What is the relationship between being conscious and knowing about
consciousness? In answering this question, constitutive panpsychists
face a delicate balancing act: their own case against physicalism
requires that being conscious reveals something of the metaphysics of
consciousness, but the stronger they make this claim of revelation, the
stronger becomes an objection to their own view sometimes called “the
revelation problem”. In this paper I argue that this balancing act,
though delicate, is not impossible: there is a plausible, well-motivated
“medium-strength” sort of revelation, strong enough to bring down
physicalism but weak enough to leave constitutive panpsychism
standing.
In section 1, I lay
out the background to the panpsychism-physicalism debate; in section 2, I
distinguish six “revelation theses”; in section 3 I analyse
the structure and varieties of the revelation problem; and in section 4 and
section 5 I outline how
to address this problem while retaining as much as possible of the
theses discussed in section 2.
Are Panpsychists Hoist by their
Own Petard?
Panpsychists think all the fundamental physical things are
phenomenally conscious, where “fundamental physical things” is a
placeholder for whatever fundamental entities feature in the true
physical theory (particles, fields, strings, spacetime, etc.). The
“constitutive” part of “constitutive panpsychism” describes the
relationship between macroexperiences (the experiences of humans and
other animals) and the postulated microexperiences of the fundamental
physical entities. This relationship should be
something like the relationship between the physical features of human
bodies (macrophysics) and the physical features of the fundamental
entities (microphysics). That relationship (which we might call being
constituted, being grounded, or being nothing over and above) generates
no “explanatory gap”: even when the details currently elude us, it seems
clear that macrophysics is fully accounted for by microphysics. When you
have the right particles, arranged in the right pattern, exerting the
right forces on one another, and the right laws governing them, there is
no further problem about how to get hands, chairs, planets, etc.: those
“come for free” when the microphysical foundations are there.
The failure of consciousness to fit into this neat picture is the
objection to physicalism that motivates most contemporary panpsychists.
Whereas the distribution of and relations among subatomic particles
seems to explain everything about my body, it leaves unexplained why
there is anything it feels like to be me, and why it feels the
particular way it does. In particular, even knowing the full story about
the particles seems to be compatible with not knowing what the
experiences are like (this is the
“knowledge argument,” cf. Jackson 1982; Nemirow 1990; Ball 2009),
and it seems that a world might have been physically identical and yet
differed from ours in respect of consciousness (the “conceivability
argument,” cf. kripke:1980?; Chalmers 2009).
There is a vast literature on whether these are good reason to reject
physicalism (see,
e.g, Chalmers 1996; Dennett 2007; Stoljar 2006; Dı́az-León 2011),
but here I will assume that they are. What comes next? In particular, is
constitutive panpsychism, often offered as an attractive non-physicalist
alternative, defensible?
Constitutive panpsychism treats consciousness as a fundamental
ingredient of nature, but tries to treat it the same as other
fundamental ingredients (mass, charge, spin, force, location, etc.).
Just as those other fundamentals are widespread in nature, with human
beings as simply one particular arrangement of them, so is
consciousness: human experience is not metaphysically special, just a
complicated combination of widespread components. Constitutive
panpsychism thus retains the monistic spirit of physicalism despite
recognising consciousness as fundamental. Importantly, non-constitutive
versions of panpsychism, on which human consciousness somehow “emerges
from” or is “caused by” microconsciousness but not literally “made up
of” it, do not secure this advantage. The macrophysical properties of
the brain seem to be wholly constituted by the microphysical properties
of its parts, so if its macroscopic consciousness is not similarly
constituted by microconsciousness, the hoped-for reconciliation of mind
and matter falls apart.
This imposes an explanatory burden: constitutive explanations of
human consciousness in terms of microconsciousness have to do better
than physicalist explanations. And one major line of criticism has been
that they do not: there is just as much difficulty in explaining how
many simple minds combine into complex minds as in explaining how
mindless things generate minds. This broad objection is often called
“the combination problem” (Seager 1995,
280; Chalmers 2017; Roelofs 2019), and has received much
discussion from both defenders and critics of panpsychism.
One specific strand of the combination problem is “the revelation
problem”: macroexperiences do not seem introspectively to be
built up out of microexperiences. And constitutive panpsychists can’t
just say: “Well they are, sometimes things aren’t what they
seem.” That would license physicalists to likewise say: “Exactly!
Consciousness seems distinct from purely physical facts, but
it’s actually not.” If being conscious doesn’t reveal the true nature of
consciousness, the case against physicalism is weakened; if it does,
then the truth of constitutive panpsychism should be introspectively
obvious, which it is not.
This talk of “seeming” and “obviousness” is not the most precise way
of presenting things. Authors articulating the sense that there is a
problem here say things like:
[…] it is hard to see how smooth, structured macroscopic
phenomenology could be derived [from microexperiences isomorphic to
microphysics]; we might expect some sort of “jagged,” unstructured
phenomenal collection instead. (Chalmers 1996, 306)
It is hard to see how [microexperiences] could somehow add up to the
phenomenal properties with which we are familiar—properties with the
specific, homogeneous character with which we are all acquainted […].
(Alter and
Nagasawa 2012, 90–91)
[Revelation is] inconsistent […] with my conscious experience turning
out to be, in and of itself, quite different from how it appears to be
in introspection: i.e., turning out to be constituted of the
experiential being of billions of micro subjects of experience […].
(Goff 2006,
57; cf. Lee 2019, 290–298)
Similar remarks were made by certain non-reductive mind-brain
identity theorists in the last century, writing about a perceived “grain
problem”:
[Any experience’s] physiological substrate, presumably, is a highly
structured, not to say messy, concatenation of changes in electrical
potential within billions of neurons in the auditory cortex […]. How do
all these microstructural discontinuities and inhomogeneities come to be
glossed over […]? (Lockwood 1993, 274)
How is it that the occurrence of a smooth, continuous expanse of red
in our visual experience can […] involve particulate, discontinuous
affairs such as transfers of or interactions among large numbers of
electrons, ions, or the like? (Maxwell 1978, 398)
Indeed, Lewis makes a very similar argument, though he rejects the
idea that experience reveals its nature and so presents the argument as
a reductio of this idea:
If we know exactly what the qualia of our experiences are, they can
have no essential hidden structure - no “grain” - of which we remain
ignorant. If we didn’t know whether their hidden “grain” ran this way or
that, we wouldn’t know exactly what they were. […] if nothing essential
about the qualia is hidden, then if they seem simple, they are simple.
