Non-reductive Realization and
the Powers-based Subset Strategy∗
Jessica Wilson†
Draft: May 22, 2010
1
Introduction
1.1
The pretheoretic appearances
Pretheoretically, so to speak, “higher-level” special science particulars and their
features (including properties, states, associated events, relations and behaviors)
appear to be dependent on, yet to some extent autonomous from, “lower-level”,
and ultimately fundamental physical, particulars and their features.1 Somewhat
more specifically, special science particulars appear to be materially composed
by lower-level particulars; and features of special science particulars appear to be
functionally dependent on features of their composing lower-level particulars—
in that, at a minimum, changes in the former features require changes in the
latter, and the latter necessitate the former with at least nomological necessity.
Yet special science particulars also appear to be to some extent ontologically
and causally autonomous from—that is, distinct from and distinctively causally
∗
Please note that this is work in progress. Many thanks for comments to Ranpal Dosanjh
and Benj Hellie; to Alyssa Ney (commentator), Derk Pereboom (fellow panelist), and audience
members at the Metaphysics and Mind symposium (2009 Pacific APA); and to the excellent
participants of the New Ontology of the Mental Causation Debate conference (Durham 2009).
†
Department of Philosophy, University of Toronto; [email protected]
1
Talk of “higher” and “lower” levels is potentially misleading in certain respects, but is
desirably efficient. To preempt confusion: in the first instance, I use these terms to track, in a
relative and prima facie way, composed and composing entities, and their associated sciences;
e.g., a molecule is “higher-level” relative to its composing atoms, as is molecular physics relative
to atomic physics. In the second instance, I also deem “lower-level” entities that are lowerlevel relational, conjunctive, disjunctive, or mereological combinations of lower-level entities,
even though these more complex entities might not be among the proper subject matter of the
associated lower-level science; hence a relational entity consisting of atoms standing in atomic
relation is again “lower-level” relative to a molecule. Of course, one might argue that molecules
are identical to relational atomic entities (or to disjunctions of such entities, etc.), which is why
the higher-level/lower-level distinction is just prima facie.
1
efficacious with respect to—the lower-level entities upon which they depend: as
per their governing laws, they have seemingly distinctive features and enter into
seemingly distinctive causal interactions.2
What metaphysical dependence relations between higher- and lower-level entities underlie these pretheoretic appearances? Physicalists disagree about the
nature of these relations (and of course, a given physicalist will likely suppose
that multiple dependence relations may be at issue, reflecting different sorts of
relata3 ). They agree, however, that these relations conform to physicalism:
Physicalism: All broadly scientific entities are nothing over and above
physical entities.
Here the “physical” entities are members of a relatively restricted set, usually
associated with fundamental physics.4 When some entities are (or are understood
to be) “nothing over and above” physical entities, in some or other sense of the
expression, I will say that the entities are “physically acceptable”.
Non-reductive physicalists (NRPists) moreover aim to vindicate the pretheoretic appearance that dependence is compatible with both ontological and causal
autonomy, as per:
Non-reduction: Some special science entities are both distinct from
and distinctively causally efficacious with respect to any lower-level
physically acceptable entities.5
2
Kim (1993a) associates these appearances with a commonly presupposed background picture consisting of “a hierarchically stratified structure of ‘levels’ or ‘orders’ of entities and their
characteristic properties. It is generally thought that there is a bottom level, one consisting of
whatever micro-physics is going to tell us are the most basic physical particles out of which all
matter is composed [. . . ]. As we ascend to higher levels, we find structures that are made up of
entities belonging to the lower levels, and, moreover, the entities at any given level are thought
to be characterized by a set of properties distinctive of that level”. Along with the caveats
about levels observed in the previous note, there is room to differ about the details (perhaps,
e.g., there is no bottom level); but in any case, metaphysicians of science usually start with
something like this picture, and go forward.
3
So, for example, there are likely dependence relations between lower-level non-relational
features and higher-level features (e.g., the relation holding between charge properties of electrons and protons and bonding properties of atoms made up of these constituents), which
appear to be instantiated in different particulars, and dependence relations between lower-level
relational features and higher-level features (e.g., the relation between structural molecular
properties and fragility properties, or the relation between brain states and mental states),
which may or may not relate features of a single particular.
4
Physicalists also differ, though not as much, concerning how to characterize the physical.
My preferred account, which I think conforms to the intentions of the majority of physicalists,
looks primarily to (appropriately accurate) physics for the extension of the physical, with
the crucial proviso that the physical entities are not fundamentally mental (that is, do not
themselves have or bestow mentality); see Wilson 2006 for details.
5
The portion of the NRPist’s thesis corresponding to Non-reduction is usually characterized
in terms just of ontological autonomy (non-identity) of higher-level entities; however, since the
2
Note that, as per the goal of vindicating the pretheoretical appearances, Nonreduction is characterized in ontological/metaphysical terms.6 So, for example,
and making explicit the assumptions (as per note 1) concerning which entities
count as lower-level relative to a given special science entity: if a given special
science entity S is composed of physical entities, then the claim that S satisfies
Non-reduction entails that S is not identical to any physical entities, nor to any
relational entity consisting of physical entities standing in physical relation, nor
to any boolean or mereological combination of such relational or non-relational
entities.
Summing up: NRPists aim to show how at least some dependent higher-level
entities that are physically acceptable, in being, one way or another, “nothing over and above” the lower-level physically acceptable (“base”) entities upon
which they depend (thus satisfying Physicalism), are also ontologically and causally
autonomous from any such base entities (thus satisfying Non-reduction).
1.2
Non-reductive realization
As noted, different relations may be relevant to a full reckoning of how special
science entities depend on (ultimately fundamental physical) lower-level entities. A common approach to an important part of this project is focused on
how certain special science particulars depend upon certain relational lower-level
particulars, where the latter are standardly assumed by all parties to be physically acceptable.7 I will make that assumption and associated classification here.
The question is then: what are the dependence relations between special science
particulars and the physically acceptable lower-level relational particulars upon
which they depend? And the larger part of interest in an answer is the question:
point of establishing non-identity is to provide a basis for causal autonomy, and because (as
we’ll see) non-identity is compatible with a failure of causal autonomy, the latter requirement
should be explicitly built into Non-reduction.
6
Non-reductive physicalism has also been formulated in epistemological terms, as directed
at the question of whether all broadly scientific entities are deducible from or otherwise explainable in terms of physical entities. This is, for example, the question that most concerns
Melnyk (2008), who argues that no viable formulations of physicalism are epistemologically
non-reductive, while leaving open that some physicalist accounts (including the version developed in Melnyk 2003) are ontologically reductive, in the sense at issue in Non-reduction.
7
If one is concerned about the assumption that certain relational physically acceptable particulars exist (on grounds that, for example, any robustly emergent or otherwise physically
unacceptable features of a special science entity S would also be features of any lower-level
relational entity upon which S depends), one may treat the assumption as a convenient heuristic, serving roughly to mark the class of aggregate particulars that reductive physicalists think
exist from the (larger) classes that non-reductive physicalists and robust emergentists think
exist. I say “roughly” since as above, reductive physicalists may also admit as part of their
candidate reduction base boolean or mereological combinations of particulars or other entities.
3
are these dependence relations such as to preserve the ontological and causal
autonomy of special science particulars?
Given Leibniz’s law, according to which the identity of particulars requires
sharing of all their features, and given, as is plausible and fairly uniformly accepted, that what a particular can do (cause) is a matter of of how it is (that
is, of what features it has), the latter question leads naturally to our target
cases, where certain features of special science particulars depend upon certain
(assumed physically acceptable) features of lower-level relational particulars, as
when . . .
• . . . having a certain temperature (understood as a property or state of a
gas) depends on having a mean molecular kinetic energy (understood as a
physically acceptable property or state of a relational entity consisting in
some molecules standing in certain molecular relations)
• . . . being phototropic (understood as a property or state of a plant) depends
on undergoing certain cellular wall weakenings and cellular expansions (understood as a physically acceptable property or state of a relational entity
consisting in certain cells of the plant standing in certain cellular relations)
And nearest and dearest to our hearts and minds:
• . . . being in a certain mental state (understood as a property or state of a
person) depends on being in a certain brain state (understood as a physically acceptable property or state of a relational entity consisting in certain
neurons standing in certain neuronal relations).
Call the “nothing over and above” relation the physicalist takes to be at issue in
the target cases “realization”.8 And call the “nothing over and above” relation
the non-reductive physicalist takes to be at issue in the target cases, which relation is supposed to compatible with the ontological and causal autonomy of the
special science entities at issue, “non-reductive realization”.
8
The term ‘realization’ has traditionally and standardly been associated with the relation
at issue in cases like the target cases. Gillett (2002) argues that an adequate account of ‘realization’ should apply not just to such cases, characteristic of what he calls the “flat” view of
realization (though unlike me he assumes that such cases always involve same-subject necessitation), but also to cases of what he calls “dimensioned” realization, involving dependence
relations between lower-level non-relational goings-on and higher-level goings-on (see note 3);
and he offers an account of realization based in the “in virtue of” relation that is supposed
to apply to both sorts of cases. Terminological decisions aside, I follow Polger and Shapiro
(2008) in thinking that such a broad-brush approach to realization obscures the underlying
metaphysical facts, which are likely different in different cases of dependence; see note 48 for
further discussion.
4
1.3
The topic
I can now state the topic of my paper. At a general level, my topic concerns the
question: How should non-reductive realization be understood? More specifically,
I will argue that an adequate account of non-reductive realization must guarantee
the satisfaction of a certain condition on the token causal powers associated
with (instances of) realized and realizing entities—namely, what I call Subset
Condition on Causal Powers. In terms of states (which I treat here as, roughly,
property instances9 ), the condition requires that the token powers had by a
realized state on a given occasion be a proper subset of the token powers had by
the state that realizes it on that occasion. Accounts of non-reductive realization
conforming to this condition are implementing what I call the powers-based subset
strategy.
