Could
There Be A Science of Consciousness?
I Introduction
In
this paper I want to consider the implications of materialism about the human
mind for a scientific understanding of consciousness. I shall argue that, while science can tell us many exciting
things about human consciousness, it won’t be able to pinpoint any specific
material property that constitutes seeing something red, say, or being
in pain, or indeed that constitutes being conscious (that is,
feeling like something rather than nothing).
Not that this means there are definite facts about consciousness about
which science must permanently remain silent.
Rather the difficulty lies with our concepts of conscious properties,
which are vague in certain crucial respects.
II Background Assumptions
Let
me begin with four assumptions which will provide the framework for my
argument, but which it will not be possible to defend properly here. (For detailed arguments, I refer readers to
Papineau, 2002.)
(1) Materialism. Conscious properties are material
properties.
Comments:
By
‘material’ properties I mean either strictly physical properties or properties
which metaphysically supervene on strictly physical properties.
Materialism
is strongly supported by the causal completeness of physics, which leaves no
room for anything which is metaphysically independent of the physical realm to
have any physical effects. If conscious
properties were non-material, they would thus be epiphenomenal ‘danglers’,
caused by physical occurrences, but themselves having no effects on physical
activities. In this respect, they would
be quite unique: nature offers no other
examples of such impotent epiphenomenal properties.
Of
course, if there were compelling independent grounds for holding that conscious
properties are non-material, then we would have no option but to accept
epiphenomenalism about consciousness.
However, I do not think that the standard anti-materialist arguments
(Kripke, 1980, Jackson, 1986) provide any such compelling grounds. So I see no reason to reject the relative
simplicity of materialism for the metaphysical oddity of epiphenomenalist
dualism.
(2) Conceptual Dualism. Nevertheless we have two quite different
kinds of concepts for thinking about conscious/material properties, which I
will call ‘phenomenal’ and ‘material’ concepts.
Comments:
‘Phenomenal’
concepts are most easily thought of as akin to demonstratives, or even
quotational terms: we use a mental
construction which can be represented as ‘that experience: ---’; we fill the gap with some current
experience or its recreation in imagination;
and we thereby form a term which refers to that experience itself
(either the type or the token). Compare
the way that the term ‘cat’ refers to the word enclosed in the quotation marks
(either the type or the token).
I
shall use ‘material concepts’ as a catch-all for all other, non-phenomenal ways
of thinking about conscious/material properties, including concepts of brain
states, concepts of causal roles, and so on.
There
are no a priori connections between phenomenal and material concepts. It is always informative to learn that that
some given experience: --- (and here we ‘quote’ some experience) is
constituted by some physical property, or plays a certain causal role, or has
some other such material feature.
I
take there to be a phenomenal concept of the determinable property conscious
(being like something rather than nothing), as well of determinates like seeing
something red or being in pain.
This is why it always makes a priori sense to ask, given any material
specification of some property, whether creatures with that property are
phenomenally conscious. (For more on
the structure of this determinable phenomenal concept of consciousness-as-such,
see Papineau, 2002, pp.185-6.)
(3) Hard Problem Research. Consciousness research aims to identify the
material referents of phenomenal concepts.
Comments:
David
Chalmers (1996) distinguishes the ‘hard problem’ of figuring out what
phenomenal concepts refer to from the ‘easy problem’ of showing how
psychological concepts of causal roles are physiologically realized. While I of course disagree with Chalmers’
eventual conclusion that phenomenal concepts refer to non-material properties,
I agree that it is the hard problem of identifying the referents of phenomenal
concepts that makes consciousness research seem so exciting.
Thus
there would be no special challenge about understanding colour vision if all we
wanted to know were which physiological mechanisms realize the psychological
concept of responding differentially to coloured objects. By contrast, the kind of consciousness
research I shall be concerned with raises such questions as which property
constitutes seeing something red—and here we mentally ‘quote’ the
experience.
(4) Subjects’ Reports. Consciousness research hinges crucially on
subjects’ phenomenal reports about present and past experiences.
Comments:
I
take subjects’ first-personal phenomenal reports to provide a kind of
observational data-base for research into the material referents of phenomenal
concepts. Compare the way in which
research into the material nature of water, say, must begin with
observationally identified samples of water.