(Lewis 1995, 142,
n.14)
Although I think all the above quotations express a similar sort of
concern, they do so with different emphasis and framing, and the exact
nature of the problem involved is far from clear. In section 3 I try to
identify the problems more precisely, and in section 4 and
section 5, I resolve
them.
The Revelation Problem and the
Revelation Thesis
Before examining the revelation problem for panpsychism, we need to
examine the background idea of a “revelation thesis” connecting
consciousness to knowledge of consciousness. There are actually several
different ideas under the broad heading of “revelation”: I will
distinguish a total of six distinct revelation theses, resulting from a
two-fold distinction permuted with a three-fold distinction.
The two-fold distinction concerns whether the claim says (a) that the
full truth about consciousness will always be manifest (a “reality\(\rightarrow\)appearance” direction of
implication), or (b) that what is manifest about consciousness is always
true (an “appearance\(\rightarrow\)reality” direction of
implication). Claims of the first sort rule out
any aspect of consciousness being “hidden” from us, while claims of the
second sort rule out any sort of “illusion” about consciousness.
The three-fold distinction is about the topic of a revelation thesis
- what kind of reality it connects with what kind of appearance. Putting
things for now in reality\(\rightarrow\)appearance terms, we can
distinguish the claims:
That someone having an experience can know that they are
presently having that token experience;
That someone having an experience can gain a special kind of
understanding of that phenomenal property;
That this understanding reveals “the complete nature” of a
certain type of experience.
The first thesis is sometimes called “self-presentation” or
“luminosity”, as distinguished from “revelation” (Stoljar 2006, 223). But in other
discussions it is presented as an integral part of a broader idea called
“revelation.” (e.g.,
Goff 2017, 109–110). The second thesis is sometimes put in terms
of forming concepts, sometimes of special sorts (e.g., Chalmers 2003a; Goff
2017, 109–110) and sometimes just in terms of “understanding”
(e.g., Stoljar 2006,
229). The third thesis is sometimes put in terms of knowing a
phenomenal property’s “essence” or “nature”, or knowing all the
essential or necessary truths about it.
Sometimes the term “revelation” or “revelation thesis” is used
specifically for one of these theses, or for the set of them together,
or for the conjunction of the second and third. But they are worth
distinguishing because, as I will show, they support quite distinct
revelation arguments against constitutive panpsychism, which need to be
addressed in quite different ways.
Moreover, we can distinguish reality\(\rightarrow\)appearance and appearance\(\rightarrow\)reality directions of each of
the three, yielding a total of six revelation theses (RT1–RT6), as
follows:
I think these six theses, though logically independent, form a fairly
natural package together, and I will refer to this package (i.e., the
conjunction RT1–RT6) as “the revelation approach”.
This package is particularly important for undergirding modal arguments
against physicalism, a role which it is held to have both by its
defenders and its critics (e.g.,
Stoljar 2009, 2013; Damnjanovic 2012; Liu 2019, 2020). Lewis, for
instance, attributes RT5 and RT6 to Kripke, as a presupposition of the
latter’s inference from the conceivability of pain without any
associated brain state to their separate possibility (Lewis 1995, 328,
n.3). Goff (2017,
74–76, 96–106) likewise argues that the conceivability and
knowledge arguments require that phenomenal concepts be “transparent”,
effectively meaning that RT5 and RT6 must be true. And
Chalmers’ version of the conceivability and knowledge arguments relies
on the premise that the primary and secondary intensions of phenomenal
concepts are equivalent (Chalmers 2003b,
2009), which implies RT5 and RT6.
Although RT5 and RT6 have the clearest role, the falsity of the other
revelation theses would also leave the anti-physicalist arguments on a
shaky footing. For instance, if RT3 were false, we could worry whether
we possessed the pure phenomenal concepts whose “transparency” drove the
arguments; if RT2 were false, we could worry that the properties these
concepts expressed were not even instantiated (as argued by, e.g.,
Pereboom 2016, 2019); and RT4 is essential to the knowledge
argument, which relies on the premise that someone who has never
experienced colour cannot know what seeing colour is like.
What is the Revelation Problem,
Exactly?
So what exactly is the supposed problem for panpsychists? How is it
distinct from other aspects of the combination problem? Fundamentally,
it concerns a perceived incompatibility between three things:
the way human consciousness appears in introspection;
the way human consciousness would be, if constitutive panpsychism
were true;
revelation: the idea that introspection gives special insight
into the reality of consciousness.
The third element makes any discrepancy between the first and second
seem fatal. Yet that third element is also something panpsychists cannot
readily give up.
How should we spell out these core elements? I think there are
actually three slightly different arguments to be made here, and then a
fourth argument which engages with the debate on a different combination
problem, the “palette problem”. Let us consider the pure revelation
arguments first, which differ primarily in whether they rely on the
appearance\(\rightarrow\)reality or
reality\(\rightarrow\)appearance
direction of implication: the first argument says, “Consciousness
appears to be X, but panpsychism implies it is not really X,” while the
second and third say, “Consciousness fails to appear to be X, but
panpsychism implies it really is X.” The first focuses on some positive
introspective appearance, and accuses constitutive panpsychists of
treating that appearance as an “illusion”. The others focus simply on
the absence of a certain appearance.
We can call the first argument the “no illusions” argument, since its
third premise is RT2, the “no illusions” thesis:
If constitutive panpsychism is true, then human consciousness is
always “particulate”.
Human consciousness (often) appears introspectively to be
“smooth”.
Consciousness can’t appear a way that it’s not. (RT2)
Being “smooth” and being “particulate” are incompatible.
