1. The plot
Perception and action are
more deeply interdependent than we usually assume. As well as having
empirical implications, this interdependence has consequences for
philosophical issues: about the unity of consciousness, relations
between mind and world, self-consciousness, cognition, behaviorism,
etc. This general theme is developed by these essays on various
aspects of consciousness and agency. While they can be read separately,
if the essays are read in sequence the attentive reader will discern
an overall plot and several subplots.
The
place of mind and its norms in the natural world has long seemed
problematic. But finding a satisfactory conception of relations
between mind and world is made more difficult by a certain tendency
of thought. This is the tendency to regard perception and action
as buffer zones mediating between mind and world. We tend to think
of perception as input from world to mind and action as output from
mind to world. This Input-Output Picture of perception and action may hold in place traditional
worries about the mind's place in the world, as well as more specific
philosophical assumptions. If perception is input from the world
to the mind and action is output from the mind to the world, then
the mind as distinct from the world is what the input is to and
what the output is from. So, despite the web of causal relations
between organisms and environments, we suppose the mind must be
in a separate place, within some boundary that sets it apart from
the world.
In
trying to understand the mind's place in the world, we thus study
the function from input to output, especially the way central nervous
systems process and transform inputs to human organisms. We argue
about whether central cognitive processes must have a language-like
structure that explains the conceptual structure of thought. But
we tend to ignore the function from output back to input, and the
way environments, including linguistic environments, transform and
reflect outputs from the human organism. The two functions are not
only of comparable complexity, but are causally continuous. To understand
the mind's place in the world, we should study these complex dynamic
processes as a system, not just the truncated internal portion of
them.
People
and other animals with minds can be seen at one level as dynamic singularities:
structural singularities in the field of causal flows characterized
through time by a tangle of multiple feedback loops of varying orbits.
Consider the circus performer who puts the handle of a dagger in
her mouth, tips her head back, balances a sword by its point on
the point of the dagger, and with the whole kit balanced above her
head magisterially climbs a ladder, swings her legs over the top
rung, and climbs back down the other side of the ladder. Each move
she makes is both the source of and exquisitely dependent on multiple,
internal and external, channels of sensory and motor-signal feedback,
the complex calibrations of which have been honed by years of practice.
An only slightly less intricate structure of dynamic feedback relations
knits the nervous system of a normally active organism into its
environment. This is what the contents of the creature's interdependent
perceptions and intentions both depend on. The whole complex dynamic feedback system includes not just functions from input to output,
but also feedback functions from output to input, some internal
to the organism, others passing through the environment before returning.
As a result, external states can be needed to explain patterns of
activity at the body surface, even if what is to be explained is
not identified in terms of external states. The dynamic singularity
is centred on the organism and moves through environments with the
organism, but it itself has no sharp boundaries.
The
dynamic singularity conception is at the subpersonal level. We should
not confuse different levels of conceptualization: the subpersonal level of
causal processes on which mind depends with the personal
level of normatively constrained mental
contents. When we conceive of intentional agents at the personal
level, we think about the relations between what they perceive and
intend, what they believe and desire, and try to make sense of them
as acting for reasons, though of course allowing for irrationality
and mistakes. By contrast, causal explanations in neurophysiological
or computational terms describe subpersonal mechanisms and functions.
The personal level contents of mental states can be seen as carried
by such subpersonal processes, or vehicles of content. But the properties of subpersonal processes,
of vehicles of content, cannot simply be projected into personal
level mental content, or vice versa. These different ways of looking
at and describing an organic system need not display isomorphisms
or map onto one another in any simple way. As with emergent properties
of dynamic feedback systems in general: Significant qualitative
differences at one level may depend on minor quantitative differences
at the other level. Discontinuities, intricate or salient structure
at one level may be invisible at the other.
But
the Input-Output Picture does confuse levels. It confuses the subpersonal-level
distinction between causal input and causal output with the personal-level
distinction between perception and action. Or at least it maps these
two distinctions onto one another in an overly simply way. Consider
instead the view that the contents of perceptual experience and
intentional action both depend on a structure of causal flows that
constitutes a complex dynamic feedback system. In other words, both
depend on relations between inputs and outputs. This provides a
two-level account of the interdependence of perception and action
that is characteristic of having a perspective. It corrects the
buffer zone view: the picture of perception as input and action
as output. Perception and action can be more intimately related
to one another than that picture allows. Neuropsychology provides
examples that illustrate and support this correction.
The
main plot in a nutshell is this. It will be argued that the unity
of consciousness has both normative, personal-level and subpersonal
aspects. The relations between these levels can be approached via
the closely related but more general idea of a perspective: the
interdependence of perception and action that perspective involves
can be explained in terms of their co-dependence on a subpersonal
dynamic singularity. This subpersonal aspect of unity does not support
sharp causal boundaries either between mind and world or between
perception and action. Moreover, it can provide an antidote to the
inward retreat of the mind in modern philosophy. At the personal
level, the self does not lurk hidden somewhere between perceptual
input and behavioral output, but reappears out in the open, embodied
and embedded in the world.
These
themes are developed in two parts. The first part focuses on the
unity of consciousness, and the second on the relations between
perception and action. We'll preview the two parts in a bit more
detail, and then several of the subplots. For ease of reference,
terms and phrases that play an important role in the essays will
be highlighted here as they are briefly and informally explained.
The
essays need not be read in the order presented. Those wishing to
dip into the book rather than read it straight through should use
the analytical table of contents to choose according to their interests
while also getting a sense of the context provided by other essays.
