring Mechanism is plausibly modularized to an interesting extent. I will also respond to the worry that
evolutionary considerations cast doubt on the theory. In section III, I review psychological
work on the concept of self, and I argue that the Monitoring Mechanism theory
suggests that there is an important notion of self that is largely neglected in
the psychological literature and that needs to be distinguished from a concept
of self that derives from the Theory of Mind. In the 4th section, I
argue that this distinction between two concepts of self helps to explain
recalcitrant philosophical problems concerning the self.
II.
THE MODULARITY OF THE MIND’S EYE
The best known and most
influential account of self-awareness in the recent literature is the Theory Theory of self awareness.[5]
According to this account, one determines one’s own mental states by using a
“theory” of mind, and this theory of mind is the very same theory that one uses
to determine the mental states of others. In philosophy, the Theory Theory of
self-awareness was first proposed by Sellars,[6]
but its growing influence in cognitive science is largely due to the work in
developmental psychology on the understanding of other minds or mindreading.
To make the view clear, it is important to review a bit of the history in
developmental psychology. The
prevailing view of how children (and adults) understand other minds is that
there is a body of information that guides psychological attribution,
prediction, and explanation. This
body of information is often referred to as the child’s “Theory of Mind”, and it
has been the subject of intense empirical and theoretical investigations. The bulk of this research explores the
child’s developing capacity to understand the beliefs, desires, and perceptions
of others. For instance, the best
known result in this area is that children under the age of 4 tend to fail the
“false belief task”. In one version
of the false belief task, the child is shown a candy-box and asked what she
thinks is in the box. After the
child says that there is candy in the box, she is shown that, in fact, there are
pencils in the box. The box is then
closed and the child is asked what another person (who is not present) will
think is in the box. Three year
olds tend to say that the other person will think that there are pencils in the
box, while children over the age of 4 tend to answer correctly that the other
person will think that there is candy in the box. [7] These sorts of findings are taken as
evidence for the development or maturation of the child’s Theory of
Mind.
The core idea of the Theory
Theory of self-awareness is that the child’s capacity for understanding her own
mind depends on the same Theory of Mind that she uses to understand other
minds. Alison Gopnik has been
perhaps the most visible advocate for this view. Here is a representative statement of
the Theory Theory from Gopnik & Andrew Meltzoff:
Even though we seem to
perceive our own mental states directly, this direct perception is an illusion.
In fact, our knowledge of ourselves, like our knowledge of others, is the result
of a theory. [8]
The Theory Theory has also
been defended in the literature on autism.
A large body of evidence indicates that autistic individuals have severe
deficiencies in their understanding of other minds, and this has led researchers
to propose analogous deficiencies in autistic individuals’ understanding of
their own mental states.[9] Uta Frith and Francesca Happé express
the view as follows:
…if the mechanism which
underlies the computation of mental states is dysfunctional, then self-knowledge
is likely to be impaired just as is the knowledge of other minds. The logical extension of the ToM [Theory
of Mind] deficit account of autism is that individuals with autism may know as
little about their own minds as about the minds of other people. This is not to say that these
individuals lack mental states, but that in an important sense they are unable
to reflect on their mental states.
Simply put, they lack the cognitive machinery to represent their thoughts
and feelings as thoughts and feelings[10].
Theory Theorists haven’t been sufficiently clear about exactly how the Theory of Mind fits into the rest of the process of self-awareness.[11] But what is clear is that they regard Theory of Mind as a necessary component of self-awareness. All access to one’s own mental states is “theoretical” in the sense that it depends on the Theory of Mind. So one cannot detect one’s own mental states without exploiting the Theory of Mind that is used for detecting others’ mental states.
In response to this growing
consensus, Stich and I argued, rather, that the mind contains a “Monitoring
Mechanism”, a special purpose mechanism (or set of mechanisms) for detecting
one’s own mental states, and this mechanism is quite independent from the
mechanisms that are used to detect the mental states of others. On the theory we develop, the Monitoring
Mechanism (MM) takes as input one’s own mental state (e.g., a belief, desire, or
intention) and produces as output the belief that one has that mental
state. So, for instance, if one
believes that p and the Monitoring
Mechanism is activated (in the right way), it takes the representation p in the Belief Box and produces the
belief I believe that p. (See figure 1). This mechanism is computationally
extremely simple. For instance, to
produce representations of one’s own desires, the MM simply copies a
representation from the Desire Box, embeds the copy in a representation schema
of the form: I desire that ___., and
then inserts this new representation into the Belief Box. Our proposal, then, was that the
Monitoring Mechanism is an independent introspection mechanism for detecting
one’s own beliefs, desires, intentions and imaginings.[12]
_files/image002.gif)
Although we argue that the
Monitoring Mechanism theory is a much more plausible account of self-awareness
than the Theory Theory, we do not deny that one can use the Theory of Mind on
oneself. Indeed, we maintain that
Theory of Mind is probably required for reasoning about one’s own mental states,
i.e., using information about one’s own mental states to predict and explain
one’s own mental states and behavior.