Human consciousness is (often) smooth. (from 2 and 3)
Human consciousness is (often) not particulate. (from 4 and
5)
Constitutive panpsychism is false. (from 1 and 6)
Obviously much turns on the meaning of the terms “particulate” and
“smooth”, but despite the frequency with which they (and similar terms
like “continuous” and “fragmented”) appear in statements of the problem,
it is unclear how to define them, and consequently unclear how plausible
premises 1, 2, and 4 are. This definitional question will be central to
my discussion in the next section.
The second and third arguments (involving a “reality\(\rightarrow\)appearance” implication) are
both suggested in Chalmers’ formulation of what he calls “the revelation
argument” (2017,
190). Chalmers notes that although constitutive panpsychism holds
consciousness to be “constituted by a vast array of microexperiences”,
this vast array is not revealed to us in introspection. This poses a
problem if we think both that introspection reveals the nature of
consciousness, and that “whatever constitutes consciousness is part of
its nature”.
I distinguish two arguments here because I think talk of
“introspection” upon “consciousness” can be taken in two quite different
ways. One is that introspection focused on macroexperiences
doesn’t reveal that they are constituted by microexperiences.
The other is that introspection focused on microexperiences
isn’t even possible. The former appears to violate what I above called
RT5, the “self-intimation” thesis: reflection upon a pure phenomenal
concept reveals the whole nature of a phenomenal property. The latter
appears to violate both what I above called RT3, the
“understanding-from-experience” thesis, and RT1, the “self-presentation”
thesis: having an experience should allow knowledge of its occurrence
and a pure phenomenal concept of it.
Focusing on either macroexperiences or microexperiences yields the
following two arguments, which I will call the “macroexperience-focused”
and “microexperience-focused” argument. The first runs thus, with RT5 as
third premise:
If constitutive panpsychism is true, each human experience
(“macroexperience”) is constituted by a vast array of
microexperiences.
A vast array of microexperiences is not revealed by reflection on
macrophenomenal concepts (i.e., phenomenal concepts based on
macroexperiences).
The nature of a phenomenal property is revealed by reflection on
phenomenal concepts based on experiences of it. (RT5)
Whatever constitutes something is part of its nature.
The natures of macroexperiences do not involve vast arrays of
microexperiences. (from 2 and 3)
Macroexperiences are not constituted by vast arrays of
microexperiences. (from 4 and 5)
Constitutive panpsychism is false. (from 1 and 6)
Clearly, the soundness of this argument depends crucially on what is
meant by talk of a property’s “nature”, since that will affect the
meaning of premises 3 and 4; this question will be at the heart of my
discussion in the next section.
The third (“microexperience-focused”) revelation argument runs thus,
with a conjunction of RT1 and RT3 as its third premise:
If constitutive panpsychism is true, consciousness is constituted
by a vast array of microexperiences.
We cannot know introspectively about microexperiences, nor form
microphenomenal concepts (i.e., phenomenal concepts based on
microexperiences).
If a subject is having an experience, they can know
introspectively that they are, and form phenomenal concepts based on it.
(RT1 and 3)
If experiences constitute a subject’s consciousness, that subject
undergoes them.
We are not undergoing a vast array of microexperiences. (from 2
and 3)
Human consciousness is not constituted by a vast array of
microexperiences. (from 4 and 5)
Constitutive panpsychism is false. (from 1 and 6)
Finally, there is an interaction between a revelation thesis,
specifically RT5, and another aspect of the combination problem, the
“palette problem”. How do the huge range of phenomenal qualities that
humans experience arise from a fundamental base which appears to involve
only a quite small number of fundamental properties? One solution is the
“small palette hypothesis”: there are only a few basic phenomenal
qualities, corresponding to the fundamental physical properties, which
are somehow “blended” to generate a plethora of different qualities for
different macroscopic creatures (see
Roelofs 2014; Coleman 2015, 2017; Chalmers 2017, 204–206), whose
pattern of similarities and differences are explained by their differing
proportions of the basic ingredients. Some critics of the small palette
hypothesis object that some of our phenomenal qualities are too
heterogeneous to be blended out of a small set of common elements,
because they are completely dissimilar, with nothing
phenomenally in common. Goff (2017, 195), for instance, claims
that, “Minty phenomenology and red phenomenology have nothing in common”
(cf. a similar
argument in McGinn 2006, 96). This line of criticism relies on
RT5 to rule out these qualities being similar in a way that we cannot
recognise (Goff 2017,
195–197). Call this the “small-palette revelation argument”, the
full structure of which is very similar to that of the
macroexperience-focused revelation argument:
If the small palette hypothesis is true, then any two phenomenal
qualities experienced by humans have something phenomenal in
common.
Reflection on some pairs of human experiences (e.g., red and
minty) does not reveal them to have anything phenomenal in
common.
The nature of a phenomenal quality is revealed by reflection on
phenomenal concepts based on experiences of it. (RT5)
The natures of two things determine whether they have anything
phenomenal in common.
If a pair of phenomenal qualities has something phenomenal in
common, reflection on phenomenal concepts based on experiences of them
will reveal this. (from 3 and 4)
Some pairs of human experiences have nothing phenomenal in
common. (from 2 and 5)
The small palette hypothesis is false. (from 1 and 6)
All four arguments have a similar four-premise form: first, a
supposed implication of constitutive panpsychism (or small-palette forms
of it); second, an introspective datum; third, an epistemological thesis
about introspection; and fourth, a metaphysical claim, given which the
other three premises entail the falsity of constitutive panpsychism (or
small-palette forms of it). But despite their common form, I will argue
that the arguments go wrong in quite different ways.
Ways of Responding to the
Revelation Arguments
The challenge for constitutive panpsychists is to rebut the above
four arguments without abandoning the revelation approach, components of
which underpin all of them. I will show how to rebut each argument in
turn, while keeping the relevant revelation theses as strong as I
can.
The No-Illusions Revelation
Argument
Consider first the “no illusions” argument, which had the following
four premises:
If constitutive panpsychism is true, then human consciousness is
always “particulate”.
Human consciousness (often) appears introspectively to be
“smooth”.
Consciousness can’t appear a way that it’s not.
Being “smooth” and being “particulate” are incompatible.