So that the essays can be read separately, some arguments and examples
are repeated in different essays. However, several of the longer
essays have been divided into parts, which are not intended to be
self-standing. The book as a whole is written for an interdisciplinary
audience, and some elements of it are likely to be more accessible
to philosophers and others to scientists. Readers with one background
should understand that some passages may be directed to readers
with a different background, and read accordingly. Philosophers
may find the degree of detail in discussion of empirical examples
and related thought experiments too great, but scientists may find
it helpful. Scientists may wish to skip over stretches of unrelieved
philosophical argument (especially in essays 2, 3, 6, 7) and may
find the discussion of detailed examples and of empirical work of
more interest (especially in essays 5, 9, 10).
2. Part 1: The unity of consciousness and
action
Some conscious states occupying
the same stretch of time are together, while others are separate.
While I talk to you, I see your face and hear my own voice. These
experiences are together or united within one consciousness. But
you also hear my voice, and your experience is separate from mine.
This is an easy case. But what is it for simultaneous conscious
states to be united or separate? Can we find a principled account
of the unity of consciousness at a time that will also apply to hard cases? Or
are there just various differences between the cases, no right answers
in some cases, and no unified phenomenon of 'the unity of consciousness'?
(Essays 2, 3, 5)
Kantian
and neo-Kantian arguments make it difficult to account for the unity
of consciousness in either of two ways. In the Critique
of Pure Reason Kant argues that the
unity of consciousness cannot be understood in terms of the contents
of experience. But he also argues, in effect, that it cannot be
understood in terms of the unity of the self as a thing in itself
either. This Kantian dilemma has a modern descendant: there are difficulties for
both subjective and objective accounts of the unity of consciousness.
(Essay 2)
According
to a traditional understanding of the subjective realm, the contents
of consciousness are in principle independent of the external world.
Such a view is vulnerable to the 'just-more-content'
argument: the unity or separateness
of consciousness cannot be accounted for in terms of the subjective
contents of consciousness, because the same question of unity or
separateness arises again for any such contents. An objective account
is needed. If there has to be an answer to the question about whether
conscious states at a given time are united or separate, then the
subjective realm is not in principle self-sufficient. We can consider
whether consciousness could be only partially unified, so that state
1 might be co-conscious with state 2 and state 2 with state 3 even
though states 1 and 3 were not co-conscious. But even if we think
such partial unity
of consciousness makes sense, it turns out that an objective account
is still needed. (Essays 2, 3)
When
we conceive of intentional agents at the personal level, we make
sense of what they do holistically: in terms of the relations between what they perceive
and intend, or what they believe and desire. What action a perception
leads to depends on the contents of the agent's intentions, and
vice versa. We also consider the consistency of someone's various
perceptions and beliefs, or various intentions and desires. Of course
people can be irrational. But we make sense of their irrationality
against a background of normative assumptions. Relations between
the contents of mental states such as perceptions and intentions
are constrained, at least weakly, by norms, such as norms of consistency
and instrumental rationality. If such norms are violated too radically,
we begin to lose our grip on why the behavior in question counts
as an expression of the mental states of an intention agent at all.
Norms
of consistency at the level of content play an important role in
the unity of consciousness. In some cases and under certain assumptions,
inconsistent contents can imply disunity. But a normative approach
to unity can only provide part of the needed objective account,
since consistency is not sufficient for unity: the contents of separate
conscious states can be consistent with each other. We can indeed
make sense of complete duplication of content in separate centers
of consciousness, and norms of consistency could not account for
their separateness. Some further component of an objective account
is needed. (Essay 3)
Neither
the contents of consciousness nor the norms of consistency governing
contents fully account for unity. Perhaps we should look instead
to subpersonal level for the needed further component of an objective
account. But now we run into another difficulty. We should not confuse
the unity of consciousness with the unity of subpersonal vehicles
of consciousness. We cannot simply project properties or structures
present at one level onto the other level. There need not be a spatiotemporally
unified locus within the brain of unified consciousness, and neuroanatomical
structure may not be isomorphic with the structure of consciousness.
Unity might take various neurophysiological forms. (Essays 1, 5)
Given
the difficulties with both subjective and objective accounts of
the unity of consciousness, we get both a clue and a lesson from
reconsidering the Kantian dilemma. Kant tries to slip between the
two horns of this dilemma by accounting for unity in terms of a
special sense of self-consciousness--"transcendental
apperception". This is neither
consciousness of an empirical self as represented in experience,
nor consciousness of the self as a thing in itself. Rather, it is
consciousness of the possibility of spontaneous conceptualizing
activity. The clue is that the unity of consciousness may essentially
involve activity in some way. It is wrong to think of perceptual
consciousness as passively receptive. The lesson is that much turns
on how the receptive and active elements of consciousness are distinguished.
Intentional actions raise all the same issues about content and
unity that perceptual experiences do. By appealing to intentional
activity in an account of the unity of consciousness we run the
danger of merely relocating but not solving the problem.
How
can this danger of regress be avoided, while recognizing the essentially
active nature of perception? The distinction between perception
and action applies at the personal level. The distinction between
input and output applies at the subpersonal level. These are both
good distinctions, but we should not project them onto one another
by assuming that perception is input from world to mind and action
is output from mind to world, as in the buffer zone view. Nor should
we simply take the content and unity of intentions for granted in
explaining the content and unity of experiences.
But
consider instead the possibility that the contents of experiences
and intentions are both functions of a complex dynamic system of
relations between input and output, including both internal and
external feedback loops. This could explain the interdependence
of perception and action, which is part of what it is to have a
perspective:
some of the ways in which what you experience and perceive depends
systematically on what you do, as well as vice versa. (Essays 2,
4, 5)
Such
a Two-Level Interdependence View can avoid the danger of regress.