However, it is quite a different matter for detecting one’s own mental states. The detection of one’s own mental states
does not depend on the capacity for
detecting or reasoning about other people’s mental states. Thus, on our account there are special
introspection mechanisms for detecting one’s own mental states but not for
reasoning about one’s own mental states.[13]
Evidence
Theory Theorists
have put forth a number of empirical arguments for the Theory Theory of self
awareness. Stich and I provide
detailed responses to these arguments and offer a few of our own arguments
against the Theory Theory.[14] I won’t rehearse all of those arguments
here, but I do want to review briefly one of the arguments against the Theory
Theory and then offer a couple of new arguments.
Development
asynchronies
The most explicit
and carefully charted argument for the Theory Theory comes from Gopnik &
Meltzoff.[15] According to the Theory Theory,
self-attributions will be subject to the same deficiencies as
other-attributions. Hence, young
children’s mistakes on attributing mental states to others should find parallels
in self-attribution. Gopnik &
Meltzoff maintain that in fact children’s understanding of their own mental
states does develop in close parallel
with their understanding of others’ mental states. However, a closer look at the
data suggests that the developmental evidence actually poses a problem for the
Theory Theory. For on a wide range
of tasks, children do not exhibit the parallel performance predicted by the
Theory Theory. Children are capable
of attributing knowledge and ignorance to themselves before they are capable of
attributing those states to others; they are capable of attributing certain
perceptual states to themselves before they are capable of attributing such
states to others; there is even some evidence that children are capable of
attributing false beliefs to themselves before they are capable of attributing
such states to others.[16] Hence, although Gopnik & Meltzoff
claim that the developmental evidence supports the Theory Theory of self
awareness, the evidence actually seems to undermine the Theory Theory in a
fairly serious way. The Monitoring
Mechanism account easily accommodates the data, however. The Monitoring Mechanism is proposed as
an innate and early emerging mechanism, and the account does not predict that
the capacity to detect one’s own mental states will develop in parallel with the
capacity to detect mental states of others.
Egocentric
attributions
A further argument against the Theory Theory emerges from the
developmental data when one considers the kinds of mistakes that toddlers make
about other minds. When asked what
another person, the “target”, thinks or wants, toddlers do not respond at
chance. Rather, for an important
class of cases, they tend to attribute their own mental states
“egocentrically”. Indeed, this is
how Theory Theorists characterize the
mistakes. For instance, in one
task, children are told to hide an object from the experimenter, and young 2
year olds failed this task. Gopnik
& Meltzoff describe the young children’s mistakes as follows: “24-month-olds consistently hid the
object egocentrically, either placing it on the experimenter’s side of the
screen or holding it to themselves so that neither they nor the experimenter
could see it.”[17] Similarly, Repacholi & Gopnik
found that 14-month old children shared the kind of food they themselves liked
rather than the food that the target exhibited a preference for. The experimenter made a facial
expression of disgust or happiness after tasting either Goldfish crackers or
broccoli. Although the 14-month
olds consistently shared the crackers (their own preference), the 18 month olds
were sensitive to the facial expressions of preference. Repacholi & Gopnik suggest the
possibility that although the 14-month olds “were beginning to acquire a
psychological conception of desire, it was, nonetheless, egocentric. Thus, although they understood that
people request things because of some underlying desire, they mistakenly believe
that everyone’s desires are the same.” [18]
Elsewhere, Meltzoff, Gopnik & Repacholi suggest that the performance of 18
month olds on this task indicates that they have a more developed Theory of
Mind: “This is a developmental achievement inasmuch as 14-month-olds did not do
this. Instead, they always gave the
experimenter crackers, their own preference, regardless of the experimenter’s
expressed desires. This work
suggests that even very young children, 18-month-olds, may have a nonegocentric
understanding of the differences between their own mental states and those of
others in some cases.” [19] These experimental findings of
egocentric desire attributions are corroborated by ecological reports.[20] For instance, when young children
help others in distress they tend to offer their own comfort objects (e.g.,
their teddy bear or blanket) to the distressed person.[21]
For the Monitoring
Mechanism theory, early egocentric errors in mindreading pose no problem. Even before the young child has an
adequate theory of, say, desire, she can use the Monitoring Mechanism to
determine her own desires and preferences, and she can subsequently attribute
her preferences to a target. By
contrast, it is hard to see how a Theory Theorist can accommodate egocentric
attributions. For if the child has
a deficient Theory of Mind, such that she is incapable of detecting the desires
of others, then the Theory Theory predicts that she should also be incapable of
detecting her own desires.