One option for constitutive panpsychists is to deny premise 1, based
on defining “particulate” in such a way that a field-based ontology, or
a substance-monist ontology, or some other account of physical reality,
renders it false that the material world, and any consciousness
isomorphic to it, is particulate (see in particular Nagasawa and Wager
2017, 120–121). If the other three premises (and constitutive
panpsychism) are accepted, this implies that the kind of consciousness
we enjoy is incompatible with some physical theories (those which make
matter “particulate”) and that we know introspectively that our world is
not any of those ways.
However, I think this approach is a mistake. Even if particles are
not ultimately real, Lockwood’s point still holds: even the simplest
experience involves billions of neurones, ions, and neurotransmitters.
Even if the space containing two sodium ions is ultimately just a set of
derivative aspects of the one substance, there is still a striking
difference in the electrical properties of different regions of that
space. To dismiss the problem because particles are not in the
fundamental ontology would be too easy. Consequently, I suggest the
following definition of “particulate”:
X is particulate iff X comprises a very large but finite
number of parts which differ significantly (in some properties) and
discontinuously (on some dimension).
This definition makes the physical brain particulate whatever the
fundamental physics turns out to be. Of course this definition will only
be as precise as “very large” and “differ significantly and
discontinuously”. The vagueness of such terms does not stop us from
taking “a trillion or more” as a clear case of “very large”, and “the
mass and charges differences between a water molecule, a potassium ion,
and a region of empty space between them” as a clear case of “differ
significantly and discontinuously”.
That leaves three remaining options: deny premise 2 (i.e., contradict
the supposed introspective observation), deny premise 3 (i.e., reject
this particular revelation thesis), or deny premise 4 (i.e., deny that
smoothness and particulateness are incompatible). But everything depends
on what “smooth” means. What is the feature of experience that is being
reported by those who feel the pull of this argument?
One option is to define “smooth” by ostension: consider some
experiences without discernible internal structure, what Lockwood (1993,
274) calls a “phenomenally flawless” experience, and stipulate
that “smooth” means the noteworthy feature of those experiences. That
would ensure the truth of premise 2, but would make it hard to
adjudicate the truth of premise 4. My preference is to define “smooth”
in such a way as to ensure the truth of premise 4, e.g:
X is smooth iff it is not particulate.
There are then a few different ways for something to be smooth: since
being particular requires parts, for instance, simple things would count
as smooth by default. Alternatively, something might be smooth if its
parts do not differ significantly in any respect, or do not differ
discontinuously along any dimension. The panpsychist must then deny
either premise 2 or premise 3: either say that experience does not
appear smooth, or say that it does but isn’t.
At first glance, both options look difficult: premise 3 is, after all,
part of the Revelation Approach (RT2), and if premise 2 is false, why
did anyone ever advance the argument in the first place?
The way out lies in scrutinising the word “appears”, and drawing a
distinction between illusions, strictly so-called, and easy
misinterpretations. Consider some non-mental examples: at first an act
appears noble, an argument compelling, a speech beautiful, and yet then
I find that upon giving the matter more thought, this appearance
vanishes, and I come to think I was mistaken. The act now appears
fanatical, the argument sophistical, the speech saccharine; I think
myself foolish for being gullible enough for the act, argument, or
speech to ever appear otherwise to me. I might say I was subject to an
“illusion”, but all this mean is that the act, argument, and speech were
such that they could be very readily misjudged.
Contrast this with a white object seen under pure red light, or a
straight stick seen half in water, or an ambitious Scottish nobleman
hallucinating a dagger. The object appears red but isn’t, the stick
appears bent but isn’t, and there appears to be a dagger, but there
isn’t. Here no reflection on the appearances will change them, and the
subject cannot hold themselves rationally accountable for being subject
to them (perhaps for forming beliefs based on them, but not for the
appearances themselves). Here we have a stronger sense of “illusion”: it
is not that these perceptions are easy to misjudge, it is that their
very content is false. Call this the “quasi-perceptual” sense of
“appears”, contrasting with the “ready-interpretation” sense (cf. Stoljar
2013; Kammerer 2018).
Premise 3 (RT2) is most plausible if read with the “quasi-perceptual”
sense of “appears”. Plausibly it makes no sense to think that my
impression of my own experience is an “illusion” in this stronger sense:
surely it would be the “impression” that deserves to be called my
experience, since this is what I am immediately aware of. To think that
consciousness might appear falsely in this way seems to involve
forgetting that consciousness is how things appear to me (cf. Liu 2020). Or at
least, this thought has some appeal, and panpsychists need not disagree
with it.
But premise 3 is less plausible if understood in terms of the
“ready-interpretation” sense of “appears”, saying that if consciousness
is readily interpreted as having some property, it must actually have
that property. After all, which interpretations come readily depends on
the subject’s expectations, background assumptions, interpretive style,
etc. An absolute principle, that no false interpretation could come
readily to anyone, would be very close to saying, implausibly,
that consciousness was never misinterpreted.
So we should read premise 3 as saying that consciousness cannot
appear a way it’s not, in the quasi-perceptual sense of “appear”. For
the argument to remain valid, premise 2 must also be read in terms of
the quasi-perceptual sense of “appear”, not the “ready-interpretation”
sense. But now premise 2 is much more deniable. We can deny premise 2,
in this strong sense, by taking the appearance of smoothness to be a
matter of what interpretations come readily, and not of how things
quasi-perceptually appear.
This is my preferred response to the “no illusions” argument: our
consciousness really is particulate, not smooth, but it is readily
misinterpreted as smooth. But this misinterpretation demands an
explanation - what is it about the way consciousness does
appear, which makes us judge it “smooth”?
One answer appeals to the difference between represented structure
and structured representations: that is, experience represents things as
being smooth, rather than itself being smooth (versions of this proposal
appear in: Clark 1989; Stoljar 2001). Critics have worried that
experience itself really does seem to display the relevant sort of
smoothness (e.g., Alter and Nagasawa 2012,
91), and that representing a smooth expanse may be insufficient
for introspectively seeming, even in the weak sense, to be smooth
(consider the sentence “space is infinitely divisible”). Another answer
is to say that many experiences quasi-perceptually appear to have, and
thus (by RT2) actually have, some property similar to, but not identical
to, “smoothness”. In section 5 I flesh out this
approach.