I will focus on the crucial case involving mental and brain states; the results may be generalized, as appropriate. I will first situate and motivate the
strategy by attention to the problem of mental causation; I will make the case,
in schematic terms, that implementation of the strategy makes room (contra
Kim 1989, 1993a, 1998, and elsewhere) for mental states to be ontologically
and causally autonomous from their realizing physical states, without inducing problematic causal overdetermination and compatible with both Physicalism
and Non-reduction; and I will show that several contemporary accounts of nonreductive realization (in terms of functional realization, mereological overlap, and
the determinable/determinate relation) are plausibly seen as implementing the
strategy. As we’ll see in the course of this discussion, implementation of the
powers-based strategy does not require that one endorse any particular account
of properties or of causation—indeed, as I’ll show, a categoricalist contingentist
Humean could implement the strategy.
The schematic location of the powers-based subset strategy in the space of
available responses to the problem of mental (more generally, higher-level) causation, as well as the fact that the schema may be metaphysically instantiated,
strongly suggests that the strategy is (when appropriately generalized and instantiated) sufficient and moreover, necessary (in being, in particular, the only
route to sufficiency), for non-reductive realization. I go on to defend the sufficiency and necessity claims against a variety of objections, considering, along the
9
States are sometimes thought of as involving reference to the particulars bearing the state
(as in states of affairs). I presume a view of states not doing so both for simplicity, and in order
to sideline the further question, to which investigation of the target cases is ultimately directed,
of whether the special science and lower-level relational particulars that bear the states at issue
are identical. See note 39 for further discussion.
5
way, how the powers-based subset strategy fares against competing accounts of
(purportedly) non-reductive realization in terms of supervenience, token identity,
and constitution.
2
The powers-based subset strategy
2.1
The problem of mental causation (MC)
The problem of mental causation is squarely directed at our crucial target case,
and at the NRPist’s hopes for treating this case in line with the pretheoretic
appearances. One starts with the question: How can mental states be causally
efficacious, as they seem to be, given their strong dependence on brain states?
The problem is then that certain commonly accepted premises about mental
states and about causation seem to give rise to an unsatisfactory answer to the
question.
The premises about mental states are as follows:
1. Mental states depend on lower-level physically acceptable relational states
(henceforth: ‘base states), in that, at a minimum, mental states (at least
nomologically) require and are (at least nomologically) necessitated by base
states.
2. Mental states and base states are real.
3. Mental states are causally efficacious.
4. Mental states are distinct from base states.
The premises about causation are as follows:
5. Every physically acceptable effect has a purely physically acceptable cause.
6. In general, effects are not causally overdetermined.
Now for the problem. First consider a case where mental state M depends, on
a given occasion, on base state P (as per 1), and where M causes special science
state M * on a given occasion (compatible with 3). M * is dependent on some
base state P * (by 1), such that P * necessitates M *, with at least nomological
necessity. Moreover, P * has a purely lower-level physically acceptable cause (by
6)—plausibly, and without loss of generality, P . If P causes P *, and P * (at
6
least nomologically) necessitates M *, then it is plausible that P causes M *, by
causing P *. So, it appears, both P and M cause M *, and given that P and M
are both real and distinct (by 2 and 4), M * is causally overdetermined, contra
(5).
Next consider a case where mental state M depends, on a given occasion,
on base feature P (as per 1), and where M causes base feature P * on a given
occasion (compatible with 3). P * has a purely physically acceptable cause (by
6)—plausibly, and without loss of generality, P . So, it appears, both P and M
cause P *, and given that P and M are both real and distinct (by 2 and 4), P *
is causally overdetermined, contra (5).
How to respond? Responses correspond to the rejection of various premises:
• Substance dualism. Deny (1): avoid overdetermination by denying that S
depends on physically acceptable P.
• Eliminativism. Deny (2): avoid overdetermination by denying that either
M or M * is real.
• Epiphenomenalism. Deny (3): avoid overdetermination by denying that M
is efficacious.
• Reductive Physicalism. Deny (4): avoid overdetermination by denying that
M is distinct from P.
• Robust emergentism. Deny (5): avoid overdetermination by denying that
every physically acceptable effect has a purely physically acceptable cause.
And finally, the NRPist’s response:
• Non-reductive Physicalism. Deny (or disambiguate) (6): allow that there
is overdetermination in this case, but deny that it is of the problematic
“firing squad” variety.
How can the NRPist make out that there is no problematic overdetermination
in the core case? NRPists typically offer up some specific metaphysical account
of the non-reductive realization relation (holding either between the states or
their associated types10 ), illustrating how this can be done. So, for example, it
has been suggested that the relation at issue is:
10
NRPists differ concerning whether realization is, in the first instance, a relation between
types or tokens. Polger and Shapiro (2008) claim that accommodating multiple realizability
requires that realization be a relation between types, reflecting that tokens (states, property
instances, etc.) are generally not multiply realizable in the intended sense; but this is incorrect,
7
• Functional realization (Putnam 1967, Fodor 1974, Papineau 1993, Antony
and Levine 1997, Melnyk 200311 ). Mental state types are functional state
types implemented by physically acceptable state types.
• Mereological realization (Shoemaker 2001, Clapp 2001, Paul 2002). Mental
states/state types are parts of base states/state types.
• The determinable-determinate relation (Yablo 1992, Wilson 200912 ). Mental states/state types are determinables of bases states/state types.
The proponent of a given account then attempts to argue that the relation at
issue is such that, were it to hold between mental and physically acceptable
states, problematic overdetermination would be avoided, compatible with both
Physicalism and Non-reduction.
2.2
Subset Condition on Causal Powers (SCCP)
There is a lot to say about each of these non-reductive realization relations, and
about whether and how they make sense as applied to the mental/physical case.
For present purposes I want to call attention to a more abstract, though still
properly metaphysical, condition that each of these relations (properly understood) arguably satisfies, and which is at the heart of the seeming promise of
these relations as entering into an account of non-reductive realization. I will
first state the condition and lay out, in schematic terms, how its satisfaction
avoids problematic causal overdetermination in the target case, compatible with
Physicalism and Non-reduction. I will then argue that each of these relations,
properly understood, satisfies the condition.
for the truth of multiple realizability claims can be accommodated even if realization is a
relation between tokens, as grounded in the types’ potentially having differently realized tokens.
Even if a given account of realization takes this to be, in the first instance, a relation between
types, the account must ensure that tokens of the types are appropriately intimately related;
and no harm comes of referring to this token-level relation as realization (for purposes, in
particular, of expressing the condition to be shortly discussed).
11
I include Melnyk here as endorsing a developed account of functional realization that
may, for reasons outlined below, serve as a basis for NRP, notwithstanding that Melnyk is
officially neutral on whether implementation of his functional account of realization guarantees
ontological autonomy (distinctness) of realized entities (as per Non-reduction), and so doesn’t
self-identify as an NRPist.
12
I am officially neutral on whether NRPism is the correct view about mental and physical states, since I think robust emergentism remains a live empirical possibility (see Wilson
2002). Of the NRPist options, however, I think the determinable/determinate relation is most
promising, for reasons I set out in my (2009).
8
2.2.1
Causal powers
The condition is a condition on causal powers (henceforth: just “powers”). What
is a power? To start, recall that (it is plausible and fairly uniformly accepted) that
what a particular can do (cause) depends on how it is—that is, on what features
it has. So, for example, it is in virtue of being massy, not magnetic, that a magnet
falls to earth, and it is in virtue of being magnetic, not massy, that a magnet
attracts steel. Reflecting that the potential production of certain effects depend
upon the possession of certain properties, we may say that the features in question
‘bestow’ or ‘have’ the powers to produce the effects in question. Moreover, as it
turns out, a given feature F will typically bestow or have powers to produce a
variety of effects, depending on what internal and external circumstances it may
occur or be instantiated in. Hence (to use Shoemaker’s example) the property
of being knife-shaped has the power to cut butter, in (internal) circumstances of
being co-instanced with the property of being wood and (external) circumstances
of being instanced proximal to some butter; and has the power to cut wood, in
(internal) circumstances of being co-instanced with the property of being steel
and (external) circumstances of being instanced proximal to some wood.
Philosophers differ over the nature of powers, their role in causation, and
whether they are essential to the features that have them. More specifically,
philosophers disagree about whether powers are dispositions, categorical features,
posits following upon “best system” regularities, and so on. Those differing about
the nature of causation may differ over whether powers are had intrinsically or
extrinsically, and so on. And philosophers differ about whether powers are essentially individuative of features, or whether the powers of a given feature can
vary, depending on what laws of nature are operative. Perhaps surprisingly, none
of these disagreements plays a role in what follows, with a small exception associated with motivations for supervenience-based accounts of realization, which I
will address down the line. This neutrality reflects that all participants to the
debate need to be able to track the potential causal contributions associated with
features of broadly scientific entities (that is, “powers”) some way or another,
and so long as the operative account of powers can make sense of one (token)
contribution’s being identical with another, the account will serve for present
purposes. Indeed, as we will see, even a contingentist categoricalist Humean can
implement the powers-based subset strategy.
9
2.2.2
Condition on Causal Powers
To start, each of the aforementioned NRPist realization relations (functional,
mereological, or determinable/determinate) arguably guarantees satisfaction of:
Condition on Causal Powers (CCP): Every power had by a mental
state M on a given occasion is numerically identical with a power had
by the relational physically acceptable (base) state P realizing M on
that occasion.13
Here and throughout, talk of ‘powers had by a state’ is to be understood as talk of
token powers had by the (token) state. As per the rejection of epiphenomenalism,
CCP is assumed to be non-vacuously satisfied. Note that reductive physicalist
accounts also guarantee satisfaction of CCP; for example, if mental state types
are identical to physical state types, then mental states are identical to physical
states, and CCP is satisfied.14
2.2.3
Subset Condition on Causal Powers
Moreover, each of the aforementioned relations arguably takes CCP to be satisfied
in such a way that the set of powers had by a realized state on a given occasion
is a proper subset of those had by its realizing state on that occasion:
Subset Condition on Causal Powers (SCCP): The powers of a mental
state M on a given occasion are a (non-empty) proper subset of the
powers of the physically acceptable state P that realizes M on that
occasion.15
(Again, keep in mind that the powers of states here are token powers.)