Egocentric mistakes indicate an asynchrony – the young child is
apparently aware of her own mental states and attributes them to others before
she is capable of detecting the other person’s distinctive mental states. Without further explanation, it is
difficult to see how a Theory Theorist can consistently maintain both that
toddlers have an early egocentric Theory of Mind and that one’s access to one’s
own mental states depends on the same theory that is used to detect the mental
states of others.[22]
Dissociations in
psychopathologies
In addition to
appeals to developmental evidence, a number of philosophers and psychologists
have recently argued that the Theory Theory of self awareness is supported by
psychopathological evidence on autism and schizophrenia.[23] The data on autism are of particular
significance for Theory Theorists, since autistic children are widely regarded
as having deficient mindreading capacities. For instance, autistic children continue
to fail the false belief task well after their mental age peers pass the
task.. And studies of spontaneous
speech indicates that autistic children basically never talk about cognitive
mental states like beliefs or thoughts. [24] Given the Theory of Mind deficit in
autism, the Theory Theory predicts that autistic children should be similarly
impaired at detecting their own mental states. Carruthers and Frith & Happé argue
that case studies indicate that autistic individuals do lack access to
their own mental states.[25] However, a close inspection of the
evidence tends to undermine rather than support the Theory Theory. [26] For instance, in a diverse range of case
studies, autistic individuals report their own mental states much better than
Theory Theorists predicted. The
Monitoring Mechanism theory has a ready explanation for this: the Monitoring Mechanism might be intact
in autism despite the deficit to Theory of Mind. [27]
Although the
evidence adduced by Carruthers and Frith & Happé seems to provide better
evidence against the Theory Theory
rather than for it, the evidence is in any case rather fragmentary, relying
largely on case studies. There is
new experimental evidence, however, that further confirms the claim that the
Monitoring Mechanism is intact in autism despite the problems with Theory of
Mind. In a recent set of studies,
Farrant and colleagues found that autistic children did remarkably well on
“metamemory” tests.[28] In metamemory tasks, subjects are asked
to memorize a set of items and subsequently to report on the strategies they
used to remember the items. In
light of arguments from Theory Theorists, the experimenters expected autistic
children to perform much worse than non-autistic children on metamemory
tasks: “On the basis of evidence
that children with autism are delayed in passing false belief tasks and on the
basis of arguments that mentalizing and metacognition involve related processes,
we predicted that children with autism would show impaired performance relative
to controls on false belief tasks and on metamemory tasks and that children’s
performances on the two types of task would be related.”[29] However, contrary to the researchers'
predictions, there was no significant difference between the performance of
autistic children and non-autistic children on a range of metamemory tasks. In one task, the subject was asked to
remember a set of numbers that were given.
The children were subsequently asked “What did you do to help you to
remember all the numbers that I said?”.
Like the other children in the study, most of the autistic children gave
some explanation that fit into the categories of 'thinking', 'listening' or
'strategies'. For instance, one
autistic child said “I did 68, then the rest, instead of being six, eight, you
put 68.” Indeed, Farrant et
al. claim that it is clear from the data that “there was no relation between
passing/failing false belief tasks and the categories of response given to the
metamemory question.”[30] Although the results flouted the
experimenters’ Theory-Theory-based prediction, they fit perfectly with the
Monitoring Mechanism theory.
For the Monitoring Mechanism can be intact even when Theory of Mind is
damaged.
The Introspection
Module
Thus, the Monitoring
Mechanism theory fits the available evidence much better than the Theory
Theory. Of course, historically,
the most venerable account of introspection is not the Theory Theory (which is
likely an invention of the 20th century), but rather that
introspection is a species of perception, inner perception. On traditional inner perception models,
one detects one’s own mental states via experience or phenomenological
features.[31] If the perception-model of introspection
is developed in this way, then the MM theory diverges in an important way. For the Monitoring Mechanism does not
rely on phenomenological features for
identifying one’s beliefs and desires. [32] In that sense, on the Monitoring
Mechanism account, introspection is quite different from perception. However, there is an important way in
which the Monitoring Mechanism might be akin to perception: it’s plausible that both systems are
modularized.