The Macroexperience-Focused
Revelation Argument
Next, consider the macroexperience-focused argument, whose premises
are:
If constitutive panpsychism is true, each human experience
(“macroexperience”) is constituted by a vast array of
microexperiences.
A vast array of microexperiences is not revealed by reflection on
macrophenomenal concepts (i.e., phenomenal concepts based on
macroexperiences).
The nature of a phenomenal property is revealed by reflection on
phenomenal concepts based on experiences of it.
Whatever constitutes something is part of its nature.
I see little prospect for denying premises 1 and 2,
and premise 3 is one of the revelation theses I want to preserve.
Chalmers, when he lays out the argument of which this is a variant,
advises panpsychists to attack premise 4: to drive a wedge between
something’s nature and what constitutes it. I agree that this is the
right tack, but everything turns on what kind of “nature” is in
question, which in turn depends on how we read premise 3, the
self-intimation thesis. I think there is a plausible and well-motivated
sense of “knowing a nature” which explains why premise 4 is false,
without undermining anti-physicalist arguments.
First consider this common gloss: knowing the nature of a property
means being in a position to know a priori every necessary truth about
that property. If I know the nature of squareness,
I am in a position to know a priori every necessary truth about
squareness (like what squares’ internal angles sum to, or what kinds of
triangles they can be divided into), though not to know contingent
truths about it (like whether it is my sister’s favourite shape).
Likewise if I know the nature of being water, I can know every necessary
truth about being water (like that water is a chemical compound, or its
molecular mass), though not every contingent truth about it (like
whether it is instantiated on Earth). This suggests that we know the
natures of mathematical properties, but do not automatically know the
natures of chemical properties, though perhaps we do now, given
scientific progress. And those results seem plausible.
But this gloss is inadequate. Consider someone who knew the nature of
squareness but not the nature of triangularity (if that were possible).
They would not be in a position to know a priori that every square can
be divided into four right-angled triangles. This suggests a refinement:
knowing the nature of some property means being in a position to know
a priori all the necessary truths about that property which
involve only other properties whose natures you also know. To put it
another way, to know a priori a necessary truth involving two
properties, you need to know the natures of both: just knowing the
nature of one is not enough. This implies, in
particular, that knowing the nature of a constituted property is not
sufficient to know about its constitution relationships to other
properties, without also knowing the natures of those other
properties.
I think this provides a plausible reading of “knowing a property’s
nature”, and thereby of RT5, which does precisely what constitutive
panpsychists need it to do: substantiate their arguments against
physicalism, without substantiating the revelation argument against
their own view. For on this reading of “knowing a nature”, that we know
the natures of macrophenomenal properties implies that for any other set
of properties whose natures we know, we are in a position to tell a
priori whether those properties are sufficient to constitute
macrophenomenal properties. And the case against physicalism is that
physical properties do not seem a priori to constitute macrophenomenal
properties. Of course, this attack only works if we know the natures of
physical properties (e.g., if we think of them as exhausted by what
physics says about them, as what Stoljar (2001) calls the “t-physical”
properties, and what Strawson (2006) calls “physicsal”
properties). It will not work if we think of physical properties as
whatever properties physical things have which in fact account for their
satisfying the descriptions given by physics (what Stoljar (2001) calls
the “o-physical” properties). But that way out is no use to standard
physicalism, which needs physical properties to be well-understood: to
say that the reason the conceivability argument fails is that there is
some mysterious hidden nature of the physical, which plays some crucial
role in accounting for consciousness, is to embrace the kind of
“non-standard physicalism” (cf. Stoljar 2006) that is no longer
incompatible with panpsychism.
But why doesn’t knowing the natures of macrophenomenal properties
substantiate a parallel argument against constitutive panpsychism?
Because panpsychists do not claim that we know the natures of
microphenomenal properties, because we are not the microsubjects who
instantiate those properties (though see the next subsection for some
complications of this claim). Without knowledge of the candidate
constituting properties, we cannot determine a priori their suitability
to constitute macrophenomenal properties. All the constitutive
panpsychist is committed to is a conditional claim: if we were
able to grasp the natures of microphenomenal properties, then we could,
in principle, see a priori that, when properly arranged, they constitute
macrophenomenal properties.
The Microexperience-Focused
Revelation Argument
Thirdly, consider the microexperience-focused revelation argument:
why can’t we introspect microexperiences like we can macroexperiences?
The premises of this argument are:
If constitutive panpsychism is true, consciousness is constituted
by a vast array of microexperiences.
We cannot know introspectively about microexperiences, nor form
microphenomenal concepts.
If a subject is having an experience, they can know
introspectively that they are, and form phenomenal concepts based on
it.
If experiences constitute a subject’s consciousness, that subject
undergoes them.
Again, I see little hope in denying premises 1 or 2,
which leaves three options: deny premise 3 (“we are undergoing
microexperiences, but cannot introspect them”), deny premise 4
(“microexperiences constitute our consciousness, but we do not undergo
them”), or show the argument to be invalid.
Goff’s approach in his (2017, 189ff.) is to deny premise 4, to
“loosen” the relation between microexperiences and macroexperiences, so
that although microexperiences in some sense constitute (or “ground”,
“compose”, or “form”) macroexperiences, the phenomenal character of the
latter contains nothing of the former. The cost of this is that the
constitution relation between microexperiences and macroexperiences is
thereby made more mysterious. If this relation were one in which both
constituted and constituter were undergone by the same subject, it could
be akin to familiar relations among macroexperiences. For instance, the
relation between my total phenomenal field right now and the component
experiences that it subsumes (sounds I’m hearing, colours I’m seeing,
twinges of physical discomfort, etc.) is plausibly something like
constitution. It would be nice if panpsychists could assimilate the
microexperience-macroexperience relation to familiar relations like
this, where a single subject undergoes all the experiences involved;
without that link it is hard to see why microexperiences should really
be said to “constitute” a macroexperience, as opposed to somehow giving
rise to it as a distinct product.