13
I include ‘on that occasion’ to make room for the possibility that a given (token) mental
state might be realized by different physical states on different occasions.
14
CCP is effectively Kim’s “causal inheritance principle”: “If mental property M is realized
in a system at t in virtue of physical realization base P , the causal powers of this instance
of M are identical with the causal powers of P ” (1992, p. 18). Kim goes on to claim that
rejecting this principle is tantamount to accepting “causal powers that magically emerge at
a higher level and of which there is no accounting in terms of lower-level causal powers and
nomic connections”. While robust emergence need not involve “magic” (see McLaughlin 1992
and Wilson 2002 and (in progressb)) it’s no surprise that physicalists typically endorse CCP,
since its denial is arguably characteristic of robust emergentism and other views incompatible
with Physicalism. I will revisit CCP’s connection to Physicalism in §4.3.
15
Note that neither CCP nor SCCP requires that the powers of a mental state type be a
proper subset of the powers of each of the types of its realizers. Of course, when the proper
subset condition holds at the level of types, it will plausibly hold at the level of tokens (again,
compare the case of identity). But the proper subset condition might not hold at the level
of types, and it still be the case that it holds at the level of tokens, on given occasions of
realization; this is the moral of Clarke 1999.
10
Let’s see, in schematic terms, how satisfaction of SCCP in our target case
avoids overdetermination while satisfying both Physicalism and Non-reduction.16
To start, a realization relation satisfying SCCP avoids problematic overdetermination. If the relation between states M and P satisfies SCCP (hence CCP),
then every power of M is numerically identical to a power of P, on that occasion.
Hence for a given effect produced by both M and P on a given occasion, there is
only one manifestation of the associated power on that occasion. There is only
one causing, not two.
Next, a realization relation satisfying SCCP satisfies Physicalism. The main
concern about physical acceptability turns on the possibility that a realized entity
might be robustly emergent. Recall that on a robust emergentist treatment,
M and P are each real, distinct, and causally efficacious; overdetermination is
avoided by denying that P causes E, or that P causes E in the same (direct)
way as M . In terms of powers: the robust emergentist maintains either (a) that
P does not have the power to cause E, or (b) that while P does have the power
to cause E, this power is not identical with that had by M (it is manifested
differently, or in different conditions; perhaps, e.g., P causes M , and M then
more directly causes E). Either way, CCP, hence SCCP, fail to be satisfied.
Satisfaction of SCCP also blocks the other live routes to physical unacceptability, associated with M ’s being non-natural (see Moore 1903) or supernatural.
To start, to be “non-natural” is not obviously to be physically unacceptable; as
Moore seems to have used the term it was mainly indicative of epistemological
irreducibility (in particular, indefinability), which is arguably compatible with
at least some forms of NRP.17 In any case, if M ’s being undefinable or otherwise
epistemologically irreducible is to pose a problem for the physicalist, this must
be because it is indicative that M ’s existence involves something metaphysically
new relative to (“over and above”) P .18 Plausibly, the non-natural novelty of a
mental state would be a matter of the state’s having a non-natural causal power,
not had by physically acceptable P . The possession of such a power is, however, ruled out if M has no powers distinct from those of its realizing physically
acceptable state P .
16
Note that the NRPist is only committed to some, not all, cases of realization’s being
non-reductive.
17
See Wilson 2002 for such an argument.
18
Similarly for accounts of robust emergence, which have sometimes been presented in epistemological terms, as involving, e.g. “in-principle” failure of deducibility. As McLaughlin
1992 notes, “[T]he Emergentists do not maintain that something is an emergent because it is
unpredictable. Rather, they maintain that something can be unpredictable because it is an
emergent” (p. 73)—that is, metaphysically ‘over and above’ what it emerges from.
11
Now, as it stands (and remaining broadly neutral on the metaphysics of
states) satisfaction of SCCP is compatible with M ’s having a non-causal aspect not had by P —say, a non-causal quiddity or epiphenomenal quale. But, as
is also reflected in the dispute between robust emergentists and physicalists, any
non-causal aspects of M are irrelevant to broadly scientific goings-on, since scientific truths involving features such as M —and indeed, scientific truths involving
any broadly scientific features—do not in any way depend on or otherwise track
whether the features at issue have non-causal aspects (much less track how noncausal aspects of seemingly distinct features are related). Hence that M has such
aspects (whether or not shared by P ) cannot undermine M ’s physical acceptability, given P ’s physical acceptability.19 Similar remarks apply to the possibility
that M is supernatural, in being the result of Malebranchean occasionalism. The
general pattern, blocking any route to M ’s physical unacceptability, is as follows:
if P is physically acceptable, and every power of M is identical with a power of
P , then any causal aspects of M are guaranteed to be physically acceptable;
non-causal aspects of M are irrelevant to M ’s physical acceptability; hence a
realization relation satisfying SCCP guarantees M ’s physical acceptability, in
conformity to Physicalism.20
19
It is also worth noting that any such non-causal aspects would also serve, at best, as a
basis for the ontological distinctness of M from P , and could not ground the causal autonomy
at issue in Non-reduction.
20
The previous considerations indicate, pace Melnyk 2006, pp. 141–143, that it is not necessary, for satisfaction of SCCP to guarantee physical acceptability of a realized state, that the
state (or state type) be identified with the collection of powers of the state (state type). Melnyk
effectively suggests that unless states are identified with clusters of token powers, satisfaction
of SCCP will not guarantee satisfaction of the “constitution” and “truthmaking” conditions,
according to which physically realized entities are constituted by their physical entities, and
truths about physically realized entities are made true by physical goings-on, since if realized
states have non-causal aspects, then even though a’s being P entails that a has (bestowed
upon it) the token powers associated with being in a state of M ’s type, it won’t follow from
satisfaction of SCCP that a’s being P constitutes a’s being M , or that a’s being P (along
with physical laws, etc.) makes true attributions of M to a. However, truths about physical
constitution or truthmaking, being broadly scientific truths, are neutral as regards whatever
non-causal aspects of features there might be. So consideration of the constitution and truthmaking conditions does not indicate that implementation of the powers-based subset strategy
requires acceptance of a causal account of properties. Similar considerations undermine another
reason Melnyk gives (pp. 138–140) for thinking that satisfaction of SCCP alone is insufficient
for Physicalism; namely, that unless realized entities are identified with clusters of powers,
then satisfaction of SCCP will not guarantee satisfaction of the “necessitation” condition, according to which a physically acceptable realized entity must (perhaps together with physical
laws, etc.) metaphysically necessitate the realized entity: “Why should it? Why assume that
along with possession of power-tokens of certain types there automatically comes possession of
a property [. . . ] that would have conferred them?” (p. 140). Given that truths about broadly
scientific entities are transparent to facts about non-causal aspects of entities, from an entity’s
possession of power-tokens of a type it does “automatically” follow that the entity tokens the
associated type, whether or not properties have non-causal aspects.
12
Finally, a realization relation satisfying SCCP satisfies Non-reduction. To
start, if M is associated with only a proper subset of P’s powers, then M is
ontologically autonomous (distinct) from P, by Leibniz’s law.
Might M also be causally autonomous—that is, distinctively causally efficacious vis-á-vis an effect that P also (by CCP) causes? Yes, supposing that one
way for a property to be distinctively efficacious is in virtue of having a distinctive set of causal powers, or as I will sometimes put it, a distinctive power
profile. The underlying basis for the powers-based subset strategy is the claim
that causal autonomy does not require that M have a distinctive power: it is
enough that M have a distinctive set (collection, plurality) of powers.
How might having a distinctive power profile suffice for causal autonomy?
One case for this appeals to difference-making or other “proportionality” considerations, in cases where M (or M ’s state type) is multiply realizable. Suppose M
is a state of feeling thirsty, which causes a reaching for a glass of water (effect E).
Now suppose that M (or another instance of M ’s state type) had been realized by
P 0 rather than P , in circumstances relevantly similar to those in which M caused
E. Then would E (or an entity of E’s type) have still occurred? Intuitively, yes,
since the “extra” powers possessed by P (say, to produce a certain reading on a
neuron detector) don’t matter for the production of E. All that matters for the
production of E are the powers associated with M (or instances of M ’s type).
That M ’s distinctive power profile contains just the powers that are crucial for
E provides a principled reason for taking M to be causally efficacious vis-á-vis
E in a way that is distinctive from P ’s efficacy vis-á-vis E.21
Another case in favor of taking a distinctive power profile to suffice for causal
autonomy appeals to the fact that certain such sets of powers are associated with
21
Both Shoemaker and Yablo take difference-making considerations to establish, not just
that M may be distinctively causally efficacious, but that M may be the only cause of E
(or that M may be a cause of E, while P is not). Hence while Yablo does not think that
(realized) determinables and (realizer) determinates compete for causal relevance or causal
sufficiency for a particular effect, he does think they compete for the status of being the cause
of that effect (1992, p. 274), and Shoemaker follows Yablo in thinking that “realized properties
can be causally efficacious, and it can sometimes be the case that it is the instantiation of a
realized property rather than an instantiation of a realizer of it that should be regarded as
causing a certain effect” (2001, p. 31). I think this is the wrong way to interpret differencemaking considerations, from a physicalist point of view. Accounts of non-reductive realization
are supposed to provide a solution to the problem of MC compatible with the efficacy of the
physical realizer, in accord with the general principle that every physically acceptable effect has
a purely physically acceptable cause (sometimes called “the causal closure of the physical”).
This principle would seem to be violated if a realized state can preempt its physical realizer
state from being a cause (though Yablo resists violation on grounds that the realizer state is
sufficient for the cause). Since autonomy can be gained just with distinctive (as opposed to
exclusive) causal efficacy, there is no need to interpret difference-making results as showing
that the physical state is not a cause.
13
distinctive systems of laws (for example, the special science treating entities of
M ’s type). Plausibly, systems of laws track causal joints in nature. Correspondingly, M ’s distinctive power profile is indicative of such a distinctive causal joint.