It can also avoid the danger of confusing properties of subpersonal
vehicles of content with properties of content. So it could in principle
contribute to understanding how the needed subpersonal component
in an account of the unity of consciousness could avoid this danger.
Suppose that content is a function of (or, if you prefer, supervenes
on, or is determined by, or depends constitutively on) various causal
processes. These may include not just internal vehicles of content,
but also external states and causal processes. Content may be a
function of distributed as well as local processes, of relations
within the organism or between organism and environment as well
as of intrinsic properties of discrete neural structures.
However,
which
function of these factors determines content cannot simply be read
off from the properties of subpersonal vehicles. This is the confusion
of levels we're trying to avoid (though without denying that there
is anything to be said about how the levels might be related). Emergence
can be unpredictable. And different types of content, such as the
contents of perceptions and the contents of intentions, might be
different
functions of the same system of causal relations: different contents
can be superposed
on the same network of relations. Which functions of these relations fix contents of certain
types could depend on still broader causal processes, such as evolution,
which explain why any subpersonal vehicles of a certain type, or
any mental states of a certain type, exist at all.
However,
we are now in a position to see how a Two-Level Interdependence
View can explain the interdependence of perception and action without
confusing the personal and subpersonal levels. If the contents of
perceptions and of intentions are different functions of the same
system of relations, then as a purely logical matter a given change
in those relations could in principle affect both the contents of
perceptions and of intentions. There could be interference or 'crosstalk' between
perception and action, as is found when different contents are superposed
on the same neural network (essays 1, 10). This phenomenon could
have beneficial functions. But for now what matters is that this
reasoning does not confuse properties of subpersonal vehicles with
properties in content, or assume that content is a transparent function
of subpersonal properties. We can respect the distinction between
the personal and subpersonal levels while saying something about
the relations between them.
The
general idea of such a Two-Level Interdependence View can be made
more specific by considering the role of motor-to-sensory feedback
in various empirical cases. In some cases, the unity of perceptual
consciousness may vary with the content of simple motor intentions
even when sensory inputs are constant. These cases both challenge
the Input-Output Picture and suggest that the unity of consciousness
may involve feedback. (Essay 5)
Could
the functional idea of a complex dynamic feedback system provide
the needed further component of an objective account of unity? To
understand the unity of consciousness we may need to understand
the interdependence of perception and action involved in having
a perspective. This is the topic of the essays in the second half
of the book.
3. Part 2: The interdependence of perception
and action
How are states of the external
world and the contents of the mind related? A traditional view,
often associated with the influence of Descartes, is that mental
content is in principle independent of the world or autonomous ('narrow').
Some modern views deny this, and hold that content is world-involving
or context-dependent ('wide'). The traditional view of content generates
skeptical worries about how knowledge of the world is possible:
why couldn't our beliefs about the external world be rampant delusions
produced by mad scientists manipulating our brains? By contrast,
views of content as world-involving can be seen to emerge from worries
about how content itself is possible. For it even to be possible
that our beliefs are mistaken, they must be about something in particular,
must have determinate content. But in virtue of what could they
count as having any particular content, if not their relations to
the world? Wittgenstein's later work is sometimes interpreted in
this way. (Essays 6, 7)
Consider
the traditional view that the contents of the mind are in principle
independent of the world, so that we might be systematically deluded.
If dualism is ruled out, it is natural to understand this view as
holding, first, that internal physical states fix mental contents
(Internalism)
and, second, that it makes sense in principle to suppose that internal
physical states can be duplicated in systematically different environments
(the Duplication Assumption). This yields the possibility of systematic delusion:
mental contents can be duplicated in different environments. (Essay
8)
Two
kinds of thought experiment are often used in current discussions
of relations between mind and world. Inverted
Earth arguments compare two situations
that duplicate the internal physical states of an organism but in
which states of the external world are different. For example, everything
red on Earth may be green on Inverted Earth and vice versa. But
red objects produce the same internal physical states in the Earthling
that green objects produce in his Twin. The question is whether
the mental contents of the Earthling and his Twin must also be duplicated
in these two situations. If content is narrow or autonomous, it
must be. If content is world-involving or context-dependent, it
may not be. The answer may depend on whether we are talking about
qualitative or representational content. (Essays 7)
Inverted
Qualia arguments compare two situations
that duplicate an organism's relations to states of the external
world in relevant respects but in which the internal physical states
of the organisms are different. For example, red objects may produce
different internal physical states in two people even though the
two types of state play exactly the same functional role for them
in their environments. Despite the difference in their internal
physical states, both people may call just red things 'red', and
so on. The question is whether their mental contents are fixed by
these duplicated environmental relations and functions, or could
vary with internal physical states. Could these two people mean
different things by 'red'? Could they experience red things differently:
say, could one see green where the other sees red? A traditional
view is that such inverted qualia do indeed make sense. (Essay 7)
Such
thought experiments provide a framework for much current debate
about relations between mind and world. Both types of thought experiment
can be (but aren't usually) applied to questions about the content
of intentional actions, as well as the content of perceptual experience.
But note that this framework involves certain assumptions, even
before the question is resolved about whether content is world-involving
or autonomous. One type of thought experiment supposes internal
physical states can in principle be duplicated though relations
to the environment are not; the other supposes environmental relations
can in principle be duplicated though internal physical states are
not. Note also that scientists are often uncomfortable or impatient
with these thought experiments. Philosophers tend to dismiss this
reaction as failing to understand the conceptual point of the thought
experiments. But the unease may itself have a deeper conceptual
point.