I think the best approach is to say the argument is invalid when
premise 3 is qualified in certain ways that are independently
necessary to make it plausible. An unqualified form of premise 3 faces
easy counterexamples: ferrets undergo many experiences, but cannot form
phenomenal concepts, or know that they are having experiences. But
plausibly this is not a counter-example to what premise 3 was intended
to say! The problem is not that ferrets’ experiences are somehow hidden
from them, but just that they lack the conceptual competence to
recognise their experiences as such. A qualified version of premise 3
would allow for this: it would say that certain kinds of knowledge and
concept-formation are possible whenever a subject undergoes an
experience and meets various other conditions. Another
plausible requirement is attention: one must focus on an experience in
order to introspect it, and if one is unable to direct one’s attention,
introspection will be impossible.
So let us consider a qualified reading of premise 3, that includes
these conditions: introspective knowledge is possible whenever a subject
undergoes an experience, and is capable of conceptualising it,
and focuses their attention on it. The argument has now become
invalid: line 5 (“we are not undergoing a vast array of
microexperiences”) no longer follows from 2 and 3. There are two reasons
why we might be phenomenally undergoing microexperiences but be unable
to know them introspectively, compatibly with this weaker reading of
premise 3: if humans cannot conceive of experiences as such, or if they
are unable to attend to microexperiences. While the first of these
options is clearly false, the second is, I think, the best option for
the constitutive panpsychist in rebutting the microexperience-focused
argument.
This implies that while microexperiences are phenomenally conscious
for us, they are not access-conscious for us. That is, microexperiences
are presented to us, “right there”, characterising the phenomenal
character of our consciousness, but they are not presented in such a way
that we can cognitively select, access, and identify them. Our
relationship to them is rather like our relationship to elements of our
experience that are very faint, which require a lot of effort to focus
on and distinguish from their surroundings, and which it is
correspondingly easier to distract us from. If something in my
peripheral vision is roughly the same colour as its surroundings, it
would be hard for me to notice it, and if I were distracted, exhausted,
or inebriated I might find attending to it all but impossible. Yet it is
still part of my phenomenology, not somehow hidden from me. The
constitutive panpsychist, I am suggesting, should claim that this
near-impossibility of attending to peripheral vision while distracted is
intensified to a real practical impossibility with microexperiences. In
section 5 I situate this
impossibility claim within a broader picture of how the mind is
constituted by microexperiences, which will help to motivate this
response to the microexperience-focused argument.
The Small-Palette Revelation
Argument
Finally, consider the small-palette revelation argument, whose
premises are:
If the small palette hypothesis is true, then any two phenomenal
qualities experienced by humans have something phenomenal in
common.
Reflection on some pairs of human experiences does not reveal
them to have anything phenomenal in common.
The nature of a phenomenal quality is revealed by reflection on
phenomenal concepts based on experiences of it. (RT5)
The natures of two things determine whether they have anything
phenomenal in common.
Since this is not an argument against constitutive panpsychism per
se, there are technically five options for constitutive panpsychists in
responding to it: deny one of the premises, or accept the conclusion.
Accepting the conclusion would mean accepting a “large palette” version
of constitutive panpsychism, with all human and animal qualities present
in the base even though that is more than there are distinct physical
roles to play (see,
e.g., Lewtas 2013). The downside is that this sacrifices the
appealing parsimony, and isomorphism with physics, that had seemed to
set constitutive panpsychism apart from traditional sorts of dualism.
Denying premise 3 is also unattractive, since it undermines the case for
panpsychism over physicalism.
Denying premise 4 here (as Lee does, 2019, 300–301)is harder
than denying premise 4 of the macroexperience-focused argument, that
“what constitutes something is part of its nature”. I denied the latter
because knowing a property’s nature is not enough to know necessary
truths about it which involve the nature of another property; we would
have to know that other property’s nature as well. But when it comes to
comparing two qualities that we do experience distinctly, it seems to
follow that we should be able, in principle, to discern every necessary
truth about how those qualities relate, and that should include their
resemblance or common constituents.
We might deny premise 4 in the same way we might deny premise 4 of
the microexperience-focused argument, by saying that although the basic
qualities constitute the macroqualities, they do not characterise them -
the “blending” leaves no trace of the ingredients at all. But this has
the same downsides discussed in the last subsection: if microqualities
in no way characterise the macroqualities, the form of constitution
involved seems mysterious.
That leaves denying premise 1 or premise 2. Premise 1 might seem
undeniable, due to the “interchangeability” of different neurons:
experiences of redness and of mintiness involve neurones made of all the
same sorts of subatomic particles, so how can one contain any ingredient
missing from the other? Any ingredient of the redness experience comes
from electrons, quarks, photons, etc., and those same things are all
present in the physical basis of a mintiness experience, so how could
they not show up in the latter? But this falsely assumes that each
macroexperience should contain every ingredient present in its neural
basis, as though each one were the independent product of one discrete
subset of neurones. It might instead be that several macroexperiences
are all grounded in the activity of the same neurones, being just
different aspects of the complex, differentiated experience produced by
those neurones.
Consider a bar magnet, whose macroscopic behaviour displays a “north
pole” and “south pole”. The north pole does not arise from one half of
the magnet, and the south pole from the other half: both macroscopic
features arise from very same microscopic physical things, because those
things are themselves internally differentiated and their different
aspects add up to what looks, from a macroscopic perspective, like two
different things. It would be a mistake to say “since all the particles
generating the magnet’s north pole also have south poles, why don’t
their south poles show up in the magnet’s north pole?” Perhaps mintiness
and redness are likewise different aspects of the same complex
experience, itself arising from the combination of a great many
internally differentiated microexperiences, combining in different ways
depending on such things as firing rates and degrees of neural
synchrony. Then they might have nothing phenomenal in common, despite
being constituted by the same things.
However, there are limitations to this response. It might allow for a
few completely dissimilar pairs to be compatible with the SPH, but not
that many - if there are a hundred completely dissimilar qualities
experienced by humans, saying that they arise from the way that
internally differentiated aspects of microexperiences are combined
starts to load microexperiences with too much structure for us to retain
the SPH. To keep the palette small, there shouldn’t be too many
completely dissimilar pairs of qualities, which is why this response to
the argument works best when combined with another: denying premise
2.
