Reduction and Elimination in Philosophy and the Sciences

A Critique of the Phenomenal Concept Strategy
Daniel Lim, Cambridge, England, UK
The Strategy
The locus classicus of the phenomenal concept strategy
is Brian Loar’s paper ‘Phenomenal States’ (1990). In it
he claims that the Knowledge Argument relies on a dubious
assumption which he dubs the Semantic Premise:
204
“A statement of property identity that links conceptually
independent concepts is true only if at least
one concept picks out the property it refers to by
connoting a contingent property of that property.”
(Loar 2004, 224)
He argues that the Semantic Premise, while true in
standard cases of a posteriori identities like ‘water is
H20’, is crucially false in cases of a posteriori psychophysical
identities. This is because the phenomenal concepts
deployed in psychophysical identity statements are
really type-demonstrative concepts of the form: ‘this
experience’. The two features of type-demonstratives
relevant for the strategy are: direct reference and conceptual
independence from physical / scientific concepts.
The first feature ensures that no new properties are introduced
and the second feature ensures that the identity
remains a posteriori.
The phenomenal concept strategy has the
makings of a powerful response to the Knowledge
Argument and defenders of the strategy claim that it
satisfies three important desiderata: (i) it respects the
kind of knowledge Mary gains after leaving her black and
white room, (ii) it is physically explicable, and (iii) it
explains why we cannot resist the illusion of ontological
distinctness concerning our conscious experiences.
Phenomenal Knowledge and
Physical Explicability
While there are independent reasons 1 for eschewing a
type-demonstrative construal of phenomenal concepts I
wish to dwell on a dilemma David Chalmers (2007) has
been keen on exposing which brings features (i) and (ii)
into tension. If phenomenal concepts are physically explicable
they will not explain our epistemic situation. On
the other hand, if phenomenal concepts can explain our
epistemic situation then they will be physically inexplicable.
Chalmers focuses his critique on issues of conceivability,
but I will focus on the putative physical mechanisms
that make phenomenal concepts possible. A nice
way of bringing out the tension in this dilemma is to use
Janet Levin’s and David Papineau’s positions as exemplars
of each horn.
Levin’s position exemplifies the first horn. She
deliberately avoids any account of phenomenal concepts
that require anything that is physically suspect:
‘quotation’, ‘partial constitution’, or ‘acquaintance’. She
opts for a physically safe version of phenomenal
concepts that is limited to “causation, reliable correlation,
and relations of physical inclusion or adjacency”. As
1 For example see Diana Raffman’s (1995) critique of type-demonstratives
based on the empirical fact that we can discriminate more colors than we
can re-identify over time.
such she argues that phenomenal concepts should be
construed as introspectively deployed demonstratives
and nothing more. All that is needed to distinguish
introspectively deployed phenomenal demonstratives
from nonphenomenal ones are ‘differences in what they
[causally] denote’. This is worrisome because it drives,
what seems to be, an unacceptable wedge between
phenomenal concepts and the properties they denote.
By relating the two by causation too much distance has
been allowed to creep into the picture. In David
Chalmers’ terminology, this makes phenomenal
concepts, in a sense, ‘Twin Earthable’. The referents of
Twin Earthable concepts will be unstable across
counterfactual worlds. When Twin Oscar thinks, while on
Twin Earth, that the substance in the lake looks
refreshing, his thoughts about the substance will be twin
water thoughts. This is because XYZ, and not H20,
causes Twin Oscar’s concept of ‘water’ to be tokened.
When Oscar thinks, while on Earth, that the substance in
the lake looks refreshing, he will entertain water thoughts
because they are caused by H20.
To apply this insight to differentially caused typedemonstratives,
we can imagine a scenario where I am
observing a lake. The lake is partitioned into sections.
Scientists, who have managed to transport a sizeable
amount of XYZ from Twin Earth, have filled some
partitions with XYZ and others with H20. While looking at
partition A I deploy a type-demonstrative ‘that liquid’ and
while simultaneously looking at partition B I deploy, what
I think is, the same type-demonstrative ‘that liquid’ and
think in my mind: ‘that = that’. I am wrong about this
since A is filled with XYZ and B is filled with H20.
However, I am wrong not because I misapplied one of
the demonstratives, but because I unknowingly deployed
two different concepts. This is because my concepts are
causally individuated by the objects they denote. While
attending to the liquid in A I may have thought that I was
deploying a water concept, when in fact I was deploying
a twin water concept.
Applying this scenario to our own phenomenal
states 2 it is possible for a normal subject to deploy a
type-demonstrative ‘that experience’ while attending to
the same phenomenal property twice in quick
succession and yet have room to rationally doubt
whether ‘that = that’ is in fact true. Levin writes:
“…it may seem epistemically odd that introspecting
subjects can be mistaken about whether
they’re using the same concepts in their thoughts
about their own phenomenal states. But when
concept difference and identity are determined
‘externally’ – that is, by the features of what’s denoted
– this shouldn’t be unexpected, even when
the subject matter is one’s own mental states.”
(Levin 2007, 108)
This makes the following scenario possible: I may mistakenly
think that I’m deploying a phenomenal concept
when in fact I am not. Let’s say that brain state p1 is my
2 This is a thought-experiment John Hawthorne (2007) develops against
direct reference theories of phenomenal concepts used to defend property
dualism.