This
framework begins to unravel when we consider cases involving spatial
perceptions and intentions. Instead of inverted red and green, for
example, we could invert or distort spatial features. We now find
systematic interdependencies between perception and action, and
dynamic feedback from output back to input, which are characteristic
of what it is to have a perspective. These create difficulties of
principle for the Duplication Assumption made in Inverted Earth
thought experiments. For example, it is hard to make sense of an
active, perspective-bearing creature whose central nervous system
could be dynamically duplicated in certain spatially distorted environments.
The difficulties are not merely technical, curable by more philosophical
imagination or a bit of science fiction. The interdependence of
spatial perceptions and intentions for an active creature sets limits
in principle to the creation of virtual reality. And because this
point is a consequence of something basic about having a perspective,
its implications are not confined to spatial cases.
It
is no accident that the same dynamic feedback factors that make
trouble for the Duplication Assumption in spatial Inverted Earth
cases also conflict with the Input-Output Picture. The picture of
perception as input from world to mind and action as output from
mind to world reinforces the assumption that at least the vehicles
of mental contents must be internal and bounded: what input is to
and output is from. We need a different way of challenging the view
that the mind is in principle independent of the world, one not
wedded to this picture. (Essay 8)
Neuropsychology
can help to undermine the Input-Output Picture as a general conceptual
framework for thinking about perception and action. For this purpose
we need to turn ourselves through ninety degrees in thinking about
content: to focus on relations between input and output rather than
on relations between internal and external states. In effect, we
shift from scrutiny of the Cartesian
mind/world cut to scrutiny of the
Humean perception/action cut. There is much scientific interest these days in
the idea of active perception, but the philosophical import of this
idea is often vague or elusive. The ninety-degree
shift provides one way of making the
idea of active perception operational for philosophical purposes.
It provides an explicit framework for understanding the philosophical
relevance of empirical work and related thought experiments. We
can understand this new framework by comparison with the standard
framework.
First,
recall that in Inverted Earth cases we suppose internal physical
states are duplicated while external relations vary, and we then
consider whether content is fixed by internal physical states. When
we make the ninety-degree shift, we suppose instead that inputs
are duplicated while outputs vary, and then consider whether perceptual
content is fixed by inputs or can vary with relations of fixed input
to output. That is, can perceptual distinctions depend noninstrumentally
on output?
Note
that the challenge to the input-output picture requires a distinction
between instrumental and noninstrumental dependence. For example,
it is obvious that perceptual content can vary with output that
in turn has an effect on input: If I move around the corner I will
see something I cannot see from where I am. This kind of dependence
of perceptual content on output is merely instrumental. It operates via changes in input; changes in output
are a means to changes in input. This is not surprising and does
not challenge the input-output picture. By contrast, noninstrumental dependence
of perceptual content on output does not operate via input, but
directly. For example, perceptual distinctions may depend on outputs
even though inputs are fixed. A simple illustration of how this
might happen is the alleged paralyzed eye phenomenon. Suppose my
eye muscles are paralyzed so that my eye cannot move. I try to move
my eyes sideways but fail, so input to my retina does not change.
And no other inputs change. Nevertheless, I may have a visual experience
as of the world shifting sideways.
Second,
recall that in Inverted Qualia cases, we suppose internal physical
states vary while external relations and functions are fixed, and
we then consider whether content is fixed by external relations
and functions. We can again shift to consider the orthogonal situation:
we can suppose inputs vary while relations to output are fixed,
and then ask whether perceptual invariances can depend on relations
to output when input varies.
Reciprocal
moves can be made concerning distinctions and invariances in the
content of basic intentions. In effect, we consider how the Input-Output
Picture may fail for distinctions and for invariances in the contents
of perceptual experiences and of intentions.
This
four-box approach can be used to analyze and interpret various empirical
(and related hypothetical) cases. These concern unilateral neglect
(neglect of the left half of the world, including their own bodies,
by otherwise normal people), adaptation to left-right reversing
and color-distorting spectacles, the use of tactile stimulation
to substitute for visual stimulation in blind patients, the acquisition
of control over bodily and brain processes through biofeedback,
the reacquisition of bodily control after deafferentation, and so
on. Intuitively, these cases have something in common; they are
surprising in a similar way.
A
diagnosis of this intuition is that they show, in various ways,
how the Input-Output Picture might fail to hold. We tend to assume
this picture as a general conceptual framework: witness the common
phrase "perceptual input, behavioral output". But these
cases suggest we should not, which is why they are surprising.
Some cases illustrate how perceptual distinctions and invariants
might depend noninstrumentally on output. Others show how distinctions
and invariants in the contents of basic intentions might depend
noninstrumentally on input. While some of the cases can be interpreted
to fit the Input-Output Picture, it is an empirical question whether
this account applies in particular cases. (Essay 9)
A
different general conceptual framework is needed, which admits the
Input-Output Picture as an empirical account in particular cases.
Generically, we need a Two-Level Interdependence View; the input-output
view can then been seen as a limiting case of nil interdependence.
This generic view can be filled out by appeal to the idea of a complex
dynamic feedback system.
Consider
different ways of departing from the dominant tradition in studies
of perception and action. The dominant tradition has an old-fashioned,
sexist view of the marriage of perception and action. They are separate but unequal:
perception is primary, the husband; action is derivative, the wife.
We should distinguish two aspects of this tradition. The first is
a primarily linear
or one-way
view of the causal flow: in from the world through sensory systems
to perception to cognition to motor systems to action and finally
out to the world again. The second is a view of perception and action
as merely instrumentally related: perception is seen as a means to action and action
a means to perception.