I
shall say more about the status of subjects’ reports later. But we can bring out the importance of such
reports by considering research into creatures who cannot make them, like
vervet monkeys, say. You can stimulate
vervet monkeys, you can get them to perform various tasks, and you can check on
what is going on in their brains. But
since the monkeys can’t you tell what they consciously experience at different
times, none of this will cast any immediate light on the monkeys’ phenomenal
consciousness. For all this research
establishes, the monkeys might share very similar experiences to our own, or
have no consciousness at all, or be somewhere in between. By contrast, humans can tell us when they
are seeing something red, feeling a pain, or experiencing something rather than
nothing, which offers a crucial handle by which to identify the material
natures of phenomenal properties.
III Empirical Research into Consciousness
Given
the points just made about subjects’ reports, it may seem that research into
the material referents of phenomenal concepts presents a standard scientific
problem. Suppose that we have a
data-base of cases for some phenomenal property Ø, derived from subjects’
reports. We can then seek some material
property M which is present whenever subjects report Ø, and is absent whenever
subjects deny Ø. This strategy thus
promises to identify M as the material nature of Ø, as desired.
I shall argue, however, that
such research will always be less than fully effective, and will inevitably
fail to pinpoint a unique material referent for any phenomenal concept. The is because there will always be a
numbers of different material properties M which fit the evidence of the data
base equally well, and between which the standard research methodology is
therefore impotent to decide.
Of course, scientific research
can do a great deal to narrow down the possible material
referents of phenomenal concepts. If
some material M is to be identical with phenomenal Ø, then it cannot be absent
when Ø is testified to be present, otherwise it won’t be necessary for Ø. Nor can it be present when Ø is testified to
be absent, otherwise it won’t be sufficient for Ø. Scientific research can show that many candidate Ms violate one
or the other of these requirements.
Take the phenomenal property of being in pain, say. Any material property that is sometimes
absent when subjects report being in pain—some very specific neuronal activity
that is peculiar to stabbing pains, for example—is thereby disqualified as
unnecessary for pain as such. And any material
property that is sometimes present when subjects say they are not in
pain—some general neuronal activity common to both pain and heat sensations,
say—is thereby disqualified as insufficient for pain.
Research of this kind can tell
us many interesting and indeed surprising things about consciousness. For example, recent investigations have
shown that subjects report no conscious experience of many cognitive processes
which we might pre-theoretically have expected to be conscious: these include the visual guidance of hand
movements (Weiskrantz, 1986, Goodale and Milner, 1992) and the instigation of
voluntary actions (Libet, 1993). There
are also cases of the converse kind, where subjects report conscious experiences
in cases where we might have expected none:
for example, subjects who take morphine to allay an existing pain will
report that they can still consciously feel the pain, even after it has ceased
to agitate them (Dennett, 1978).
Yet science can only take us so
far. It may be able to rule out certain
candidates as the referents of phenomenal concepts, often surprisingly so. But the trouble is that, even after science
has narrowed things down in this way, there will always remain a plurality of
Ms in play for any given phenomenal Ø, and no further means of deciding which
is the real referent of the phenomenal concept.
My eventual diagnosis of this situation will not be
that there are definite facts about consciousness to which we lack
epistemological access—that there is some material property that really
constitutes being in pain, say, but which we can’t find out about. Rather, I shall argue that our phenomenal
concepts of conscious states are vague—nothing in the semantic
constitution of phenomenal concepts determines precisely which of the candidate
material properties they refer to.
IV Too Many Candidates
To
illustrate the basic difficulty, suppose that we have identified some strictly
physical property P that is present whenever humans testify to being in pain,
and absent whenever they deny this. If
this is so, then will also surely be a number of more structural properties
which are similarly perfectly correlated with human reports of being in
pain. Simply focus on some detailed
level of causal organization ensured by P, but abstract away from the specific
physical way that this organization is realized in human beings. The resulting structural property S will
also be present whenever humans report being in pain, and absent whenever they
deny it. So there will be no way of
using standard scientific research into humans to decide whether P or S
constitutes being in pain.
To make the point graphic,
consider a ‘silicon doppelganger’ who shares some detailed structural property
S which co-varies with pain in humans, but whose basic chemistry is
silicon-based rather than carbon-based.
If you prick this being with a pin, you will produce S, but not the
strictly physical P which also goes with pain in humans. So standard consciousness research leaves us
with an obvious but unanswered question.