Perceptual systems are the
paradigm examples of modules, as modularity is developed in Fodor’s classic
treatment in The Modularity of Mind. The central feature of modularity for
Fodor is informational encapsulation. [33] A cognitive mechanism is encapsulated if
it has little or no access to information outside of its own proprietary
database. Perceptual systems tend
to be encapsulated – there are restrictions on the kinds of information that are
processed by perceptual systems.
The classic illustration of perceptual encapsulation is the fact that the
Müller-Lyer illusion persists even after one knows about the illusion. Apparently, the perceptual system is
insensitive to the knowledge that the lines are the same length. Fodor maintains that there are a number
of other features that tend to co-occur with encapsulation. And Fodor maintains that perceptual
systems also have these correlated features. Among other things, perceptual systems
tend to be dedicated to particular tasks, they have characteristic ontogenies,
and they exhibit characteristic patterns of breakdown, and perceptual processing
tends to be very fast.
The Monitoring Mechanism, like the perceptual systems, has many of the
features of modules. Like
perceptual systems, the Monitoring Mechanism is dedicated to a particular task,
it has a characteristic and early ontogeny, it seems to exhibit a characteristic
pattern of breakdown[34],
and it seems to be selectively spared in autism. It also exhibits fast, but restricted
processing. The last point is of
some significance. The processing
capacity of the Monitoring Mechanism is extremely limited – it simply plugs a
representation into the self-attribution schema. Thus, while it may not be strictly
speaking encapsulated, the Monitoring Mechanism resembles encapsulated
mechanisms in that it does not engage in any remotely intelligent
general-purpose reasoning. We might, then, think of the Monitoring Mechanism as
the Introspection Module.[35]
Possible Functions of
Introspection
The evidence suggests, then, that the capacity for detecting one’s own
mental states is subserved by a Monitoring Mechanism that is quite independent
from Theory of Mind. One might
wonder, though, why we would have
such a mechanism.[36] The problem is especially acute if
one assumes that the Monitoring Mechanism evolved before
Theory of Mind. As noted above, the
developmental evidence indicates that the Monitoring Mechanism emerges earlier
than Theory of Mind. So one might
suppose that, if ontogeny recapitulates phylogeny, then the Monitoring Mechanism
should be phylogenetically older than Theory of Mind. Of course, the claim that ontogeny
recapitulates phylogeny is hardly a strict law.[37] So it is possible that the Monitoring Mechanism
emerged after or contemporaneously with Theory of Mind. Nonetheless, the claim that ontogeny
recapitulates phylogeny has been a good heuristic, and it’s probably better to
be on the side of the heuristic rather than against it.
The clearest way to develop and confirm evolutionary accounts of
cognitive mechanisms is by appeal to evidence comparing distantly related
species that share a cognitive mechanism or related species in which the
cognitive mechanisms diverge.[38] Unfortunately, in the case of
introspection, we have no comparative evidence that speaks to this issue
directly, so it’s difficult to identify the evolutionary function of the
Monitoring Mechanism with any confidence.
However, to forestall the criticisms that such a mechanism would have
been good for nothing, let me make clear a couple of different ways in which the
Monitoring Mechanism could have been adaptive.[39]
One of the arguments I presented above against the Theory Theory is the
fact that young children tend to attribute their own mental states egocentrically. Although egocentric
attribution typically shows up most clearly in the mistakes that children (and
adults[40])
make, in fact, the practice of egocentric attribution likely forms a large and
productive part of our mindreading abilities. In trying to figure out another person’s
mental states, one quite successful strategy for a large set of states is to
attribute one’s own mental states to the target. For we typically share a broad
background of similar beliefs with those we attribute beliefs to. This is also true for a wide range of
preferences – my conspecifics and I typically have similar tastes and basic
desires. Indeed, even in the Goldfish crackers & broccoli experiment, it’s
likely that the young infants who “mistakenly” attribute their own preferences
to the target are adopting a fairly effective strategy. Thus, a mechanism that provides access
to one’s own mental states might provide a basis for attributing mental states
to others. And there are lots of
reasons to think that it’s adaptive to be good at attributing mental states to
others.[41]
In a quite different way, monitoring one’s own mental states might have
been useful to enable more efficient planning. Perhaps the most influential account of
planning in recent philosophy and artificial intelligence is Michael Bratman’s
“planning theory of intention.”[42] Although it is not entirely explicit in
Bratman’s planning theory, the capacity to detect one’s own mental states can
play a crucial role in enabling efficient planning, as I hope to
explain.