Denying premise 2 means denying that redness and mintiness have
absolutely nothing at all in common. After all, our ability to recognise
two things as akin to one another is usually enhanced by our ability to
recognise and attend to the features they share, and if we never
experience their shared features in isolation, we may take them to be
entirely unlike even if they are not. Sometimes, of course, two
qualities seem inarticulately alike even without an identifiable shared
feature; this is why we routinely describe qualities of one modality
using terms drawn from another (warm, harsh, sweet, soft, loud, etc.).
The SPH and RT5 can both be retained as long as idealised scrutiny of
these inchoate likenesses would reveal a system of qualitative
connections over our entire experiential range. This view is defended by
Coleman:
[…] just as it’s possible to move across the colour spectrum in tiny,
almost undetectable steps, it must be possible to move from tastes to
sounds, sounds to colors, and so on, via equally tiny steps. Tiptoeing
between modalities already seems conceivable in certain cases,
perhaps even actual. We know that what we experience as “taste” is
really some kind of fusion of qualia sourced from the nose and from the
tongue […]. To address qualitative incommensurability we must stretch to
conceiving of such continuities as the rule rather than the exception.
(Coleman
2017, 264, emphasis in original; cf. Coleman 2015; Hartshorne 1934,
35ff.)
This claim does not seem to me obviously false, but it is at least
dubitable. Consequently, the revelation approach may be most threatening
to constitutive panpsychists not through any of the three pure
revelation arguments, but through intensifying the palette problem.
Accepting revelation pushes constitutive panpsychists towards either a
large-palette solution like Lewtas’s, or towards Coleman’s very bold and
ambitious form of the small-palette hypothesis.
Confusion and Revelation
Identifying a premise of an argument that might be false is often
not, by itself, an effective way to persuade critics. For all that I
have said so far, this “medium-strength” version of revelation,
interpreted and qualified so as to undermine arguments against
panpsychism while substantiating arguments against physicalism, might be
technically consistent but ad hoc and unmotivated, a dingy corner of
logical space which panpsychists can awkwardly retreat to. But in fact,
these qualified revelation theses flow from a reasonable picture of the
limits of human self-knowledge, on which the introspective ignorance
that constitutive panpsychism implies differs only in degree from
familiar forms of introspective ignorance.
It is commonplace to say that when two experiences become
phenomenally unified, they form a composite experience which subsumes
them: they still exist, and are still undergone by the subject, but they
are now “undergone together”. We easily recognise this when we can
discern introspectively not just the composite experience but also its
components: but what if the discernibility of the component experiences
is not an automatic consequence of the composite experience being
composite? We might consider the idea that it depends instead on having
the right structure of informational relations among the components. Perhaps if these relations make the
subject’s overall dynamics differentially sensitive to multiple distinct
features of the experience, the composite experience will be
characterised by contrast among those features: they will stand out as
distinct things. If not, those features will be present in the composite
experience in an undifferentiated way, as a single element whose
phenomenal quality is a seamless blend of its components. In short: the
component experiences all go in together, but the way they are present
in the composite experience depends on how they are organised.
What explains why experiences should compose in this way is a further
question, which I cannot here address (though see Roelofs 2016;
2019, 123–125, 166–170). But suppose some conditional like this
were true: when distinct experiences are unified, they can be
distinguished by the subject only if they have the right informational
structure. Although the human brain is an exquisitely structured
processor of information, it has limits. The overall dynamics of the
brain can perhaps be sensitive to whether a neurone fires, but not (as
far as we know) to which ions in that neurone played which roles in its
firing. Since individual events at the microscopic level are
informationally inaccessible, they will be experienced by the whole in a
blended way. They each make a minute difference to the quality of some
element of the whole’s experience, but they do not stand out as distinct
elements of it. To use a term made famous by Leibniz, they are
“confused” with one another, the way that the sounds of each bit of
water striking the shore are “confused” in the roar of the sea.
I have elsewhere elaborated more fully on the idea of confusion as I
understand it (2019,
126–129), but the essential idea is captured in the following
definition:
Two experiences are confused with each other, relative to a
subject, iff that subject cannot distinguish them by attending to one
without simultaneously attending to the other.
It is important to emphasise that confusion is not a matter of a
subject “perceiving” things outside themselves so poorly that they
cannot distinguish the parts of that outside thing. Confusion is a
matter of how the subject’s own states are related, not a relation
between them and something external. For example, someone viewing a
pointillist painting, for whom the many dots of paint “blur together”,
is not thereby subject to confusion, if they simply have a single
experience that is the product of many external objects. A better
example would be someone with an untrained palate, who drinks coffee and
experiences (let us stipulate) the same diversity of taste and flavour
experiences as a practiced connoisseur but experiences them together as
a single blended flavour, without being able to pick out the bitterness
from the aroma, etc.
Confusion may depend on circumstances. When we are tired, distracted,
or drunk we often cannot distinguish things which we could under better
conditions. Then our experiences are confused only relative to those
circumstances. Confusion can also depend on a subject’s conceptual
repertoire: sometimes we cannot distinguish two things using their
present concepts, but would be able to if we learnt new ones. Call
confusion which can be removed by adjusting the subject’s bodily
surroundings or condition, or improving their conceptual repertoire, or
in some similarly mild way, “shallow confusion”, and call confusion
which persists even into ideal conditions, “robust confusion”.
In between shallow and robust is confusion which persists until the
subject becomes distinctly acquainted with a token of the same type as
the confused elements. For example, suppose the sensory component of
pain is robustly confused with the unpleasant affect pain involves,
except for subjects who have experienced “pain asymbolia”, the rare
condition of feeling pain without finding it at all unpleasant (cf. Grahek
2007; Klein 2015). If they regain normal pain experiences, they
might find themselves newly able to attend to its sensory element in
isolation. If this were to happen, we might say that their original
confusion was “nearly-robust”: removable only by somehow acquainting
them with (a token of the same type as) one of the confused elements on
its own.