When you prick the doppelganger, does it feel pain? If pain = S, then it does. If pain = P, then it doesn’t. And if standard consciousness research on
humans can’t decide between S and P as the referent of being in pain,
then it can’t tell us what the doppelganger feels.
Note that there won’t only be
one alternative structural S here, but a whole sheaf of them. We can characterize humans physiologically,
neuronally, and computationally, as well as strictly physically (cf. Lycan, 1987),
and we are likely to find Ss at all these further levels that correlate
perfectly in humans with any strictly physical P that goes with pain in
humans. Empirical research will
therefore be unable to decide between the strictly physical P and any of these
candidate Ss as the referent of any given phenomenal concept.
It gets worse. As I shall show
in section VII below, there is also a plurality of material candidates in the
‘broad-narrow’ dimension, as well as in the physical-structural dimension. Still, let us keep things simple for the
moment, and stick with a simple pair of candidates, a strictly physical P found
in humans, and some structural S which is common both to humans and their
silicon doppelgangers.
IV More Data
It
may seem as if the problem is just one of insufficiently varied data. After all, research into the nature of some
kind always needs a heterogeneous data base.
If we lack sufficiently varied data points, we are in danger of picking
out some narrower or wider category than the target kind itself. So for example, if all our examples of pain
were stabbing pains, we might be led to posit an overly precise material nature
for pain, by fixing on the material basis of stabbing pains, rather than of
pain as such. Conversely, if none of
our examples of pain absence were heat sensations, we may be led to an overly
permissive material nature for pain, by fixing on some material property common
to both pain and heat sensations. To
get things right, our sample must thus include sufficiently varied examples of
pains and non-pains. (For a careful
discussion of how to maximize such variation in research on humans, see Frith
et al., 1999.)
Isn’t just this our present difficulty? You might think that if only we had appropriately varied examples
of pains and non-pains, then we could decide between P and S as the material
nature of pain easily enough. In normal
humans, P and S always go hand in hand, so normal humans won’t supply the
requisite data points. But if we could
find some beings who have P without S, or vice versa, and could see whether
they still have pains, then that would do the trick.
In fact there is only a possibility of dissociation in one direction
here, since the carbon-based details P will determine the structural property
S. Still, we can well have S without P,
as in the silicon doppelganger. And
this doppelganger itself might then seem to offer the requisite data point, in
principle at least. We simply need to
check whether the silicon doppelganger with S but not P is in pain. If it is in pain, then this will eliminate
the carbon-based P as unnecessary for pain.
And if it isn’t in pain, then that will eliminate the structural S as
insufficient for pain.
But this doesn’t work. There is
a deeper difficulty, which will remain even after we incorporate a silicon
doppelganger into our research. The
trouble is that our canonical method for telling whether or not we have a case
of pain is to rely on subjects’ reports.
In the case at hand, we will thus ask the doppelganger whether or not it
consciously feels pain when it is pricked with a pin. However, we know beforehand what the doppelganger will say,
whether or not it actually feels any pain.
Since it is structurally organized just like us, it will behave
observably just as we do, and in particular will utter the words ‘I am in pain’
when it is pricked with a pin. But this
tells us nothing at all, since we already know it will utter these words both
in the case where it lacks real pain and in the case where it has it.
Clearly, it would beg the question at issue to take the doppelganger’s
utterance at face value. Our question
is whether structural S or physical P constitutes the material nature of pain. If structural S did constitute pain, then
the doppelganger would feel pain, and its words ‘I am in pain’ would report as
much. On the other hand, if physical P
constituted pain, then the doppelganger wouldn’t feel pain, and its words would
have some other significance. So, in
order to interpret the doppelganger’s words, we need to know whether S or P
constitutes pain, which was the question we started off trying to answer.
VI
What’s So Special About Humans?
At this point an obvious question may occur to some
readers. If we can’t take the
doppelganger’s reports at face value, then why isn’t there a similar problem
with the phenomenal reports of other humans?
As I have described the methodology of consciousness research, we are
looking for a property which is present whenever humans say ‘I am consciously
experiencing phenomenal property Ø’ and which is absent whenever they deny
this. This methodology thus takes it
for granted that the phenomenal reports of other humans provide accurate
information about Ø. But why is there
an asymmetry between humans and the
doppelganger here? Shouldn’t we be
suspicious of the reports of other humans too?