Causal joints may overlap—both in respect of M ’s power to cause E being identical with P ’s power to cause E (as per CCP), and in respect of E.22 Still, if the
joints as a whole are different, this provides a principled basis for taking M to be
distinctively efficacious vis-á-vis E, in that M produces E as part of a different
system of laws (causal joint) than P .23
2.3
NRP accounts satisfying SCCP
Let’s now see how the aforementioned accounts of non-reductive realization each
arguably work to guarantee satisfaction of SCCP.24
First, consider “functionalizability” accounts of realization, according to which
realized types are second order types associated with causal roles that are played
by the entities realizing tokens of these types (on a given occasion). A causal
role is just a collection of powers. Ergo, if M is of a functional type, then, on
any given occasion, every power of M will be numerically identical with a power
of the physically acceptable state P that plays M ’s causal role on that occasion.
Functional realizability accounts of the relation between mental and physical
states thus satisfy CCP.
Does functional realizability accounts moreover satisfy SCCP? One might
think not, on grounds that a functionally realized feature inherits all of the
token powers of the feature realizing it on that occasion:
A functional reduction of pain has the following causal and ontological implications: Each occurrence of pain has the causal powers of
its neural realizer; thus if pain occurs by being realized by N , this
occurrence of pain has the causal powers of N . [. . . ] In general, if M
22
Supposing laws express type-level goings-on, the overlap will be, in the first instance, at
the level of types of the states and effect at issue.
23
This is, I think, Macdonald and Macdonald (1995) have in mind when they say, in arguing
that mental properties may be causally autonomous (“relevant”), that “any instance of a causeeffect relation can be an instance of more than one pattern” (p. 71) and that mental properties
do stand in such distinctive patterns. (That said, the MacDonalds hold that mental property
instances (states) are token-identical with their physical realizers; I discuss such “hybrid”
accounts down the line.) And it is what I have in mind in Wilson forthcominga, § 5.2.1.
24
As we’ll see, these results are most straightforwardly accomplished under the assumption
that M ’s type is multiply realizable. However, satisfaction of SCCP can be gained even assuming the realized type is only singly realized, if the realized and realizing types enter into
different laws (where, as per the second consideration showing how a distinctive power profile
makes room for causal autonomy) distinctive systems of laws, tracking distinctive causal joints
in nature, go hand in hand with distinctive power profiles. See Wilson forthcominga for details.
14
occurs by being realized by N on a given occasion, the M -instance
has the causal powers of the N -instance (Kim 2006, p. 554).
But when a functional role may be played by multiple realizers, there is a case to
be made that the realized feature has, on a given occasion, only a proper subset
of the token powers of the realizing feature. To see this, recall the analogy initially motivating functionism (see, e.g., Putnam 1967), to cases where multiple
hardware systems may implement the instructions associated with a given piece
of software. Here the realizing systems are similar with respect to powers needed
to implement the software, but different with respect to powers associated with
their distinctive hardware. More generally, in cases where a type of functionally
characterized higher-level feature may be multiply realized, it is plausible that
each of its realizing types will have all of the powers associated with its functional
role, and more besides.25 Correspondingly, a proper subset relation will hold between the powers of the realized type and those of any of its realizing types. This
relation between powers will plausibly hold on any occasion of realization involving tokens of the types at issue; hence functional realization arguably satisfies
SCCP (hence both Physicalism and Non-reduction).
Second, consider mereological accounts of realization (see Shoemaker 2001,
Clapp 2001).26 Proponents of such accounts suppose that mental states (Shoemaker) or state types (Clapp) are proper parts of the base states/state types
upon which they depend, reflecting that the relation between part and whole
appears to satisfy the NRP desiderata. In particular, proper parts are distinct
from and yet in a sense nothing over and above wholes, and may be distinctively
efficacious w.r.t. certain effects also caused by corresponding wholes without inducing overdetermination, as when both I, and my eye, cause a wink. As Clapp
(2001) puts it,
25
Hence Antony and Levine (1997) say that for causal autonomy of functional properties,
“What we really need is a “realization-indifferent” regularity: a contingent regularity that
essentially involves the second-order property, and that applies to any instance of the property, no matter the form of realization. We think that such regularities do exist. [Consider]
the second-order property of dormativity [and] this generalization: “dormative substances, if
ingested before driving, cause traffic accidents.” This generalization is true of anything that
induces sleep, regardless of the mode of realization. [. . . ] We contend that the availability
of this sort of generalization certifies “dormativity” as a genuine property, associated with a
distinctive set of causal powers” (p. 93, my emphasis); they then argue that mental properties
are similarly associated with such “realization-independent” generalizations.
26
Certain mereological accounts of the constitution relation as holding between objects (e.g.,
as per Paul 2002) might be easily repurposed as mereological accounts of realization as holding
between features. In this vein its worth observing that among the main puzzles of material constitution is a variation on the theme of causal overdetermination, and non-reductive responses
to this puzzle are likely to require something like satisfaction of SCCP (as I think Paul’s does).
Consideration of such extensions would, however, take me too far afield.
15
[M]ultiply realized mental properties, though real and causally efficacious, are better thought of as parts of their physical realizors. [. . . ]
Just as there is no causal and/or explanatory competition between a
whole and its parts, so there is no causal and/or explanatory competition between instances of mental properties and instances of their
physical realizers.
Correspondingly, if mental states/state types are parts of base states/state types,
then the former look to be ontologically and causally autonomous from the former.
What is the connection between a mereological account of realization and
satisfaction of CCP or SCCP? On both Shoemaker’s and Clapp’s accounts, the
mereological account follows from a more general understanding of realization as
involving CCP or SCCP. Hence Shoemaker (2001) first presents an account of
realization in terms of CCP:
Property X realizes property Y just in case the conditional powers
bestowed by Y are a subset of the conditional powers bestowed by
X (and X is not a conjunctive property having Y as a conjunct).27
(p. 78)
He then goes on to claim that multiply realized features satisfy SCCP:
Where the realized property is multiply realizable, the conditional
powers bestowed by it will be a proper subset of the sets bestowed
by each of the realizer properties. (pp. 78-9)
Shoemaker’s reasoning here broadly follows that at issue in functional realization, with the main difference being that he supposes that all broadly scientific
properties, not just functional properties, are associated with distinctive sets of
powers—causal profiles, if not causal “roles”. If such a feature is multiply realized, its realizing types will share all the powers of the realized type, but will
moreover differ from each other in respect of further powers. This relation will
plausibly hold on any occasion of realization of tokens of the types at issue; hence
a feature realized as per a powers-based account will satisfy SCCP. Filling in this
fairly schematic account, Shoemaker notes that in cases where a proper subset
27
Shoemaker’s terminology of ‘conditional causal powers’ is intended to flag the distinction
between internal and external circumstances: a conditional causal power is a power (to cause
certain effects, if certain external circumstances are in place) conditional on certain internal
circumstances being in place. So, for example, the power to cut wood (bestowed on a knife) is
a conditional power of being knife-shaped—conditional on the knife’s also being made of steel.
This distinction plays no role in what follows, so here we may treat powers as conditional on
whatever internal and/or external circumstances are requisite to their being manifested.
16
relation between powers is at issue, we may see realization as a variety of (proper)
parthood:
Likewise, the instantiation of a realizer property entails, and might
naturally be said to include as a part, the instantiation of the functional property realized. (p. 81)
Clapp similarly supposes that a mereological understanding of realization is motivated by the supposition that types, hence instances of, realized features have
a proper subset of the powers of their realizing features, as per SCCP.
Alternatively, rather than take satisfaction of SCCP as motivating a mereological account of non-reductive realization, we can “backwards-engineer” the
need to satisfy SCCP from a mereological approach, as follows. First note that
if properties are allowed to have non-causal aspects (e.g., quiddities, or primitive
identities, or epiphenomenal qualia) the fact that a realized state mereologically
properly overlaps a realizing state need not indicate that the realizing state has
any powers at all (hence is at all efficacious), much less that it is distinctively
efficacious, since the overlap might be just in respect of the non-causal aspect(s).
Hence if mereological overlap is to provide a basis for distinctive efficacy of a
realized feature, then the overlap must be specifically in respect of powers, as
per SCCP.
Importantly, nothing in the preceding line of thought requires accepting that
the powers of a broadly scientific feature are essentially or exclusively individuative of it (as per Shoemaker 1980, Ellis 2001, and others). Above we saw that
Shoemaker and Clapp motivate a broadly mereological understanding of nonreductive realization as following from the holding of a proper subset relation
between powers of the features at issue; in turn they see a powers-based account
of realization as naturally flowing from a causal or dispositional essentialist view
of properties: if features are exhaustively as well as essentially characterized by
powers (type or token) then a feature satisfying SCCP will clearly be nothing over
and above its realizing feature. But endorsement of a causal theory of features
is not necessary either for endorsement of a powers-based account of realization
or for implementing the powers-based subset strategy. As previously (see §2.2.3,
especially note 20), any non-causal aspects of features there might be (e.g., noncausal quiddities or epiphenomenal qualia) are irrelevant to issues of physical
realization (or constitution) of broadly scientific features. For similar reasons,
the metaphysical possibility of a feature’s existing while having powers different
from those it actually has or bestows is irrelevant to whether and how such features are physically realized in worlds governed by the same laws as ours—the
17
only worlds of interest, or so it seems to me, when issues of dependence between
broadly scientific entities are at issue. And indeed, not just a contingentist but
even a Humean categoricalist may endorse a powers-based account of realization
and associated subset strategy: here the powers will be contingently given by
regularities among the categorical features, and the proper subset relation between collections of powers will be grounded in one regularity’s being part of
another.
On the other hand, it is worth noting that if properties are thought to have
non-causal quidditistic or categorical aspects, then it will clearly not suffice for
the holding of the proper subset relation between powers that one property be
a part of the other. In that case some other metaphysical relation or relations
should be cited as illustrative of powers-based realization (e.g., functional realization or the determinable-determinate relation), if such illustration is needed.
Finally, consider accounts of non-reductive realization in terms of the determinable/determinate relation (see Yablo 1992, Wilson 2009), the relation of
increased specificity paradigmatically holding between colors and their shades.
Yablo (1992) suggests that taking mental properties/states to be determinables
of physical property/state determinates avoids problematic overdetermination
(“competition”):
[W]e know that [determinables and determinates] are not causal rivals. This kind of position is of course familiar from other contexts.