Since there are always indefinitely many possible courses of action and
the world is constantly changing, ideal practical reasoning is quite
impossible. It would require a
constant assessment of the best thing to do, and humans have neither the leisure
nor the capacity for such endless calculation. Bratman argues that one way out of this
problem for a resource-limited creature is by committing to an intention in such a way that one no
longer deliberates all things
considered about what to do.
Rather, one’s commitment to an intention constrains and structures
subsequent planning and decision making.
One crucial part of the theory is that once you commit to a plan, there
is a wide range of incompatible options that you don’t even consider – they are
“filtered” out.[43] For instance, if I commit to going
to England on July 15th, then I don’t even consider the option of
having a dinner party on July 16th. Of course, this deliberative
neglect of a range of options carries a certain cost – for some of the options
that aren’t considered might have been adopted had the options been considered
in a thorough process of deliberation. However, in many other cases, the options
that are neglected would not have been adopted in any case, so in those
(presumably, more typical) cases, one saves the time and energy of deliberation
without incurring any costs. The
other crucial feature of committing to an intention is that this structures
one’s subsequent deliberations, e.g., about the best way to execute the
plan. So, once I’m committed to
going to England, I generate the subplan that I need to get my passport
renewed. In the long run, this
filtering and structuring plausibly makes one’s planning much more efficient
than a constant calculation of utility maximization.
In her work in Artificial Intelligence, Martha Pollack has argued that
another way that resource-bounded agents can make their reasoning more efficient
is by “overloading” their intentions.[44] The idea is that if an agent has a goal,
she can try to determine whether that goal can be satisfied in the course of
executing some plan that has already been adopted. So, for instance, suppose I realize that
I need to buy a present for a party that will occur tomorrow. I can then consider whether I already
have a plan that will take me near an appropriate vendor. If I had already planned to go downtown
to the post office, I can “overload” this prior plan to include a trip to the
store. Not only will this make the
errand-running more efficient, Pollack suggests that by overloading one’s
intentions in this way, one’s reasoning is also more
efficient.
How does all this connect to the utility of the Monitoring
Mechanism? What I want to suggest
is that overloading or committing to one’s intentions is enabled by having a
mechanism that delivers beliefs about one’s intentions. Since Bratman and
Pollack develop their theories in contexts in which it is simply assumed that
the reasoning agents (be they human or artificial) can represent their own
intentions, this part of the theory is never made explicit. However, unless one
knows about one’s own intentions, it’s difficult to see how one can overload
one’s intentions. Similarly if you
don’t know which intentions you’re “committed” to, it’s hard to see how those
intentions can subsequently inform one’s planning. The obvious way to implement
the kind of reasoning that Bratman & Pollack promote would involve having beliefs about your own intentions. If the agent has a belief about what she intends to do,
this belief can structure her further decision making. Of course, the Monitoring
Mechanism would serve the function of delivering such beliefs. Hence, it seems that this mechanism
might play a crucial role in facilitating efficient planning. Again, I’m not arguing that the
Monitoring Mechanism actually evolved to serve this function. To make such a claim plausible one would
need a body of comparative evidence.
Rather, the point is to show that there could be fairly direct advantages
to having a Monitoring Mechanism, so the mechanism can’t be faulted on general
evolutionary grounds.
To summarize this section, I’ve offered a number of arguments for why the
Monitoring Mechanism account of introspection is more plausible than the rival
Theory Theory account. I’ve also
proposed some reasons to think that such a mechanism could have been
adaptive. And I’ve suggested
that this Mechanism is modular.
There certainly is not sufficient evidence for this claim to parade it as
an obvious truth. But given the
available evidence, it is a plausible conjecture that there is a modular
Monitoring Mechanism – a dedicated mechanism for detecting one’s own mental
states that is independent of the capacity to detect mental states of others.
III.
THE SELF-CONCEPT IN COGNITIVE SCIENCE
If the foregoing account of introspection is right, it might have
significant implications for psychological work on the concept of self.[45] For the Monitoring Mechanism account
suggests that there is a basic concept of self that has been largely neglected
in the psychological literature.