My answer is that when we are dealing with other humans, rather than
silicon doppelgangers, we have good independent reason to suppose that they
share our experiences, and thus that they refer to those experiences when they
form phenomenal concepts by ‘quoting’ their own experiences.
This might seem rather quick.
What about the familiar challenge of the inverted spectrum? Isn’t it possible that, when you look at
lush grass, you have the experience that I have when I look at ripe tomatoes?
Well, I concede that this is a priori possible, even
given our similar behavioural responses to colours. But there is also a posteriori reason to think that
such inverted spectra are not actual. I
have in mind here the general empirical arguments in favour of materialism
mentioned at the beginning of this paper:
given the (a posteriori) causal completeness of physics, we have good
empirical reason to reject the metaphysical danglers of dualism in favour of
the relative simplicity of materialism;
so conscious properties must be determined by physical properties; which means we can be sure that any beings
who share relevant physical properties will share conscious properties. Since we also have empirical reason to
suppose that normal humans are alike in relevant physical details when they
look at ripe tomatoes, we can conclude that ripe tomatoes occasion the same
experience in all normal humans.
In support of this line of argument, note how we would not rely on the
phenomenal reports of human beings whom we knew not to be physically
similar in relevant respects. Thus
consider a brain-damaged human who has had its cerebral pain mechanisms
replaced by some structurally equivalent but silicon-based prosthesis. Even though this is unequivocally a human
being, it will count with the silicon doppelganger as far as consciousness
research goes: it is no good asking
this person whether it feels a pain when it is pricked, in the hope of
discovering whether S (which it has) or P (which it lacks) is the material
basis of pain. For its structural
equivalence with normal humans ensures that that it will respond ‘Yes’, whether
or not it feels pain. (Similarly, it is
no good engineering somebody in whom my ‘red’ brain state realizes the detailed
‘green’ structure, and then asking it whether it sees something green when it
looks at grass, in the hope of discovering whether colour experience goes with
structure or realizer. For again this
person will say ‘Yes’, whether or not it sees something green.)
Indeed a similar point would apply if it were to turn out that normal
healthy humans varied in the material underpinnings of their conscious
states. Suppose, just for the sake of
the argument, that we discovered that the pain mechanisms of men and women were
realized by somewhat different neuronal arrangements, even though these
mechanisms were computationally identical, with both sexes agreeing entirely in
their subjective reports about pain.
Then there would be some computational structure that co-varied with
pain in both men and women, while the neuronal realizers of this structure
differed between men and women. In this
situation, women could take the subjective pain reports of other women at face
value, and be confident that other women meant the same as them when they said
‘I am in pain’, on the grounds that women shared all relevant material
properties . But women could not take
the same attitude to male pain reports, since they could not rule out the
possibility that pain experience depended on neuronal underpinning rather than
computational structure, and that men were therefore referring to something
different from female pain with their phenomenal concept of ‘pain’.
So
far I have argued that scientific research will be impotent to decide between
strictly physical and structural properties as the referents of phenomenal
concepts. Let me now show that a
similar problem arises, not only with ‘vertical’ alternatives among narrow
material properties, but equally with ‘horizontal’ alternatives that emerge
once we entertain the possibility that conscious properties may be broad.
The point is best introduced by considering views of
consciousness which are not just materialist but also representationalist—that
is, views which seek to equate phenomenal properties with representational
properties, where representation is somehow explained in material terms. Now, most current materialist accounts of
representation make representation a broad matter: to represent that there is water in your glass, or that an
elephant is approaching, or that your wife is leaving, it is not
enough that your brain be appropriately internally arranged—you also need to
have appropriate causal or historical relationships to water, or elephants, or
your wife. A physical duplicate on a distant planet would not be able to
represent these things.
At first pass, representational accounts of
phenomenal properties thus appear to have the surprising consequence that your phenomenal
states, along with your other representational states, depend on your causal or
historical relations to your environment, and not just on current matters
inside your skin. Some philosophers of
consciousness are prepared to swallow this (Dretske, 1995, Tye, 2002). However, there is also the option of
factoring out a narrow component of representational properties—which depends
only on matters inside the skins of water/elephant/wife representers—and
equating phenomenal properties with these narrow representational properties
instead.