When confusion is shallow, we have an easy way to tell that we suffer
from it: we remove it and contrast the resulting distinction with the
earlier confusion. With sufficiently robust confusion, we would not have
such means of recognising it; we could not tell that we were confused.
And if we suffered from confusion that was “nearly-robust”, it would be
undetectable, except by means of independent acquaintance with elements
of the same type as the confused ones. We could, that is, be subject to
a lot of confusion without being able to tell, introspectively. And if
constitutive panpsychism is true - in particular, if micro-experiences
corresponding to all the physical details of our brains were somehow
present in our consciousness - then we should expect just that: all the
experiences of our microparts would be confused relative to us. Call
this the Radical Confusion Hypothesis.
Confusion is defined functionally, but that does not imply that
confusion is a purely functional fact that makes no phenomenal
difference. My suggestion is that undergoing two confused experiences
feels different to undergoing two distinguishable experiences, even if
those experiences are the same in all intrinsic respects. When the
components of an experience are distinguishable by the subject, they are
phenomenally present as discernible, separate, parts - there is an
experience of phenomenal contrast, of things standing out against other
things. But when they are confused, they are present qualitatively, as
contributions to the total quality of the experience they blend
into.
How would the Radical Confusion Hypothesis help with the four
revelation arguments? Recall that in response to the “no illusions”
argument, I denied premise 2: that human consciousness positively
appears introspectively to be “smooth” (there defined as “not
particulate”). I maintained that this is false if “appears
introspectively” is read in a strong, quasi-perceptual sense; it is true
only if “appears introspectively” is read in a weaker sense, as meaning
“it is easy and natural to interpret experience this way”.
Now I can say why this misinterpretation is easy and
natural: because many human experiences display something close to
“smoothness”, namely, all their component experiences are
nearly-robustly confused with each other, distinguishable only by a
subject who already knows what to look for. A subject who lacks any
distinct acquaintance with the ingredients will be unable to distinguish
them or discern their internal structure. We might say that experiences
all of whose components are confused with one another are
“pseudo-smooth”, and it is true (and introspectively obvious!) that many
of our experiences are pseudo-smooth. But to infer genuine smoothness
from pseudo-smoothness is a metaphysical over-interpretation which goes
beyond the introspective deliverances: it is inferring absence of
structure from the failure of structure to be manifest in a certain way
(it is thus very similar to the “headless woman illusion” discussed by
Armstrong
(1968), where not seeing someone’s head gives us the vivid but
false impression that they have no head). The noticeable quality that
some experiences have, which prompted the “no illusions” argument, is
just what radical confusion feels like.
Second, in response to the macroexperience-focused argument I denied
premise 4, that whatever constitutes something is part of the “nature”
that is revealed to us by pure phenomenal concepts. I suggested that a
priori reflection tells us only those necessary truths that involve
only properties whose nature we know - such as whether one
could constitute the other. But just knowing the nature of one property
does not tell all the things that could constitute it, nor what
constitutes a particular instance of it.
I can now elaborate on this distancing of constitution from “nature”.
Macroexperiences are composite experiences composed of many
microexperiences confused with one another. Their phenomenal character
is determined by combining the phenomenal characters of those component
experiences, which they subsume in fundamentally the same way that a
person’s total experience at any one time subsumes the partial
experiences they are having at that time. But just as two composites
might end up sharing certain properties despite being constituted by
different sets of parts, and despite their properties being mere
combinations of the properties of their parts, two composite experiences
might have the same phenomenal character, despite being constituted by
different sets of microexperiences. The particular parts might be
essential to the particular macroexperience, but not to the property
that it is an instance of.
I also said, in response to the small-palette revelation argument,
that distinct macroexperiences might arise from the same neural basis:
we need not assume that each distinguishable element of our
consciousness contains the entire phenomenal nature of one discrete
subset of physical entities. The radical confusion hypothesis reinforces
this point: it says that which experiences phenomenally contrast or
phenomenally blend with one another in human experience reflects the
informational structure of the brain, so a single macroexperience might
not correspond to any discrete section of the underlying physical
substrate. Instead, it will correspond to a set of features of the
substrate such that information about them collectively is extracted and
used by the brain, but information about them individually is not. Thus
different macroexperiences based in the same brain area might have
different, even non-overlapping sets of phenomenal ingredients, because
they reflect different features of the same microexperiences.
Finally, in response to the microexperience-focused argument I
suggested that our ignorance of microexperiences is compatible with our
undergoing them, if we cannot attend to them. Now I can add that our
inability to attend to microexperiences is part-and-parcel of their
being confused for us. Their radical confusion is explained by the
limitations discussed above on how much information about microscopic
brain events can be extracted by the rest of the brain.
Because radically confused experiences cannot be distinctly attended to,
we cannot know them or their natures, even though the experiences
“present themselves” in the sense that if their subject could attend to
them they could know them and their natures by introspection.
An opponent might object that even though attending to particular
experiences can be harder or easier, depending on, e.g., architectural
facts about the brain, it cannot be strictly impossible for me
to attend to an experience, if it is really is an experience I am
undergoing. I reply that distinctly attending to microexperiences is
not strictly impossible, just impossible in practice (as discussed in Lee 2019,
296–297). They are manifest in our consciousness, but incredibly
difficult to pick out. After all, it is very difficult for the
large-scale dynamics of our brain to be sensitive to changes in a single
particle, but there is no in-principle impossibility in there being such
sensitivity, perhaps using advanced technology or strange altered states
of consciousness.
Conclusions
The idea of “revelation”, that having an experience provides a
special insight into its nature, is a key weapon in the armoury of
anti-physicalists. But for constitutive panpsychists there is a risk it
will blow up in their faces. I have argued, however, that a
suitably-qualified form of the revelation approach can bring down
physicalism while leaving panpsychism standing: a form which reconciles
the profound fallibility of the human mind’s self-knowledge with the
perfect transparency of its access to its itself. Although nothing does
or could “conceal” our own experiences from us, we are nevertheless
limited in our ability to attend to their elements, prone to
misinterpret them, and consequently unable to tell introspectively just
how composite they might really be.