We
can depart from the tradition in either or both of these respects.
But rejecting either aspect of the dominant tradition while retaining
the other aspect is unsatisfactory.
'Output-side'
views of perception such as behaviorism depart from the instrumental but not the one-way
aspect of the traditional view. Behaviorism gives no special role
to feedback or the dynamically loopy character of causal flows:
it casts actions as output. But it does view action as constitutively
related to perception. This combination leaves behaviorism open
to the objection that it is verificationist: that it collapses the distinction between perceptual
experiences and their effects in action, effects that are merely
evidence for perceptual experience, not constitutive of it.
Ecological
or Gibsonian views of perception depart from the one-way view, but
not the instrumental view. They emphasise the dynamic loopiness
of the causal flows that make for perception: the importance of
sensory feedback from movement. Nevertheless, Gibson insisted that
passive movement would do as well as active: the role of feedback
was merely instrumental, in that action merely provided a means
to higher-order patterns of input. Perception and action are interdependent,
on his view, but merely instrumentally interdependent.
By
contrast, consider a pair of views that can make both moves away
from the traditional view. First, they emphasize dynamic feedback.
Second, they show how perception and action might in some cases
be constitutively interdependent, not merely instrumentally. These
two views are the motor theory of perception and the control systems
theory of action as the control of
perception. These theories fit together, in that they both appeal
to dynamic systems in which feedback has a complex role: it may in principle operate both instrumentally
and constitutively, and both externally and internally. The combination
of these theories yields the view that perception and action both
depend on relations between input and output: a version of the Two-Level
Interdependence View. Perceptual distinctions and invariants and
basic intentional distinctions and invariants emerge together from
the complex dynamic system and constrain one another. To describe
such a system as perceiving is to describe it as functioning to
represent; to describe it as acting is to describe it as functioning
to control. But these are different ways of describing the same
complex system. And such a system's functions of representation
and control do not in general map tidily onto a distinction between
input and output.
This
is an improvement on both of the one-sided departures from the separate-but-unequal
tradition. It shows how, despite their rivalry, ecological and motor
theories of perception can benefit from one another's insights.
And it avoids standard objections to behaviorism: perception and
action can be constitutively related without threatening the holism
of the mental. (Essay 10)
4. Some subplots.
Some themes are repeated with
variations and cross-referencing in different essays, but are not
treated comprehensively in one place. Other themes are present only
implicitly. A few of these subplots are sketched here, to help see
their relationship to the main plot. The sketches below use broad
strokes and don't provide fully argued positions. Rather, they may
consolidate points scattered about the essays or fill in suggestions
or implicit context.
5. Divide and conquer?
We are not trying to produce
a grand unified theory of consciousness. Indeed, there is room for
skepticism about whether there is a grand unified problem of consciousness.
Our approach is rather to break consciousness down into aspects
and address those. Perhaps a view of consciousness as a whole can
be built up by considering the relationships between these various
aspects of it. There are lots of different issues about consciousness.
Here are four broad categories of issues.
A. What does the presence or absence
of consciousness consist in? What is it for states to be conscious
or not? Under this heading philosophers wonder: Could there
be zombies, who are just like people except that their 'mental
states' are not conscious? If a system has the right structure
of functional relations, is it thereby conscious, or does it
matter what materials the system is made of? What work does
consciousness do? What evolutionary function might it have?
Is it dispensable in science?
B. What is it for conscious
states to be like they are, of the phenomenal
types they are? What is it to
be whatever it is like to be a bat, or for pain to be like pain,
or for something to look like red things look rather than like
green things look? How is phenomenal type related to representational
type?
C. What is it for consciousness to be
unified
at a time?
D. What is self-consciousness? Does consciousness require self-consciousness?
Are there varieties of self-consciousness? Does self-consciousness
require conceptual abilities?
Instead of hammering at the
front door of the difficult issues about presence or absence, maybe
we can slip round the back. Maybe if we begin with issues about
unity and self-consciousness, we will end up learning something
about phenomenal types and even the presence or absence of consciousness.
But we should not assume that the problems of consciousness will
fall like a row of dominoes. We should tease issues apart and create
a network of small insights. Perhaps together they may impose some
overall order on the subject matter. But the correct view may be
eclectic and messy. It may have components drawn from traditionally
opposed camps, such as normative and functional components. Different
aspects of consciousness may require different theoretical tools.
We can't rule out in advance the possibility that consciousness
organizes itself out of disparate and ill-assorted materials, and
that there is no independently discernable overall structure from
which its various attributes can be derived.
6. Parallels, myths, and a new angle
Perception and action can
be regarded as parallel and isomorphic though separate systems,
with different directions of 'fit' to the world. The parallels between
them provide ways of elaborating the view of perception and action
as buffer zones around the self, which interfaces between input
and output. But as we follow perception and action further inward,
the self seems always to recede.
It
can be illuminating to pursue the parallels. Consider the parallel
inward retreats of knowledge and of responsibility. Philosophers
are often wary of the Cartesian supposition that we only know for
certain the contents of our own experiences. They are less often
wary of the Kantian supposition that we are only truly responsible
for the contents of our own intentions. Yet the effect of luck in
contracting the scope of responsibility can be seen as parallel
to the effect of the skeptic's hypotheses in contracting the scope
of knowledge. (Essay 7)
Following
the parallels through can also help to see what is wrong with the
Input-Output Picture. By taking a new angle on perception and action
we allow the self to reappear in the world, where it seemed to be
to start with.