So any phenomenal property Ø will once more present
us with a choice: should Ø be equated
with some broad representational property B, or with its narrow distillation
N? Well, we could decide this empirically
if we had an effective test case. We
need a being with N but not B, or vice versa, and a way of telling whether it
has Ø.
Again there is only a possibility of dissociation in
one direction here, since broad representational properties will include their
narrow components. However, we can
imagine a being who is internally just like you, but lacks your causal or
historical environment (it lives on a distant planet, say, or was constituted
by chance in a swamp). So all we need
to know is whether this being feels Ø when you do. But now we face the same old problem. The way to find out whether it feels Ø is to ask it. But we know beforehand what it will say. Since it is just like you in narrow
respects, it will certainly utter the sounds ‘I am feeling Ø’. Yet it would utter these sounds even if
phenomenal properties were broad, and it did not feel Ø at all. Once more, empirical research cannot decide
between the alternative material referents for phenomenal concepts.
VIII The Issue In General Terms
It
may be helpful at this point to move away from details, and lay out the
dialectical situation in general terms.
I began with abstract considerations resting on the causal completeness
of physics: these implied that any
given conscious property Ø must be identical to some material property or
other. The identification of some
specific material property was then turned over to empirical
investigations. Ideally, we would like
the abstract considerations and the empirical investigations to complement each
other, with the empirical investigations identifying some unique material
property M that is perfectly correlated with phenomenal property Ø in humans,
and the abstract considerations then allowing us to conclude that M and Ø
aren’t just correlated, but identical.
The trouble, however, is that
empirical investigations seem unable to play their part. They can’t identify a unique material
property, since there are too many material properties which are perfectly
correlated in humans with any given phenomenal Ø—structural properties as well
as strictly physical ones, broad as well as narrow. So empirical investigations seem unable to pinpoint the material
referents of phenomenal concepts.
Some readers may feel that the
way forward is to beef up the abstract considerations. Perhaps we are asking too much of the
empirical investigations. Rather we
should first seek to decide, on general abstract grounds, whether we are
looking for a structural or a strictly physical property, or a broad or a narrow
one. Then, once the target has been
narrowed down in this way, there need be no barrier to empirical research
locating unique material referents for phenomenal concepts.
So perhaps we could decide
against structural properties, and in favour of strictly physical ones, on the
grounds that only strictly physical occurrences have the kind of strong causal
efficacy appropriate to conscious causes (cf. Kim, 1998). Again, perhaps we could decide against broad
properties, and in favour of narrow ones, on the grounds that only narrowness
is consistent with the authority which subjects have over their conscious
states (cf. Block, 1990).
But to my mind this is asking
rather a lot of the abstract arguments.
While the completeness of physics does argue strongly against
metaphysically independent dualistic properties, it does nothing to rule out
metaphysically dependent structural properties, and it is by no means clear
that these structural properties are lacking in the kind of causal strength
that is possessed by conscious properties (Papineau, 2002, section 1.7). Again, it is not obvious that first-person
authority about conscious states is incompatible with their external
constitution; an analogous
compatibility has been widely defended by externalists about propositional
attitudes, and externalists about consciousness have appealed to similar
arguments (Tye, 2000).
In any case, this is not the
place to pursue these issues. As I said, I am pessimistic. So I shall assume, for the sake of the
subsequent argument, that general abstract considerations are not sufficient to
decide whether conscious properties are structural rather than physical, or
broad rather than narrow, or between other similar alternatives. My main point so far has been that, if I am
right about the impotence of general abstract arguments, then it is no use
turning to empirical science, for that won’t help us either.
If
abstract analysis combined with empirical investigation cannot identify unique
material referents for phenomenal concepts, does this mean that we must remain
permanently ignorant of the basic material facts about consciousness? Is there some epistemological barrier that
stops us finding out whether the silicon doppelganger is actually in pain?
Not at all. I do not take the above arguments to show
that there are real truths about consciousness which lie beyond our
epistemological access. Rather, the
underlying problem is that our phenomenal concepts are vague. Because of this, there is simply no fact of
the matter as to whether the doppelganger is in pain. Our phenomenal concept of pain is not focused enough to decide
the issue. Nothing in the semantic
constitution of the concept decides whether it refers to the physical mechanism
P which realizes pain in humans, or to the structural S which we humans share
with the doppelganger. So not even God,
who knows everything, would be able to tell whether or not the doppelganger
with S is in pain, for roughly the same reasons that God would be unable to
tell whether I am bald or not.