Take for example the claim that a space completely filled by one object can contain no other. Then are even the object’s parts crowded
out? No. In this competition wholes and parts are not on opposing
teams [. . . ]. Determinables and their determinates, like objects and
their parts, are guaranteed to be on the same team (p. 259).
As Yablo’s claim that determinables and determinates “are not causal rivals”
(p. 259) suggests, and as is developed in Wilson 1999 and 2009, the determinabledeterminate relation satisfies CCP. Consider a paradigm case of determination,
holding between the properties (state types) red and scarlet, and consider a patch
that is red, and in particular scarlet. On a given occasion, Sophie the pigeon,
who has been trained to peck at any red patch, is presented with the patch,
and she pecks. Presumably, the patch’s being red on the occasion in question
caused Sophie to peck—after all, she was trained to peck at red patches. But
also presumably, the patch’s being scarlet caused Sophie to peck—after all, to
be scarlet just is to be red, in a specific way. Nonetheless, Sophie’s pecking was
not problematically overdetermined. And the reason why this is so—why determinables and their determinates are not causal rivals—is because it is plausible
18
that each power of the instance/state of red (on the occasion giving rise to the
pecking) was numerically identical to the instance/state of scarlet determining
the instance/state of red on that occasion. Hence the determinable/determinate
relation satisfies CCP.28
Does the determinable/determinate relation satisfy SCCP? Again, one might
be inclined to think not, on grounds that determinable and determinate tokens
are identical (as per MacDonald and MacDonald 1986 and Ehring 1996), in
which case a higher-level feature will inherit all of the token powers of the feature
realizing it on that occasion. But again, there is a case to be made that instances
of determinables have only a proper subset of the latter token powers (against
token identity, about which more anon). Consider again the patch and the
pigeon. Given her training, Sophie would have equally pecked if the patch had
been a different shade of red (burgundy, say), but not so for Sophie’s cousin
Alice, who was trained to peck at only scarlet patches. This suggests that both
the determinable type and associated instance (state) of red has fewer powers
than the determinate type and associated instance (state) of scarlet (the latter,
but not the former, can enter into causing Alice to peck).29 More generally:
broadly scientific determinables are associated with distinctive sets of powers;
moreover, it is a typical and even, some think, essential feature of determinables
that they may be “multiply determined” by mutually distinct determinates. For
the familiar reasons, then, determinable types will have only a proper subset of
the powers of the determinate types that may determine them them. This proper
subset relation will plausibly hold on any given occasion of realization; hence the
determinable/determinate relation satisfies SCCP.30
28
In Yablo’s terms (involving the comparison of the “essences” of mental and physical events):
“Whenever a mental event M occurs, there occurs also a subsuming physical event p, that is,
a physical event whose essence includes m’s essence” (p. 268). Here the mereological and
determinable/determination-based accounts of realization broadly agree.
29
In Yablo’s terms, “[I]f there is an essential difference between mental events and physical
ones, it is not that physical events’ essences are mentally impoverished. Instead, I suggest,
it is the other way around: the essences of mental events are physically impoverished. [. . . ]
Therefore [. . . ] For every mental event m, and every physical events P which subsumes m, P
subsumes M properly and so determines it” (p. 268).
30
See Wilson 2009 for fuller discussion of and defense of determination as the basis for a
viable account of non-reductive realization (for the mental/physical case, in particular).
19
2.4
The main objection diagnosed, and the necessity and
sufficiency claims
The main objection to NRP, due to Kim (1989, 1993a, and 1998) is that NRP
cannot avoid problematic overdetermination in cases of realization without collapsing either into reductive Physicalism (contra Non-reduction) or expanding
into robust emergentism (contra Physicalism). But, I have argued, attention
to SCCP provides a schematic account of how problematic overdetermination
may be avoided compatible with both Physicalism and Non-reduction. Moreover, several accounts of non-reductive realization currently on offer arguably
satisfy SCCP. We will shortly consider certain direct and indirect objections to
the strategy, but I want to start by flagging the core presupposition that gives
rise to the first horn of Kim’s dilemma, the denial of which makes room for the
powers-based subset strategy.
The first horn of Kim’s dilemma is supposed to turn on the need for the
NRPist to endorse the CCP (the “causal inheritance principle”), which is indeed
plausibly required for Physicalism. Kim has observed that CCP may be satisfied
via SCCP.31 But in his discussions of the dilemma, Kim does not consider the
available cases seeming to show that SCCP suffices for ontological and causal
autonomy.
Similar remarks apply to Pereboom’s (1991, 2002) claim that satisfaction of
SCCP does not suffice to gain a sufficiently “robust” sort of autonomy. He says,
with respect to our crucial case, that if a “token-identity thesis for these causal
powers” is held,
. . . then the causal powers to which the psychological explanation
refers would in the last analysis, in fact, be microphysical. Psychological explanations might then presume a classification that clusters
microphysical causal powers in a way distinct from how microphysics
31
These observations are fairly recent; in earlier work Kim presupposed that instances of
realized entities inherit every power of instances of their realizer entities. This supposition is,
for example, the basis of Kim’s rejection of the disjunctivist strategy for reducing multiply
realized properties (in favor of a species-specific reduction) in his (1992). Kim rejects this
strategy on grounds that Fodor’s objection to identifying multiply realized properties with
disjunctions of their realizers—namely, that such disjunctions would be too heterogeneous to
enter into laws, hence fail to be natural kinds—applies just as well to the realized mental
kinds; but the inheritance of heterogeneity holds only if an instance of a realized entity inherits
all the powers of its realizing instance on a given occasion. Relatedly, Kim (1993a) seems
to have supposed that, if the NRPist accepts that real (broadly scientific) states have powers
(“Alexander’s dictum”) and that some real realized states are irreducible, then the NRPist must
accept that some real realized states have irreducible powers; but this supposition neglects the
possibility of the NRPist’s rather locating irreducibility in the having of a distinctive power
profile.
20
sorts them, but this would not compromise the microphysical status
of those causal powers (2002, 500).
And he concludes, in agreement with Kim, that accounts requiring token identity
of powers are incompatible with “robust non-reductive materialism” (p. 502).
Such remarks, however, do not constitute an argument against the powers-based
subset strategy as they serve to describe it. Why, then, do Kim and Pereboom,
among many others, think that there is no way for higher-level features to be
genuinely reductive if they satisfy CCP—even if, as per the subset strategy, they
moreover satisfy SCCP?
The reason, I believe, is that these philosophers are recognizing only a single
form of causal efficacy—namely, that associated with the having and manifestation of a power. Consequently, they recognize only a single way for a higher-level
feature to be distinctively efficacious; namely, in virtue of having a distinctive
power. If CCP, or SCCP, is satisfied, then this route to the distinctive efficacy
of higher-level features is precluded, in which case the prospects of grounding
any kind of NRP will seem dim indeed. Moreover, if distinctive powers are the
only route to distinctive efficacy, it’s easy to see how the NRPist gets pushed to
the “robust emergentist” horn of Kim’s dilemma, for if the distinctive efficacy
of a higher-level feature lies in its having a distinctive power—that is, a power
that its realizer doesn’t have—then it’s hard to see how such a feature could be
physically acceptable.
As above, however, there is arguably another route to distinctive causal efficacy, associated not with distinctive powers, but with distinctive collections of
powers. I suggested, in particular, that such an alternative conception of efficacy
may be grounded either in proportionality or difference-making considerations
or in distinctive systems of laws or causal joints; more generally, the alternative
conception might be seen as locating efficacy in extrinsic rather than intrinsic
grounds.
So stands my diagnosis of the ultimately non-fatal grounds of the main objection to NRP. Now on to my broader positive claims.
To be sure, it is a further question whether a given relation satisfying SCCP
is in fact suited to relate mental and physical states. So, for example, some deny
that mental states are appropriately seen as determinables of physical determinates (but see Wilson 2009). Supposing a given relation is so suited, however,
then satisfaction of SCCP—that is, implementation of the powers-based subset strategy—is, I have argued, sufficient to show that the relation serves as an
adequate basis for non-reductive realization, suited to the NRPist’s purposes.
21
I moreover claim that implementation of the powers-based subset strategy is
necessary for an adequate NRPist account of realization. My reasons for thinking
this reflect that, of the schematic options for resolving the problem of mental
causation, only one of these avoids both horns of Kim’s dilemma, combining
an endorsement of CCP (thus preserving Physicalism) with an endorsement of
SCCP (thus ensuring Non-reduction). If I am right that the powers-based subset
strategy appropriately characterizes this option, and that the associated apparent
satisfaction of both Physicalism and Non-reduction is genuine, then there is really
only one route to non-reductive realization, though one may travel it in diverse
ways.
I turn now to defending the sufficiency and necessity claims against various
objections.
3
Objections and responses to the sufficiency
claim
I have seen three sorts of objections to the sufficiency claim: the objection from
eliminativism, the objection from reductionism, and the objection from token
multiple realizability.
3.1
The objection from eliminativism
The powers-based subset strategy supposes that, when a given higher-level type
is associated with a distinctive set of lower-level powers, it may be reasonably
supposed to be distinct from and distinctively efficacious with respect to its lowerlevel realizer. The objection from eliminativism proceeds by noting that one
need not take the appearance of such higher-level features ontologically seriously,
rather taking this appearance to reflect merely pragmatic ways of classifying or
talking about lower-level types.32 Perhaps higher-level predicates do not pick out
any properties,33 notwithstanding that they are useful for explanatory or other
pragmatic purposes34 .
The response proceeds by recalling the status of the present dialectic. Debates
32
Here eliminativism is contrasted with reductionism; hence nominalists who offer one or
other ontological reduction of properties of whatever level or type, should not immediately
be counted as eliminativists. Of course, nominalists might be eliminativsts about higher-level
features, if these do not satisfy their preferred reduction.