There is a vast literature on the self in psychology elaborating, often
in great detail, how people think of themselves. Much of this work focuses on issues that
relate to self-esteem and self-worth.[46] But here I want to focus on the ontogeny
of “self-conception”, i.e., when children develop a concept of self. Perhaps the best known method for
addressing this question is the mirror self-recognition task.[47] More recently, in light of the research
on mindreading, researchers have suggested rather that the child’s concept of
self depends on the Theory of Mind.[48]
If the Monitoring Mechanism theory is right, then there is a notion of self that
isn’t captured by either of these approaches.
The
Body’s “I”
There is a long tradition in developmental psychology and primatology of
using “mirror self-recognition” tests to descry awareness of self in infants and
nonlinguistic animals. The first study on mirror self-recognition was reported
by Gordon Gallup.[49]
Gallup found that chimpanzees respond in self-exploratory ways to their images
in mirrors; for example, they inspect parts of their body that are difficult to
see without a mirror, and they will reach up to investigate marks that were
surreptitiously placed on the forehead before the mirror was made available.
Subsequent research on human children showed that children begin to exhibit this
kind of behavior by 18-24 months. [50] A wide range of further comparative
research has found, surprisingly, that most species do not exhibit this kind of
self-exploratory behavior. For
instance, it has not been convincingly demonstrated in any species of monkey.[51]
Although mirror self-recognition is sometimes treated as evidence that
chimpanzees have a psychological understanding of themselves[52],
there are fairly obvious alternative explanations. For instance, Daniel Povinelli maintains that passing the mirror tasks
doesn’t require a concept of self as psychological subject: “Gallup believes that chimpanzees
possess a psychological understanding of themselves. In contrast, I believe these apes
possess an explicit mental representation of the position and movement of their
own bodies – what could be called a kinesthetic self-concept.”[53] On Povinelli’s account it is the
“contingency between the self’s actions and the actions in the mirror” that
“triggers the formation of an equivalence relation between the organism’s
internal self-representation and the external stimuli (the mirror image)”.[54] In fact, as Povinelli points out, this
is similar to the proposal Gallup made in his initial paper reporting the
findings. “Gallup’s (1970) initial
speculation was that ‘self-directed and mark-directed behaviors would seem to
require the ability to project, as it were, proprioceptive information and
kinesthetic feedback onto the reflected visual image so as to coordinate the
appropriate visually guided movements via the mirror’ (p. 87)”.[55] So, while chimpanzee & human
toddler’s self-exploratory behavior in front of mirrors is plausibly evidence
that chimpanzees and toddlers have a concept of the self-as-body, the Body’s “I”, it does not show that they
have the concept of self as a mind.
The capacity for kinesthetic feedback may suffice for generating mirror
self-recognition.[56]
What does the absence of
self-exploratory behaviors in front of mirrors show? It does not show that the creature lacks
any concept of self. Indeed, it may
well be the case that creatures can “fail” the mirror task while having a
psychological understanding of the self. If, for instance, Povinelli’s theory of
mirror self-recognition is right, then a system of kinesthetic monitoring plays
a crucial role in mirror self-recognition.
As a result, creatures that lack this kinesthetic monitoring might well
fail the mirror self-recognition task.
But it’s quite possible that a creature can have an understanding of the
self-as-mind without kinesthetic monitoring. A subject that lacked any
kinesthetic monitoring might know that people (including himself) have beliefs
and desires without being able to detect the contingencies between his bodily
movement and that of his mirror image.
Indeed, the possibility of a dissociation between mirror self-recognition
and self-awareness has recently been given some empirical support. Breen and colleagues report cases in
which subjects have deficits in mirror self-recognition (i.e., they do not
recognize themselves in mirrors) but apparently no deficit in self-awareness.[57]
The upshot is that while mirror self-recognition provides evidence of
some concept of self, the mirror tasks don’t tell us much about the subject’s
understanding of herself as a subject of psychological states. It’s possible to
exhibit mirror self-recognition while having no understanding of the
self-as-mind, and it’s possible that creatures can have an understanding of the
self-as-mind even if they don’t exhibit mirror
self-recognition.
The Theory of Mind’s ‘I’
Although mirror tasks seem to provide no evidence that a creature understands itself as having psychological states, the work in Theory of Mind provides a wealth of evidence on this in humans. There is little question now that pre-school children understand that they (and others) have psychological states. Thus, the child’s Theory of Mind obviously provides the basis for a psychological understanding of the self.