Actually, an analogy with
baldness is less than perfect, since baldness is a matter of degree, vague
because there is no sharp cut-off point.
Phenomenal concepts, by contrast, are vague because it is indeterminate what
kind of property they refer to, rather than where to divide some
continuum. A better analogy is provided
by the traditional Eskimo word for whale oil (Block, forthcoming). At some point in Eskimo history, a petroleum
product which looked and functioned just like natural whale oil was introduced
as a substitute. Did the original
Eskimo word (let’s pretend it was ‘whale oil’) apply to this new substance or
not? If ‘whale oil’ referred to a
biologically or chemically identified type, it did not; but if ‘whale oil’ referred to anything with
the requisite appearance and use, then it did.
But of course there is likely to be no answer here. Nothing in previous Eskimo linguistic
practice or usage need have determined which way ‘whale oil’ should be read
when this question arose. And then
there would be no fact of the matter as to whether the new stuff is whale
oil. Similarly, I say, with pain. There is no fact of the matter as to whether
the doppelganger is in pain, because nothing in previous usage fixes whether it
should apply to physical P or to structural S when the question is raised.
Now, this line on consciousness would be uncomfortably ad hoc if my only reason for saying that phenomenal concepts are vague were that we can’t find answers to the questions they pose. It is a mistake to infer an indefiniteness of questions directly from an unavailability of answers. This would require the mistaken verificationist doctrine that our concepts cannot lay claim to matters beyond our epistemological access. However, I think that there are good independent reasons for taking phenomenal concepts to be vague.
I take phenomenal concepts to be basic unstructured
terms, whose referential powers depend on the information they (are designed)
to causally track (Papineau, 2002, chapter 4).
At bottom, I take this to be a matter of tracking the experiences of
human beings. But experiences are
material, if they are anything, and, as we have seen, there are various
different material properties, physical and structural, broad and narrow, which
will serve to make the same classifications among normal human beings. So there is no reason to suppose that there
is anything in the semantic constitution of phenomenal concepts to focus them
precisely on one rather than another of these humanly coextensive material
properties—just as there is no reason to suppose that anything focuses the
Eskimo term for ‘whale oil’ on one possible referent rather than another.
No doubt some readers will feel that this analogy
with ‘whale oil’ is misplaced. In the
whale oil case, it seems clear that nothing is at issue but a matter of
conceptual refinement. If we read
‘whale oil’ as referring to a chemical substance, then the new stuff doesn’t
count as ‘whale oil’—but if we go the other way, it does. The pain case, by contrast, doesn’t seem
like a matter of conceptual decision at all.
True, if we were to read ‘pain’ as referring to the structural S, then
the doppelganger would be in the extension of ‘pain’, and if we were to read
‘pain’ as referring to the physical P, it wouldn’t be. But in this case we feel there remains a
further question. We want to know
whether the doppelganger is in pain, not how to apply some term, and we
feel that any decision as to whether ‘pain’ refers to S or P should respect
this concern, with ‘pain’ referring to whichever material property in fact
fixes real pain.
Still, I want to stand by the
analogy with the whale oil case. In my
view, the pain case seems different only because we find it so natural to think
about conscious experiences dualistically.
We unreflectively assume that pain is metaphysically distinct from any
material properties like S or P.
And once we do this, then of
course we take there to be a fact of the matter as to whether the doppelganger
has this distinct property or not. But
this is all a mistake. There isn’t any
distinct pain property, different from physical and structural material
properties, or broad and narrow ones, any more than there is a distinct whale
oil property, different from chemical constitution, biological origin,
appearance and use. So the person who
insists on asking whether the doppelganger is in pain is like an Eskimo who
says ‘Yes, I can see that this new manufactured stuff is like the old stuff in
appearance and use, while different in chemical constitution and biological
origin. But that doesn’t yet answer my
question, which was whether it is like the old stuff in being made of whale
oil.’ This person would be
assuming that being made of whale oil is some extra property, over and above
constitution, use and so on. Similarly,
the persistent desire to know whether the doppelganger is really in pain
stems from the assumption that pain is some extra property, over and above
physical and structural ones.