33
As John Heil, David Robb, and Sophie Gibbs suggested
34
As Alyssa Ney and Derk Pereboom suggested
22
over the ontological status of special science entities start with the “pre-theoretic”
appearances, according to which (in particular) some higher-level features are
both ontologically and causally autonomous from any lower-level physically acceptable features. Certainly there are well-rehearsed ways of explaining away
these appearances, but absent independent reason for going some such way, the
present burden on NRPists is simply to show how, assuming the appearances
are genuine, they may be metaphysically accommodated. Indeed, independent
reasons for eliminativism about higher-level features (not to be confused with
reductionism about higher-level features, to be treated shortly) are in short supply. Considerations of Ockham’s razor or ontological parsimony alone do not
provide support for eliminating any and all seemingly higher-level features, for
such a general strategy of elimination would be seriously revisionary, entailing, in
particular, that all special science claims were either false or meaningless (since
containing non-referring terms) or else mean something different than they seem
to mean. Less revisionary would be a selective eliminativism, allowing that some
higher-level features exist and that some special science claims are true—namely,
those that can be ontologically reduced—but eliminating any purportedly nonreductive such features. The primary motivation for such a moderate eliminativism is (one or other variation on) Kim’s claim that attention to the problem of
mental (higher-level) causation indicates that one cannot have one’s physicalist
cake and one’s non-reductionism, too. But the proper subset strategy, assuming
it works as I have suggested, undercuts this claim, showing both schematically
and by attention to various contemporary approaches how non-reductive realization may be gained. Relatedly, the strategy provides clear reasons for thinking
that a properly metaphysical reading of features having distinctive power profiles
can be given. Of course, someone with a taste for minimalist landscapes may
resist following the NRPist here. But no matter; the NRPist does not need to
knock the higher-level eliminativist off their horse; they only need a principled
way to stay on their horse.
3.2
The objection from reductionism
A trickier objection grants the reality of the higher-level feature, but presents
a challenge: given that (as per SCCP) every power of the higher-level feature
is identical with a power of any physically acceptable feature that realizes it
on a given occasion, why isn’t this higher-level feature appropriately deemed a
23
physically acceptable feature?35
It is first necessary to refine the objection, since as stated the NRPist may
simply respond by agreeing that higher-level features are physically acceptable,
just like all features, as per Physicalism. What the NRPist more specifically
claims is that higher-level features are, while physically acceptable, not reducible
to any lower-level physically acceptable features. And what this means, as per
Non-reduction, is that a given higher-level feature satisfying SCCP is not identical
to any feature of individual physically acceptable entities, nor to any relational
feature consisting in lower-level entities standing in lower-level relations, nor to
any feature consisting in a boolean or mereological combination of such relational
or non-relational lower-level features. Call any of the latter sorts of features
“reduction base features”. More precisely, then, the challenge posed by the
objection from reductionism is: given that (as per SCCP) the powers of the
higher-level feature are each identical with powers of the (presumed physically
acceptable) feature that realizes it on a given occasion, why isn’t this feature
itself appropriately deemed a lower-level physically acceptable feature, in being
identical to one or other reduction base feature?
There are two strategies for answering the challenge. The first appeals to
strong multiple realizability considerations to indicate that it is implausible that
any such identification with a reduction base feature is in the offing.36 Here the
NRPist points out that, notwithstanding their endorsement of Physicalism as an
appropriate account of the broadly scientific entities around here, it is nonetheless possible that a given higher-level feature might be realized by features that
are not physically acceptable. As such, any identification of a higher-level feature with some lower-level relational, boolean or mereological combination of its
realizers must incorporate, somehow or other, reference to physically unacceptable realizers it might have. Since no reductive base features incorporate such
reference, higher-level entities cannot be identical with any such features.
The second proceeds constructively, and does not require appeal to strong
multiple realizability, and indeed, may be implemented even if a given higherlevel feature has only a single realizer. The strategy proceeds by arguing that
none of the features in the reduction base could have the same powers as are
35
Though Kim has not treated the powers-based subset strategy in detail in print, this is
plausibly the heart of his concern about the powers-based subset strategy as providing the
basis for a viable NRP, as expressed in a panel on Shoemaker’s (2007) at the 2009 Pacific APA.
Jim John and Timothy O’Connor independently raised a similar objection in Q&A pertaining
to this paper.
36
Many thanks to Ranpal Dosanjh for identifying the following strategy for blocking ontological reduction.
24
associated with the higher-level feature in question; hence by Leibniz’s law, such
a higher-level feature cannot be identified with any reductive base entity. The full
implementation of this strategy is somewhat intricate (see Wilson forthcominga);
here I will sketch the approach.
To start, as per SCCP, the higher-level feature has a proper subset of the
powers of any feature that realizes it, on a given occasion. Hence the higher-level
feature is not appropriately identified with any one of these realizers. Nor is it
clear how disjoining, conjoining, or fusing any of these lower-level realizers would
result in a feature that would, like the higher-level feature in question, have only
a proper subset of the features of its realizer on a given occasion. So, for example,
consider a feature consisting of a disjunction of lower-level realizers—that is, the
sort of feature sometimes suggested as a candidate for identity with a multiply
realized feature. Arguably, any instance of such a disjunction of lower-level realizers will be identical to an instance of one of its disjuncts (namely, whatever
disjunct is instanced on the occasion in question); but any such instance will be a
lower-level realizer, and hence, by assumption, such that it has a proper superset
of the higher-level feature in question on the occasion in question. So higher-level
features of the sort satisfying SCCP cannot be identified with a disjunction of
their lower-lower level realizers. Other modes of reductively acceptable combination similarly do not seem to result in any decrease in specificity of the resultant
feature’s powers, on any occasion of its instancing. So the challenge, it seems, is
on the reductionist’s side, to show that some feature in the reduction base has
powers that are, like the powers of the higher-level feature, restricted in the way
that SCCP requires.
3.3
The objection from token multiple realizability
Pereboom (2002) provides a positive argument for rejecting CCP, thus indirectly
arguing against the sufficiency claim (since SCCP entails CCP). He starts by
arguing that, in our case, (token) state M can be multiply realized, and so is not
identical with state P :
Is token mental state M identical with P , its actual token microphysical realization base? Suppose that M is realized by a complex neural
state. It is possible for M to be realized differently only in that a few
neural pathways are used that are token distinct from those actually
engaged. [. . . ] [I]t is evident that this alternative neural realization is
itself realized by a microphysical state P that is token distinct from P .
It is therefore possible for M to be realized by a microphysical state
not identical with P , and thus M is not identical with P . (p. 503)
25
He continues:
Furthermore, this reflection would also undermine a token-identity
claim for mental causal powers—should they exist—and their underlying microphysical causal powers. For if the token microphysical
realization of M had been different, its token microphysical causal
powers would also have been different. We therefore have good reason to suppose that any token mental causal powers of M would not
be identical with the token microphysical causal powers of its realization.
There are two ways that a proponent of CCP, or of SCCP and the powers-based
subset strategy, may respond to Pereboom’s argument.
First, they may deny that M is token multiply realizable. What is (fairly)
uncontroversially true is that states of M ’s type are multiply realizable. Pereboom’s case in support of that very token M ’s being multiply realizable isn’t
compelling, and may be rejected. Compare: could that very instance of red,
currently realized by an instance of scarlet, have been realized by an instance of
burgundy? Intuitively not; and I don’t see that (or why) intuitions would change
if the shade is only “slightly” different. In any case, it seems reasonable to deny
that a given instance of red could be realized by a very different, or even by a
slightly different, shade than that which actually realizes it. One may similarly
reasonably deny that M (as opposed to a different instance of M ’s type) might
be realized by a physically acceptable relational state different from P , whether
this alternative state be very different from P , or only different with respect to
“a few neural pathways”.
Second, the proponent of CCP/SCCP may allow that M is token multiply realizable, but maintain that token powers of M are relativized to realizations/occasions. Pereboom seems to presuppose that M has its token powers
essentially; but why think this? As he later observes, “stable tokens [. . . ] often
retain their identity over certain changes in their constitutions and configurations” (p. 529). A proponent of CCP may similarly maintain that M can retain
its identity across changes in its realizers and associated token causal powers.
CCP requires only that M ’s token powers on a given occasion be numerically
identical with the powers of the state realizing it on that occasion. So CCP may
be true even if M is token multiply realizable.
26
4
Objections and responses to the necessity claim
Must an adequate account of non-reductive realization implement the powersbased subset strategy? Here I will consider a representative sample of NRPist
accounts on which the relation at issue is not guaranteed to satisfy SCCP, associated with supervenience-based accounts, token-id accounts, and with Pereboom’s
“robust nonreductive materialism”, respectively. I will argue that superveniencebased accounts and token-id accounts fail to establish that a realized state is
causally autonomous (contra Non-reduction), and that supervenience-based accounts and Pereboom’s account fail to establish that a realized state is physically acceptable (contra Physicalism). Moreover, as we’ll see, these difficulties
can each be traced directly to the failure of the account to entail satisfaction of
either CCP or SCCP. I will conclude that there is considerable inductive support
for the necessity claim.
4.1
Supervenience-based accounts
Supervenience is a relatively abstract relation, in which dependence of higherlevel entities on lower-level entities, and/or determination of higher-level entities
by lower-level entities, is cashed in terms of correlations alone, leaving aside any
specific account of what underlying metaphysical relations might be responsible
for the correlations.37 There is a history of attempts, starting with Davidson and
continuing to the present day, to characterize non-reductive realization by appeal
to one or other form or strength of supervenience as holding between (families
of) the types at issue (such that, e.g., mental properties supervene on physical
properties).
Supervenience-based accounts of non-reductive realization are doubly problematic, from an NRPist perspective. First, such accounts fail to guarantee Nonreduction. Supervenience of types is compatible with identity of tokens (as on
Davidson’s anomalous monism); hence, for reasons I will present when discussing
token-id approaches in §4.2, do not guarantee ontological or causal autonomy of
states.
Second, supervenience-based accounts do not guarantee Physicalism. Physicalists seem to have hoped that the holding of supervenience correlations of
sufficient modal strength, in particular, would suffice to transmit physical acceptability from base to supervenient entities. Such hopes are encouraged by a
37
See Yoshimi 2007 for an insightful discussion of the notions of dependence and determination, and of how these enter into the standard formulations of supervenience.