Consider
the "myth of the given": the conception of perceptual experience as
the given, as reflecting pure input from world to mind with no active
contribution from the perceiver. Both Kant and Wittgenstein reject
this idea. Both emphasize, though in different ways, the contribution
of our activity to what we experience. Kant appeals to the spontaneity
of synthesis in explaining the unity of consciousness and conceptual
unity ('combination does not ... lie in the objects'). Wittgenstein
appeals to the role of our practices in determining meaning and
content ('this
is what we do'). Their rejections of the given have profoundly influenced
the modern doctrine that the way content classifies the world is
'up to us', not exogenously fixed but a result of our conceptualizing
activity.
But
let's now follow the parallels through. Consider such conceptualizing
activity. Is it intentional agency or not?
If
it is, what should be said about the contents of conceptualizing
intentions? How are they determined? A regress lurks. We should
not suppose appealing to agency gets beneath the issues about content
and unity, as if agency were a matter of pure output from mind to
world. We can call this the myth of
the giving. It is just as much a mistake
as the myth of given, the idea of perceptual experience as pure
input. The issues about content and unity arise equally for the
contents of intentions. Agency may depend on perception as deeply
as perception depends on agency.
On
the other hand, if intentional agency is not at issue, in what sense
are classification and conceptualization up to us? Only in the sense
that they result from events and processes that occur in our bodies?
They also result from events and processes that occur in the world
outside our bodies.
Up-to-us
doctrines run a danger: the danger of recoiling from a conception
of experience as "the given" into an equally mistaken
conception of agency. Both myths should be avoided. The revolution
that began with Kant's arguments about perceptual experience should
be carried through to agency. Action is no more pure output than
perception is pure input. The whole of the Input-Output Picture
should be rejected, not just half of it. (Essays 2, 6)
We
get a new angle by making the ninety-degree shift: by making the
focus of our scrutiny the perception/action cut rather than the
mind/world cut. The constitutive interdependence of perception and
action when both depend on complex dynamic feedback systems has
several consequences. At the subpersonal level of causal flows,
the dynamic singularity matters but the significance of boundaries
like the skin or the skull fades. At the personal level of mental
content, however, the inward retreat of the self is preempted. Perception
and action are not buffer zones around the mind. The self is out
in the open, where it seems to be. And both myths are avoided. Perception
and action are mutually and symmetrically dependent. (Essays 8,
9, 10)
This
new angle also brings the philosophical relevance of results in
neuropsychology into relief. Various empirical cases suggest the
constitutive interdependence of perception and action. But we observe
something striking. The action in question in these cases is not
spontaneous conceptualizing or pure content-establishing activity:
not the fancy kind of action that runs the myth of the giving danger.
It is simple, real-world, motor action, contentful in a way that
is in turn dependent on perceptual capacities. It is this kind of
action that is the essential partner of perception. (Essays 8, 9)
It is a mistake to get sophisticated about
experience while remaining naive about action. Someone who wouldn't
be caught dead thinking of the content of perceptual experience
as autonomous rather than world-involving might well still think
of the content of intentions in this way. That makes it easy to
make another mistake, which encourages idealism and relativism:
to think the kind of activity experience depends on is the fancy,
mythical, content-establishing kind. The interdependence of perception
and real-world action offers no such encouragement.
Conceptualizing
is not something we must do in order to experience or think or do
anything else, but is a fairly sophisticated activity. What happens
in our brains to make perception, action, and conceptualization
possible is not necessarily something we do. To the extent the conceptualization
of information in experience is something we do, as opposed to something
that happens partly in us, we do it by acting in the world, often
by linguistic action, but not by acting in a ghostly precontentful
kind of way somewhere in our brains. For example, I may point to
some architectural details and ornaments in a baroque church and
say to a friend: "You can see those lines as Art Nouveau".
7. Perspective and access: self-consciousness,
conceptual abilities, agency, and life
Does consciousness require
self-consciousness? Self-consciousness is usually thought of as
conceptual in character. But conceptual abilities aren't necessary
for consciousness, for example, in nonhuman animals. Perhaps consciousness
may involve forms of self-consciousness that do not require conceptual
abilities. (Essay 4)
Consider
perspectival self-consciousness. Having a perspective means in part that what you
experience and perceive depends systematically on what you do, as
well as vice versa, and that you can keep track of some of the ways
in which this is so, even if not in conceptual terms. In this sense
having a perspective involves self-consciousness. A conscious animal,
moving through its environment, has the ability to keep track of
relationships between what it perceives and what it does. This ability
enables it to use information about itself and its own states and
activities as well as information about its environment to meet
its needs. It doesn't follow that it has the ability to reason systematically
about aspects of itself, others, and the environment in a variety
of ways detached from its needs. Its abilities need not have the
generality, richly normative character, and systematic decompositional
and recombinant structure of conceptual
abilities. Perspectival self-consciousness
may but need not involve conceptual abilities.
Consider
also self-consciousness in the sense of access to the contents of conscious states. For people with
conceptual abilities, it is arguable that consciousness requires
self-consciousness in the sense of self-evidence: cognitive access to the contents of conscious states.
But even for creatures without conceptual abilities, consciousness
may require self-consciousness in a weaker sense of access, intentional access
to the contents of conscious states. Intentional access is closely
related to the ability to use information explicitly, an ability
that is missing in various examples of covert processing, such as
blindsight. Intentional access to content requires the ability to
form an intention whose content is provided by certain information
and to act on it just for the reason that information provides.