Why is it so compelling to
suppose that conscious experiences involve extra properties, over and above
material properties, when this seems absurd for whale oil? This is a large question, which I shall not
even attempt to answer here. My Thinking
About Consciousness (2002) is a sustained attempt to locate the source of
this near-inescapable ‘illusion of distinctness’. For the present, let me simply state that I take this illusion to
stem from the special structure of phenomenal concepts, and in particular from
the way they ‘quote’ the experiences they refer to: this quotational feature makes us feel that other, material ways
of referring to experiences ‘leave out’ the experiences themselves, and thus
that the phenomenal concepts must refer to something extra non-material property.
XII Relation to Earlier Debates
Philosophical
debates about alternative material referents for mental terms are not new. The question of whether terms like ‘pain’
refer to higher-order role properties or to realizer properties has been a staple
of analytic philosophy of mind for some decades. (See the essays in Block (ed.), 1980.) Similar issues about broad and narrow properties are also
familiar. However, these debates are
normally taken to be about the referents of ‘psychological’ concepts defined in
terms of theoretical roles. In this
paper I have focused instead on the referents of phenomenal concepts. I think is it is helpful to recognize that
psychological and phenomenal concepts raise different issues.
If our only concern were the referents of
‘psychological’ role concepts, then it would be quite mysterious why ‘pain’
strikes us so differently from ‘whale oil’.
The psychological concept of pain is defined in terms of canonical
causes like bodily damage and effects like avoidance behaviour. Now, does the term refer to any being
displaying this structure of causes and effects, or only to those who share the
physical property that mediates this structure in humans? So far this looks like a simple question for
decision, quite analogous to the ‘whale oil’ case—after all, concepts defined
in terms of theoretical roles can be taken either way, so what more is needed
but a decision about how to use the term ‘pain’?
To make sense of
the importance attached to these debates, we need to read them as
focused on the referents of phenomenal concepts. This is because it is specifically
phenomenal concepts that give rise to the illusion of extra, non-material
referents, in the way indicated at the end of the last section. And it is this illusion that makes us feel
that there must be something more to our questions about reference than
opportunities for terminological decision.
It is specifically because we think that phenomenal pain involves
something metaphysically independent of physical nature that we feel that the
doppelganger issue cannot simply be settled by terminological fiat alone.
XII What It Is Like
Does
it really make sense to hold that it is vague whether the doppelganger is in
pain? How can this be? Either the doppelganger feels a certain way
or it does not. It can’t be vague how
it feels.
But I do not want to argue that
there is something less than definite in the doppelganger’s experience
itself. The doppelganger definitely
feels as it does, however that is. My
thesis is only that it is indeterminate whether it is pain, where I take
this indeterminacy to arise from a vagueness in our human term ‘pain’. The indeterminacy doesn’t lie in the
doppelganger’s experience itself, but in whether that experience is similar
enough to cases of human pain to fall under our term ‘pain’, a term whose
content derives from exemplars provide by human pains.
At this point some readers may
want to ask about the general phenomenal concept of consciousness-as-such,
as opposed to specific phenomenal concepts like pain or seeing
something red. Maybe it makes sense
to hold that it is vague whether doppelgangers count as in ‘pain’, on the
grounds that it is unclear how similar their experiences are to human
pain. But surely I don’t want to say
whether it is vague whether they are conscious at all. Mustn’t this at least be definite, however
it is with their more specific modes of experience?
But I do want to say that even phenomenal
consciousness-as-such is vague. So far
in this paper I have focused on specific phenomenal concepts like pain. But everything I have said about these
applies equally to the general phenomenal concept conscious-as-such.
As I observed at the beginning of this paper, there
are no a priori connections between the phenomenal concept conscious-as-such
and any material concepts. So it will
be matter of empirical research to identify the material referent of the
phenomenal concept conscious-as-such.
However, when we look for some material property common to all those cases
where humans report they are phenomenally conscious, and absent whenever they
deny this, we will inevitably find a variety of material candidates which fit
perfectly with this data base. Apart
from any strictly physical property, there will also be various structural
properties. Again, there will be broad
properties as well as narrow ones. In
addition—though this is not the place to go into details—there will be candidates
involving Higher-Order Thoughts as well as candidates involving only lower-order
attentional properties (Papineau, 2002, section 7.11 to 7.13).