27
tendency to assume that broadly causal relations of the sort that might be at
issue in robust emergence hold with only nomological necessity, whereas certain
paradigm relations preserving physical acceptability (identity, for one) hold with
metaphysical necessity. But ultimately these hopes are in vain, for no matter
how strong the supervenience correlations at issue, it might be that a physically unacceptable source is responsible (see Wilson 2005 for discussion of the
following cases). So, for example, mental states might supervene with metaphysical necessity on physically acceptable relational states, thanks to a consistent
Malebranchean god, who occasions certain mental states (of a type) upon the occurrence of certain physical states (of a type) in every possible world. Relatedly,
supervenience is compatible with robust emergentism, given certain live metaphysical and physical theses about powers and laws. For example, if properties
(states) are essentially individuated by their powers (as per Shoemaker 1980), if
laws of nature are metaphysically necessary (as per Swoyer 1982 and Bird 2007),
or if the laws of nature are holistic (as is arguably presupposed in physicists’
attempts to unify fundamental interactions), then any world where the physical
state exists is a world where the mental state exists, even if the latter is partly
caused or constituted by a non-physical fundamental interaction.38
Proponents of supervenience-based accounts of non-reductive realization sometimes deny that such scenarios pose problems for their view, on grounds that the
scenarios violate Hume’s Dictum, according to which there are no metaphysically
necessary connections between distinct existences. One thing to note about this
line of defense is that arguments (much less good arguments) for the dictum are
in short supply.39 In any case, an account of non-reductive realization shouldn’t
rule out live independent physical and metaphysical theses concerning properties
and laws, or intuitive Malebranchean possibilities. The yet deeper point is that
we shouldn’t expect mere correlations to do the work—of illuminating higherlevel dependence, in particular—of a properly metaphysical realization relation.
Finally, it is worth noting that attempts to beef up the supervenience relation
in such a way that physical acceptability is preserved succeed just to the extent
that the resulting “supervenience” relation guarantees satisfaction of CCP. So,
for example, Horgan’s (1993) account of “superdupervenience” requiring that
38
A non-physical (physically unacceptable) fundamental interaction is a fundamental interaction that first comes into play at the level of complexity associated with mentality. See
Wilson 2006 for discussion of why such an interaction should count as non-physical.
39
As I have argued, the dictum is neither analytic, nor grounded in intuitions we have no
good reason to question, nor motivated as required by the best account of counterfactuals (see
Wilson 2005, forthcomingb, and in progressa).
28
supervenient entities be robustly ontologically explainable doesn’t do the trick,
since one variety of robust ontological explanation is causal explanation, which
is compatible with non-identity of token powers of supervenient and base entities
(effects typically having powers that their causes don’t have), and hence with
robust emergentism (see Wilson 1999). By way of contrast, Chalmers’ (1996)
account of “broadly logical” supervenience does guarantee physical acceptability
of supervenient entities, precisely because the required conceptual connections
establish that the supervenient entity is of a functional type whose distinctive
causal role may be played by the base entity—such that, as argued in §2.3, CCP
and indeed SCCP is satisfied.
4.2
Token identity accounts
A properly metaphysical strategy for avoiding overdetermination in our core case
is to identify M and P (see, e.g., Davidson 1970, Macdonald and Macdonald 1995,
Ehring 2003, and Robb 1997). On such an approach, the relation between M
and P satisfies CCP, but not SCCP. Above, I characterized this as a form of
reductionism; and indeed, reductive physicalists will suppose that the states are
identical. But token identity leaves open whether mental state types are identical
to their realizing state types (or, in cases of multiple realizability, disjunctions
thereof). Hence the associated account of realization might be non-reductive at
least in not identifying the state types.40
The token-id approach avoids overdetermination at the level of states. It
doesn’t gain the ontological autonomy of M , but one might think that M’s being
distinct from P isn’t as important as M’s being real and efficacious. It remains,
however, that mental causal autonomy requires not just that M be efficacious,
but distinctively so.41
40
That is, assuming that the operative notion of state isn’t such that identity at the level of
tokens entails identity at the level of types; it was in anticipation of this point that I said that
I was here treating states “roughly” as property instances. It was also why I chose to focus
on states rather than events, since by some lights (see Kim 1976) events have feature types as
constitutents, such that identity of event tokens entails identity of event types.
41
This claim reflects what Horgan (1989) calls “the problem of mental ‘quausation”. As
Yablo (1992) puts it, “To reply [to the problem of MC] that mental events just are certain
physical events, whose causal powers they therefore share, only relocates the problem from the
particulars to their universal features. [. . . ] Mental events are effective, maybe, but not by
way of their mental properties; any causal role that the latter might have hoped to play is
occupied already by their physical rivals” (pp. 248-9). And as Shoemaker (2001) puts it: “If
what is in fact an instance of a mental property causes something [. . . ] but does so in virtue
of being an instance of a physical property rather than in virtue of being an instance of a
mental property, then the causal efficacy of the mental does not seem to have been adequately
vindicated” (p. 28).
29
How might a mental state be distinctively efficacious, on a token-id account?
It cannot be so in virtue of its distinctive powers, or in virtue of having a distinctive cluster of powers. The remaining option, endorsed by some NRPist token-id
theorists (MacDonald and MacDonald, Ehring), is that a mental state may be
distinctively efficacious in virtue of falling under a distinctive state type.42
The appeal to non-identical state types as the ground of the causal autonomy
of states reintroduces a (or the) threat of causal overdetermination, and associated threat of exclusion. As Ehring (2003) puts it, the concern is: “Since mental
types are not identical to physical types (because of multiple realizability) even
if mental [states] are identical to physical [states], there are no causes of physical
effects that are efficacious in virtue of mental property types” (p. 364). In order
to gain the autonomy of mental states, the token-id theorist needs to explicitly
address the relation between the types associated with M and P , respectively,
and show that the associated means of gaining autonomy is not such as to induce
causal overdetermination, at either the level of states or state types.43
Ehring, who endorses a token-id account of states (hence CCP but not SCCP),
provides an account of the relation between types aiming to show that this threat
is avoided. Being a trope theorist, Ehring understands the states at issue as
occurrences of tropes, and the types at issue as collections of resembling tropes;
he then argues that the mental and physical types at issue are related as part
to whole. Here the order of the part/whole relation between types is reversed
42
Davidson cannot avail himself of this strategy, since he supposes that mental states (events)
cause what they do only in virtue of falling under physical types (and associated laws); Robb
chooses not to do so, rather maintaining that the problem of exclusion does not crop up again
at the level of types, at least if the token features at issue are tropes (but see note 42).
43
Robb (1997) denies that issues of efficacy or overdetermination crop up again at the level of
trope-theoretic types: “[I]t would be odd if once the qua problem were solved for [Davidsonian]
events, it arose again regarding the very thing used in the solution; the objector seems to be
making a kind of category mistake. Tropes are not causally relevant qua this or that, they
are causally relevant (or not), period”; and again: “we cannot use the same considerations to
raise the qua problem for causally relevant properties, whether or not we view those properties
as tropes. A causally relevant property F simply does not have various aspects such that
one can legitimately ask whether some but not others are responsible for F’s being causally
relevant. Suppose, for example, that a red ball dropped on a sheet of metal causes a dent qua
massive thing, but not qua red thing. It would then be odd to object: ’Yes, perhaps its mass
is causally relevant here; but is it causally relevant qua mass?’. Surely a good answer here
is: ‘Its mass is not relevant qua this or that, it is just causally relevant, period’.” But the
red ball case is disanalogous from the mental/physical case precisely in that the red trope and
mass trope of the ball are not identical, in which case there is no point in asking if the mass
trope is efficacious qua mass trope. Analogy aside, the reference to a resurrected qua problem’s
involving a “category mistake” may reflect Robb’s view that tropes are not themselves suited
to be the bearer of tropes. I’m differently inclined, but in any case it is not crucial that the
resurrected problem take exactly the form of the original. Even if tropes do not have tropes,
we can still ask: is the one mental/physical trope efficacious qua member of the set of mental
tropes, or qua member of the set of physical tropes?
30
from Clapp’s understanding: for Clapp, a realized state type is part of each of
its realizing state types; for Ehring, a realized state type is a whole, having all
the subclasses (collections of resembling tropes) of its realizing types as parts.
What is interesting for present purposes is that Ehring takes appeal to a typelevel powers-based subset strategy to be required in order to establish that distinct mental and physical types stand in relations fit to ground causal autonomy
without inducing problematic overdetermination. He first motivates the view for
the type red and associated shade types (trope-theoretically understood):
It seems clear that the class of red tropes as a whole, the type “red”,
has certain causal powers. [. . . ] [W]e are still left with the question
of how the causal powers of this class as a whole are related to the
causal powers of the subclasses of each determinate shade of red. I
believe the answer is that the causal powers of the type “red” are those
exactly similar causal powers shared by each of these subclasses. [. . . ]
For any causal power of a shade of red not matched by an exactly
similar causal power belonging to each of the other shades of red,
“red” lacks any such power. (p. 374)
That is to say, the powers of the type red are a proper subset of those of its
constituent determinate types. Ehring then notes that a similar line of thought
indicates that the type of a given mental tropes has powers which are those
exactly similar powers shared by its realizer state types. Hence on Ehring’s
view, mental state types have a proper subset of the powers of their realizer
state types.