If
a creature without conceptual abilities can have intentions and
act for reasons, it can have intentional access to contents and
perspectival self-consciousness. Intentional
agency occupies a kind of normative
middle ground. It is not mere response to stimulus. It requires
at least rudimentary practical rationality, holistic relationships
between perceptions and intentions, and weak normative constraints
on contents. But the richly structured constraints and inferential
norms associated with conceptual abilities go well beyond this.
What
light do the notions of perspective and access shed on the relations
among conceptual abilities, intentional agency, life, consciousness,
and self-consciousness? Life appears to be doubly dissociable from
intentional agency understood in normative terms, as above. Though
both are alive, animals may be intentional agents while plants are
not. Robots might be intentional agents, even if they were not alive
and their contentful states were not conscious. Both perspective
and access to content involve agency, but not necessarily conceptual
abilities. For creatures with conscious states, both perspective
and access to content may count as forms of self-consciousness.
But even if both are necessary for consciousness, neither perspective
nor access to content seem to be sufficient for consciousness. It
seems that a robot could have both, yet be a "zombie"
without conscious states. Could adding in conceptual abilities keep
such zombie worries at bay? Or, could it be sufficient for consciousness
that a living
thing have both perspective and access to content?
8. Vehicles, contents, and boundaries
We should distinguish properties
represented in content from properties of vehicles of content: the
subpersonal states or processes that carry content. In particular,
we should distinguish the personal and subpersonal levels of conceptualization.
The contents of the mental states of subjects/agents are at the
personal level. Vehicles of content are causally explanatory subpersonal
events or processes or states. We shouldn't suppose that the properties
or relations of vehicles must be projected into the personal-level
contents they carry, or vice versa. This would be to confuse the
personal and subpersonal levels. Nevertheless, there may be something
to be said about the relations between the processes that determine
mental contents and the processes that carry mental content.
These
distinctions are familiar in various applications. Several further
applications are made in the essays.
First,
to the unity of consciousness. We shouldn't suppose the structure
of content, in particular its unity, is isomorphic with the structure
of vehicles of content. A unified consciousness need not depend
on neuroanatomical unity, and neuroanatomical unity doesn't rule
out splits in consciousness. Yet the concept of a subpersonal dynamic
singularity may contribute to an account of the unity of consciousness.
Second,
to the relations between perception and action. The Input-Output
Picture confuses the personal-level distinction between perception
and action with the subpersonal-level distinction between input
and output. This is kin to vehicle/content confusions. Yet the interdependence
of the contents of perceptions and intentions may reflect their
superposition on the same complex network of dynamic relations.
(Essays 1, 5, 8, 9)
A
third application of the vehicle/content distinction is also relevant,
though less of a focus in these essays. But we can use it to illustrate
points made in various essays and to suggest how they might be related
and further developed. The cognitive abilities of persons display
generality and systematicity; content at the personal level has
decompositional, recombinant conceptual structure. What explains
this structure at the level of subpersonal processing? In particular,
must subpersonal vehicles of conceptual content themselves have
a language-like or syntactic structure, so that structure in personal-level
content is explained by isomorphic subpersonal structure?
Their
answers to this last question distinguish classical, 'language of thought' approaches to cognitive architecture
from connectionist approaches. Classicists answer yes, on either a priori or empirical grounds, while connectionists
allow that it may be possible to explain conceptually structured
cognitive abilities in terms of neural networks without syntactic
structure. The assumption that the processes that support true thought
must have a classical architecture (even if they are implemented
by a connectionist network) imposes a isomorphism
requirement. (Essay 10)
Might
our conception of ourselves as genuine thinkers impose this constraint
on relations between levels? If the conceptual structure of content
were not explained by isomorphic subpersonal structure, would displays
of 'cognitive ability' be mere patterns of behavior, mere mimicry
of true thought? This is another category of zombie worry, this
time about thought rather than consciousness. So some argue that
if connectionism does give the right account of supposed cognitive
abilities, then true thought is eliminated.
Much
here turns on two questions. First, should the isomorphism requirement
be applied, if at all, to what explains particular thoughts and particular displays of cognitive ability,
or to what explains why any thought content or cognitive ability
of a given type
exists at all? Vehicles of content explain tokens: particular thoughts
on particular occasions. They are token-explanatory processes. But if the isomorphism requirement holds
at all, does it apply to token-explanatory processes or to type-explanatory
processes, which explain why a certain type of content exists at
all? (Essay 8)
Second,
must whatever explains conceptually structured cognitive abilities
for purposes of the isomorphism requirement be internal? Or can it be world-involving? Can the requirement
of isomorphism be satisfied relationally, by processes or states
that cross the boundary between organism and environment? If it
can be, does it matter whether it is satisfied all over again internally?
This
second question takes two forms, as applied to type-explanatory
and to token-explanatory processes. The answer to the first is obvious.
Type-explanatory processes can be world-involving: consider upbringing,
linguistic practices, evolution, and so on. These are often appealed
to in accounts of what determines content. The answer to the second
is less obvious. Could token-explanatory vehicles of content also
be world-involving, relational states of persons? Could processes
that carry content, as opposed to those that determine content,
go external or relational? Arguments against (as well as for) Internalism
about what determines content typically presuppose Internalism about
what carries content, or vehicles. But is this presupposition of
Vehicle Internalism justified?
There
is nothing spooky about relational properties of persons: they correspond
to intrinsic properties of something bigger where causality reigns
as usual. Context-dependence is relative to a boundary. We are used
to the idea of relations internal to a nervous system as vehicles:
relations between events in a distributed process within the brain
need not be monitored by a single cell or local module. Such internal
relations themselves may be the vehicles of content, rather than
the intrinsic properties of some internal monitor of these relations.