Just as before, I do not take this to show that we
are trapped in ignorance about the true material nature of
consciousness-as-such. Rather, the
fault lies with our phenomenal concept of this property, which is vague in the
crucial dimensions. We somehow
construct this concept by taking exemplars of human conscious states, and use
the resulting concept to classify humans depending on whether or not they are
in some such conscious state or not.
There is no reason to suppose the concept so constituted is precisely
enough focused to discriminate between the physical and structural properties
shared by conscious humans, or between the broad and narrow properties, or
between the Higher-Order and attentional properties.
Still, readers might wish to press me, how can it be
vague whether a silicon doppelgangers is conscious at all? (Surely a light is on, or it is not.) But, as before, my thought is not that it is
vague how it is for the doppelganger itself.
Rather, the vagueness lies in our concept, which fails to decide whether
the doppelganger’s mode of being is sufficiently like our own to count as
‘conscious’ or not.
But, you may wish to persist, isn’t it either like
something to be the doppelganger, or not?
How can that be vague? (Surely a
light is on, or it is not.) But I am
sceptical as to whether the phrase ‘like something to be’ will bear the weight
of this insistence. There are many
modes of being apart from our own, including those of cats, fish, slugs,
plants, stones, possible silicon humanoids, and no doubt many others. I see no reason to suppose that the phrase
‘like something to be’ will draw a sharp line across this range, with some beings
coming out as clearly conscious, and the rest not. The various cases will all display some material similarities
with human consciousness, and some differences. Given this, there will be nothing to determine whether they fall
under our term ‘conscious’ or not.
XIII Conclusion
Over
the past decade or so, there has been a growing wave of enthusiasm for
scientific research into consciousness.
After a long period when consciousness was regarded as beyond the
scientific pale, many scientists now aim to penetrate the mysteries of
consciousness and identify the material mechanisms that underlie conscious
experience. If the argument of this
paper is sound, however, their hopes will at best be partially fulfilled.
The most we can expect from scientific research is
the identification of a range of material properties that correlate in human
beings with pain, say, or colours, or indeed being conscious at all. However, this won’t pinpoint the material
essence of any such conscious state, for there will always be a plurality of
such human material correlates for any conscious property. And in the face of this plurality science is
impotent, for the methodology of consciousness research offers no handle by
which to separate material properties that are perfectly correlated in normal
humans.
So scientific research into consciousness is fated
to disappoint its more extreme enthusiasts.
This does not mean, however, that we would do better to turn to some
alternative mode of investigation. It
is not as if conscious properties have true material essences, yet science is
unable to discover them. Rather the
whole idea of identifying such essences is a chimera, fostered by the
impression that our phenomenal concepts of conscious states are more precise
than they are.
References
Block,
N (ed.). 1980. Readings in the Philosophy of Psychology, vol. 1. London:
Metheun.
Block,
N. 1990. ‘Inverted Earth’, in Philosophical Perspectives 4, ed. J.
Tomberlin. Northridge, Calif: Ridgeview.
Chalmers,
D. 1996. The Conscious Mind. Oxford: Oxford University Press.
Dennett,
D. 1978. ‘Why You Can’t Make a Computer That Feels Pain’, in his Brainstorms.
Cambridge, Mass: Bradford Books.
Dretske,
F. 1995. Naturalizing the Mind. Cambridge, Mass: MIT Press.
Frith,
C., Perry, R., and Lumer, E. 1999. ‘The Neural Correlates of Cognitive
Experience: An Experimental Framework. Trends in Cognitive Neuroscience,
3.
Goodale,
M. and Milner, A. 1992. ‘Separate Visual Pathways for Perception and Action’. Trends
in Neurosciences, 15.
Jackson,
F. 1986. ‘What Mary Didn’t Know’. Journal of Philosophy, 83.
Kim,
J. 1998. Mind in a Physical World. Cambridge, Mass: MIT Press.
Kripke,
S. 1980. Naming and Necessity. Cambridge, Mass: Harvard University
Press.
Lycan,
W. 1987. Consciousness. Cambridge, Mass: MIT Press.
Papineau, D. 2002. Thinking
about Consciousness. Oxford: Oxford University Press.
Tye,
M. 2000. ‘Inverted Earth Meets Swampman’, in his Consciousness, Color and
Content. Cambridge, Mass: MIT Press.
Weiskrantz, L. 1986. Blindsight.
Oxford: Oxford University Press.