Unfortunately, the task of characterizing the young child’s concept of
self has not yet received sustained attention by researchers drawing on the
Theory of Mind tradition. However,
the view that the concept of self depends on Theory of Mind has been suggested
by various authors in different ways.[58] It’s also implicit in the Theory Theory
of self awareness, since on that theory, the child’s understanding of her own
psychological states depends on the Theory of Mind. Perhaps the most explicit
statement that the concept of self depends on Theory of Mind comes from Henry
Wellman:
our understanding of
ourselves partakes of and is limited by our framework belief-desire psychology….
Thoughts, desires, basic emotions, actions, perceptions and so on, as specified
by belief-desire psychology, are the basic categories that frame specific person
concepts. Everyday theory of mind
provides the infrastructure for self-conception.[59]
Wellman doesn’t directly
consider mirror self-recognition, but it’s reasonable to assume that he would,
in line with the discussion above, maintain that mirror self-recognition does
not show anything about the child’s understanding of the self as a psychological
subject. Once we put the concept of
self-as-body to the side, Wellman seems to be suggesting that the child’s
concept of self depends crucially on the Theory of Mind. We might call this
concept the Theory of Mind’s
‘I’.
As noted above, there is disappointingly little written in this tradition
on the child’s concept of self.
Earlier developmental work on the child’s concept of self produced the
bizarre finding that children seemed to identify the self primarily with
physical features in free recall tests.[60]
For example, when young children are asked “What will (not) change about
yourself when you grow up?”, 7 year olds tended to refer to physical
characteristics (e.g., hair color) and very seldom referred to psychological
characteristics.[61]
Indeed, one prominent view was that young children had only a “physicalistic”
conception of the self. [62]
This body of findings seemed increasingly peculiar in light of the emerging body
of evidence on the young child’s extraordinary capacities in Theory of
Mind. The research on Theory of
Mind clearly demonstrates that young children attain considerable sophistication
about the mind well before the age of 7.
Hence, it’s puzzling that they should identify themselves solely with
their physical features.
In a set of recent experiments motivated partly by these considerations,
Daniel Hart and colleagues found that although children tended to appeal to
their physical features in standard free-recall tasks about the self, the same
children regard their psychological
characteristics as most important to their self.[63]
Hart and colleagues used philosophically-inspired thought experiments on
personal identity to explore this.
In one condition, the child is shown a model of a “person machine” and is
told the following:
This is a person
machine. What the machine does is
make persons. The person behind
this door gets an exact copy of your body and looks exactly like you. But this person… does not have your
thoughts and feelings…. The person behind this door… has an exact copy of your
thoughts and feelings… But this person does not have your body or look like
you.[64]
The subject was then asked,
“Which of these two persons is closest to being you? The one with your body and appearance or
the one with your thoughts and feelings?”
The researchers compared children’s responses to the person machine task
with their responses to standard free-recall tasks, and found that in the free
recall tasks, the children mentioned physical characteristics most often, but in
the person machine task, they regarded their psychological qualities as most
important for the self. Hart and
colleagues describe the results as follows:
When asked to judge which
set of characteristics was most important for establishing similarity between
the self and a hypothetical person, the 7-year-olds in this study most
frequently claimed that it is their psychological features; indeed, over half of
the children claimed that the psychological characteristics are superior for
preserving personal identity for all the hypothetical transformations posed in
this study.[65]
Philosophers will no doubt
notice that this isn’t really a question about personal identity, since it is explicitly about
similarity between simultaneously existing persons. However, many children would be upset by
imagining that they are dismantled shortly after stepping into the person
machine, and ethics review boards would be unlikely to approve such a study
should a sadistic experimenter propose it.
At any rate, as Hart and colleagues intimate, the person machine task is
likely more revealing than free-recall tasks for uncovering the child’s theory
of the core features of the self.
Hart and colleagues developed their task explicitly in the context of Theory of Mind research. And the Theory of Mind is plausibly implicated in the person machine task since the task requires the subject to judge the importance of psychological properties for similarity across individuals. The findings on the person machine task begin to tell us a bit about the Theory of Mind’s “I”, then. They suggest that on the concept of self delivered by the Theory of Mind, the most important features of the self are one’s psychological properties.[66] Presumably children don’t have this understanding of self until they have a Theory of Mind. Hence, at least for this concept of self, Wellman is right: “Everyday theory of mind provides the infrastructure for self-conception.”
The Mind’s
“I”
So, the Theory of Mind apparently delivers a concept of self as a mind,
and this concept needs to be distinguished from the concept of
self-as-body. However, the
Monitoring Mechanism account suggests that there is another, more basic concept
of self that is independent of Theory of Mind, what one might call the Mind’s “I”.