But if internal relations can qualify as vehicles, why not external
relations? Given a continuous complex dynamic system of reciprocal
causal relations between organism and environment, what stops the
spread? The idea that vehicles might go external takes the notion
of distributed processing to its logical extreme. (Essay 8)
The
isomorphism requirement is usually applied to internal, token-explanatory
vehicles of content. But might the intuition that genuine thought
requires an isomorphism between the level of content and the level
of causal explanation instead be captured by appealing to external,
type-explanatory processes? Perhaps this intuition can be satisfied
by a whole system, including embedding linguistic environment, even
if it is not satisfied by internal vehicles. (Essay 10)
9. The classical sandwich vs. horizontal
modularity
A view of perception and action
as separate input and output systems compliments a view of thought
and cognition as 'central' and in turn separate from the 'peripheral'
input and output systems. The virtual processing of cognition is
seen as central, even if its implementation is distributed; input
to it is provided by perception, and it issues output that generates
action. The subpersonal underpinnings of the mind are conceived
as vertically modular, with cognition interfacing between perception and
action. Against this background, a standard view is that connectionist
approaches are at their strongest for the peripheral processes (e.g.
sensorimotor control, pattern recognition), but that the central
cognitive interface must have classical structure.
This
classical sandwich conception can be attacked from the center, by arguing
against the view that true thought requires isomorphic structure
among the internal vehicles of thought. But such attacks often accept
the conception of perception and action as peripheral buffer zones,
which holds the central interface conception of cognition in place.
Another way to undermine the classical sandwich is to attack from
the outside, as it were: to challenge the buffer zone view and the
separation of perception and action from central cognitive processes.
The
two angles of attack can be complementary and mutually supporting.
Traditional presuppositions about the separateness of perception
and action may hold other philosophical assumptions in place. To
illustrate this suggestion very briefly, consider first the assumption
that the mind decomposes vertically so that cognitive processes
are central and distinct, and second the assumption of a dichotomy
between internal classical structure and behaviorism.
First,
we've seen that the Input-Output Picture runs together the personal-level
distinction between perception and action and the subpersonal-level
distinction between input and output. By contrast, a Two-Level Interdependence
View sees perception and action is as interdependent because co-dependent
on a complex dynamic system of causal relations, which may extend
into the environment. On such a view, instead of seeing the mind
as vertically modular we can see it as horizontally
modular. Each horizontal module or
layer is a content-specific system that loops dynamically through
internal sensory and motor processes well as through the environment.
This
change may help to understand how the conceptual structure of cognition
might emerge from dynamic sensorimotor systems and the role environmental
structure might have in such systems, even in the absence of internal
isomorphic structure. Complex dynamic systems can have very striking
emergent properties. Indeed, emergence in complex dynamic systems
could be argued to underlie and explain the vehicle-content distinction.
An essential lesson we learn from studying such systems is that
discontinuities and structure at a higher level can emerge unpredictably
(though deterministically), with no isomorphism at the underlying
level.
Such
emergent structure could register in evolutionary and cultural processes.
A horizontal module might couple perception and action together
in specific ways that are functional or adaptive in certain conditions:
consider imitation. However, considered in isolation such a specific
coupling might be too inflexible to respect the holism of the mental
and the constraints of rationality. But just evolution and culture
might select specific layers, they might also twist the layers together
into still more complex systems that display further, qualitatively
novel structure. Rational flexibility and cognitive structure might
emerge from intricate higher-order relations of inhibition and facilitation
between horizontal layers. But damage to these higher-order relations
might reveal the underlying horizontal structure, in the form of
various neuropsychological pathologies and dissociations. (Essays
5, 10)
Second,
recall the zombie worry about 'cognitive abilities' without explanatory,
classically structured internal vehicles. This worry can be encouraged
by a spurious dichotomy: either the right explanatory internal structure,
or mere mimicry, mere patterns of behavior. The Input-Output Picture
supports this dichotomy by type-casting behavior, including linguistic
behavior, in the role of effect and ignoring its other talents.
As the mere effect of thought, behavior is merely evidence for thought.
So worries about behaviorism and verificationism get started.
The
change from the Input-Output Picture to a Two-Level Interdependence
View can underscore the obvious but neglected point that to participate
in natural language is to perceive and to act, and depends on sensory-motor
interactions with linguistic environments. Behavior, including linguistic
behavior, is as much cause as effect within a complex dynamic feedback
system; it is not merely evidence for something else that's doing
the work. (Essay 10)
10. Affinities and implications
Decentralization, self-organizing
systems, context-dependence, feedback, emergence: these are themes
of some fascinating current work in cognitive science and related
disciplines. The sketched subplots have clear affinities with developments
in connectionism, dynamic systems theory, artificial life, as well
as evident antecedents in cybernetics. These essays can be seen
as a philosophical complement, and bridge, to this work in other
disciplines. But this book develops these shared themes relatively
independently, beginning (at least) from mainstream philosophical
concerns, and it puts primary emphasis on their relevance to consciousness
rather than to cognition. By tracing continuities that may not be
apparent on the surface, I hope this book may help to bring mainstream
analytical philosophy into fruitful contact with this work in other
disciplines.
The
Input-Output Picture that is here criticized has had significant
implicit influence in ethics. Large questions for further thought
are these: What are the implications for ethics, and for social
and political philosophy, if this view can not be justified? More
generally, to what extent should these normative subjects be studied
in ways that are relatively insulated from our advancing understanding,
both empirical and theoretical, of the mind?
top | back
to symposium index | |