The Monitoring Mechanism sketched in section II produces beliefs about
one’s own mental states. That is,
it produces representations of the form I
believe p, I desire q, I imagine r, etc. Thus, as the source of such
self-attribution, this mechanism delivers representations in which the concept,
I, is the subject of mental
states. Since the Monitoring
Mechanism is presumed to be innate, one of the implications of the account is
that this concept of the self as subject is also innate.
The Mind’s “I” is distinct from both of the self-concepts discussed
above. It’s distinct from the Body’s “I” since the Mind’s “I” is explicitly the
subject of psychological states.
There is no a priori reason to
think that this concept of self will covary with the concept of self that is
revealed through mirror self-recognition.
It’s certainly possible that the concept of the Mind’s “I” is present in
creatures that lack the concept of the Body’s “I”. For instance, if Povinelli is right,
mirror self-recognition depends on a keen system of kinesthetic monitoring in
humans and chimpanzees. But of course, the evolutionary pressures that led to a
keen system of kinesthetic monitoring may well be quite different from those
that led to mental state monitoring. And, as noted earlier, a creature might
well be able to recognize its own mental states without recognizing the
contingencies between its movements and the movements in the mirror.[67]
Thus, the Monitoring Mechanism delivers a concept of the self as a mind,
rather than a body. However, this
is a concept of self-as-mind that does not depend on Theory of Mind. Rather, it is present in humans before
Theory of Mind has matured, and it may be present in creatures that don’t have Theory of Mind. In contrast to Wellman’s claim, then,
there is a basic concept of the self as a psychological subject for which the
Theory of Mind does not provide the infrastructure. And while the concept of self that
depends on Theory of Mind may well exhibit cross-cultural variation[68],
the Mind’s “I” is unlikely to be cross-culturally variable for it is a direct
output of an innate and early emerging module.
It’s worth emphasizing that on the Monitoring Mechanism account, the fact
that this basic concept of “I” is delivered by introspection in no way suggests
that there is a phenomenologically salient sense of the self as psychological
subject. That is, this basic
concept of the self as a psychological subject comes from a specialized
computational module, not from any phenomenological features that the self might
have.[69]
Thus far, I’ve focused largely on what this concept of self is not; it is
harder to provide a positive characterization of the concept since psychologists
have not studied it systematically.
By hypothesis, one of the functional properties of this concept of self
is to underwrite self attribution, and the concept is exploited by young
children’s early attributions of mental states to themselves; this concept is
presumably also connected to action systems. But what does this self-concept specify
as the core features of the self?
It’s likely that this concept of self contains virtually no information
about the essential features of the self.
As we’ve seen, the concept of self that depends on Theory of Mind seems
to provide a rich characterization of the essential features of the self on
which a person’s psychological properties are crucial to the self. By contrast, the computational
characteristics of the Monitoring Mechanism suggest that, while this simple
mechanism does deliver a concept of self, this concept does not include any
specification of the core features of the self.
Despite its exiguous
content, this concept of self can support judgments of personal
identity. I currently desire that
it not rain, since I’ve discovered that my roof is leaking. And the Monitoring Mechanism can detect
this desire. However, before I
discovered the leak, I desired extensive rain to relieve the dustbowl that is my
backyard. So, several weeks ago,
the Monitoring Mechanism produced the belief that I want it to rain, and that
belief was converted into the memory that I wanted it to rain. More recently, my desires have changed,
and the Monitoring Mechanism produces the belief that I currently desire that it
not rain. So, the Monitoring
Mechanism is the basis for the representations I wanted it to rain and I want it not to rain.
And it’s plausible that I can use such beliefs as the basis for the
judgment that although I wanted rain previously I do not want it now. Of course, I can have several
mental states at once, and the Monitoring Mechanism might deliver, e.g., the
belief that I currently desire to go to the store and to listen to music. The upshot of this is that the
Monitoring Mechanism provides the basis for making judgments of personal
identity both synchronically and diachronically. I am the same person who wanted it to
rain several weeks ago and currently wants it not to rain. I am also the same person who wants to
go to the store and to listen to music.[70]
Although it hasn’t been
systematically studied, it’s likely that these kinds of judgments of identity
are implicit in self-attributions of young children. Consider, for instance, the following
exchanges from the CHILDES database:
Abe (3;3): I didn’t
get you a surprise.
Adult:
You didn’t. I’m
sad.
Abe:
No, don’t be sad. I thought
I would ‘cept I didn’t see one for you.