More generally, token-id theorists typically avoid problematic overdetermination at the level of non-identical types by implementing, explicitly or tacitly, a
subset strategy at the level of types. So, for example, Macdonald and Macdonald
(1995) can be seen as suggesting that type-level overdetermination can be avoided
by taking mental state types to be relevantly analogous to determinables, since
(as per §2.3) determinable types arguably have a proper subset of the powers of
their realizing determinate types.44
Such hybrid approaches are problematic, however. To start, the view is odd:
shouldn’t the subset relation between powers at the level of types be mirrored at
the level of states? As Shoemaker (2001) notes:
[I]t seems doubtful that we should identify the mental property instance with the instance of the physical property that realizes it—or
44
Robb (1997) also follows Ehring in taking mental trope types as having a proper subset of
the powers of their realizing physical trope types, on grounds, however, of multiple realizability
considerations
31
that we should identify the instance of red and the instance of scarlet. If we think of the instantiation of a property as the conferring on
something of the conditional powers associated with that property,
then when properties confer different sets of conditional powers, the
instantiation of one of them is not identical with the instantiation of
the other. (p. 28)
A related but more serious concern is as follows. According to the hybrid
view, M ’s type does not have powers that differ between its realizer types; but
M (being identical with P ) can have such powers. Hence M can have powers
that M ’s type doesn’t have. That seems wrong. Perhaps it makes sense for
a token state to have fewer powers than its state type, reflecting restrictions
associated with circumstances in which the state occurs.45 But it makes no sense
to see a token state’s having more powers than its state type, at least if types
are supposed to track similarities among tokens of the type. If a state has more
powers than a given state type, that is, or so it seems to me, compelling reason
to think that the state is not of the type.46
Avoiding this difficulty requires that the subset strategy hold at the level of
tokens as well as types. More precisely, it requires imposing SCCP if the account
is to be a version of NRP. Alternatively, the proponent of token-id could go
reductive at both token and type levels, or else go robust emergentist at the type
level.
Let’s recap the previous argumentative trajectory:
• Identifying mental and physical states does not, in itself, suffice to establish
the causal autonomy of mental states.
• Causal autonomy of mental states, on a token-id account, requires that
state-level efficacy occur in virtue of mental state types not identical with
physical state types.
45
Again, this is the lesson of Clarke 1999.
David Robb suggested in discussion that this is not problematic, by analogy to a case
where, e.g., I may be admitted to a club in virtue of satisfying some condition, compatible
with my satisfying other conditions as well. But it is unclear that the analogy succeeds. The
club case plausibly reflects that a single substantial particular may be a member of many
different types due to having many different properties; taking the analogy at face value the
suggestion would be that one can make sense of a trope’s having more powers that its given
type on grounds that the trope can have some properties that are not relevant to its being a
member of the type. Robb doesn’t think that tropes can have properties (tropes), but even
if they can that wouldn’t show how a trope of a give type could have more powers than are
associated with a given type, since the powers associated with a given property (trope) aren’t
(on any view I’m familiar with, anyway) properties of that property (trope).
46
32
• Grounding causal autonomy of states in the non-identity of mental and
physical state types gives rise to a new threat of overdetermination.
• Avoiding overdetermination while retaining type-level non-identity requires
implementing a subset strategy at the level of types.
• Implementing a subset strategy at the level of types conflicts with identity
at the level of tokens (a state can have more powers than its state type).
• Summing up: resolving the conflict compatible with NRP requires rejecting
token identity and also implementing the subset strategy at the level of
tokens—that is, imposing SCCP.
At the end of the day, token identity accounts of non-reductive realization either
do not establish autonomy of states, contra Non-reduction; or else impose SCCP
and so are not really token identity accounts.
4.3
Pereboom’s account of non-reductive realization
Pereboom (2002) endorses what he calls “robust non-reductive physicalism”. On
the associated account of our target case, M is neither type nor token identical
with the base feature P upon which it depends on a given occasion; but contra both CCP and SCCP, M ’s powers are not numerically identical with (are
“irreducible to”) powers of P : “robust non-reductive physicalism affirms varous
token-diversity claims for mental causal powers”. Taking M to have powers P
doesn’t have will clearly make sense of M ’s ontological and causal autonomy.
But how are M ’s irreducible powers supposed to avoid problematic overdetermination, in a way compatible with Physicalism?
According to Pereboom, this is because M ’s powers are “constituted” by P ’s
powers, in a way piggybacking on the notion of token (state, event, property
instance) constitution:
Token causal power constitution: The causal powers of a token of
kind F are constituted of the causal powers of a token of kind G just
in case the token of kind F has the causal powers it does in virtue of
its being constituted of a token of kind G.47
The notion of constitution of one token (state, etc.) by another is broadly primitive, but is (as per Pereboom in progress), to be grasped as relevantly analogous
to the “made up of” relation, as holding between one substantial particular and
47
See Pereboom and Kornblith 1991 and Pereboom 2002.
33
one or more others (e.g., a statue and a lump of clay). The account of state
constitution, coupled with the account of token power constitution (above), is
intended to motivate taking the powers of a non-reductively realized feature to
be, while irreducible to, still nothing over above the powers of its realizing feature
(on a given occasion, as per usual), and so compatible both with Physicalism and
the avoidance of overdetermination:
[Though M ’s powers are irreducible to P ’s] there would be a sense
in which the token causal powers of M would be “nothing over and
above” the token causal powers of P . . . M 0 s causal powers would
nevertheless be “absorbed” or “swallowed up” by P’s causal powers.
But there are importantly distinct modes of this sort of absorption:
identity and constitution without identity. [. . . ] token mental causal
powers are wholly constituted by token microphysical causal powers.
(2002, pp. 503-4)
So far, such appeals to token state and power constitution do not show that
satisfaction of CCP (hence SCCP) is unnecessary for non-reductive realization,
for in themselves they do not establish that rejection of CCP is compatible with
Physicalism. “Constitution” is a term of art, mainly applied (as Pereboom notes)
to objects. Where token states are at issue, and where the relation at issue
is supposed to conform to Physicalism, “constitution” is usually just another
name for “realization”.48 But as argued in §2.1, standard cases of non-reductive
realization presuppose satisfaction of CCP. The expression “in virtue of” entering
into the account of token power constitution is also a term of art, compatible
with many underlying relations, from identity and the determinable/determinate
relation (satisfying CCP) to emergence-involving causation (not satisfying CCP,
and contra Physicalism).49
48
Kim (1992) says, “talk of “realization” is not metaphysically neutral: the idea that mental
properties are “realized” or “implemented” by physical properties carries with it a certain
ontological picture of mental properties as derivative and dependent. There is the suggestion
that [. . . ] the instantiation on a given occasion of an appropriate physical property [. . . ] simply
counts as, or constitutes, an instantiation of a mental property on that occasion” (pp. 4-5).
49
Analyses or accounts on which certain goings-on hold “in virtue of” other goings on are
notoriously ill-defined. Hence Polger and Shapiro (2008) complain against Gillett that his formulation of realization “asserts that the entity in which an instance of property G is realized
has the powers individuative of G “in virtue of” the properties that it or its constituents have.
But this runs together all the different ways—all the different dependence and determination
relations—by which the properties of parts may contribute to the property of a whole. [ldots]
it seems to us that other kinds of dependence relations—relations already handy in any metaphysician’s toolkit—are in fact doing all the explanatory work” (p. 220). My similar point in
re Pereboom’s use of “in virtue of” is more specifically that this umbrella term covers relations
not preserving physical acceptability; see Melnyk 2008, p. 1292 for a similar observation.
34
Pereboom offers further considerations in support of irreducible mental powers’ being compatible with Physicalism and with the avoidance of overdetermination, but these do not, I think, establish his case. In re compatibility with
Physicalism, he says that that “correlated with the possibility of this sort of constitutional explanation is the fact that the existence and nature of token higherlevel causal powers would be predictable in principle from their microphysical
constituents together with the laws governing them” (p. 504). But if the powers of M , at either the type or token level, are not identical with the powers of
(any) of its realizers, then what guarantees that the powers of M would be so
predictable? Perhaps such predictability could be guaranteed if every power of
M was type-identical (though token-distinct) with a power of the state realizing
it on a given occasion. But this understanding looks to give rise to the worst
kind of overdetermination.
Pereboom resists, saying that “no competition arises in the case of mere
constitution”:
For if the token of a higher-level causal power is currently wholly
constituted by a complex of microphysical causal powers, there are
two sets of causal powers at work which are constituted from precisely
the same stuff [. . . ] and in this sense we might say that they coincide
constitutionally. (p. 505)
But to the extent that I understand why constitutional coincidence blocks overdetermination, this is because (the relevant sort of) coincidence would entail identity of the powers had by states of the “stuff”. Pereboom acknowledges “that
they now coincide in this way might tempt one to suppose that these causal powers are token identical, but, as we have just seen, there is a good argument that
they are not”. Here Pereboom is referring to his argument from the supposed
token multiple realizability of M to the non-identity of M ’s token powers with
P ’s powers; but as we saw in §3.3, that argument does not work.
Finally, Pereboom provides an indirect argument that problematic overdetermination can be avoided even if CCP is rejected. He starts by saying, “If
identity [of powers] and not just constitutional coincidence were necessary for
[. . . ] noncompetition, then there would be features required for noncompetition
that identity has and current constitutional coincidence lacks” (p. 506). He goes
on to say that the features possessed by identity but not constitutional coincidence are coincidence at all other times and worlds; then notes that lack of such
features surely could not induce overdetermination. But what identity of powers has and constitutional coincidence of powers lacks, and which is necessary
35
for non-competition, is not a matter of goings-on at other times or worlds. It
is rather simply that identity of powers on an occasion guarantees noncompetition on that occasion, whereas constitutional coincidence on an occasion without
identity of powers on that occasion seems not only not to avoid competition, but
positively to invite it (in particular, if M and P are token distinct, but typeidentical). Summing up: Pereboom never provides reason to think that denying
CCP is compatible either with Physicalism or with avoiding problematic overdetermination, much less jointly so compatible.
To show that the denial of CCP is so compatible, Pereboom has to get more
metaphysically specific about what it is for one power to constitute another; but
it’s hard to see how to proceed. To start, there is little hope of implementing a
variant of the powers-based subset strategy holding between individual powers,
for it’s not clear that powers have powers, and even if (for example), M ’s powers
have powers, it’s unclear how the latter would be implicated in the production
of M ’s effects. Alternatively, Pereboom might attempt to make out that one or
other of the metaphysical relations that, when holding between states or properties, arguably guarantee the holding of the associated proper subset relation
between the (token or type) powers of these entities, can also be seen to hold
between powers themselves. But it is unclear that such relations are appropriately extended to powers. So, for example, if M ’s power to cause E is to be a
determinable of P ’s power to cause E, then P (or perhaps P ’s associated effects)
must manifest, relative to M , the sort of increase in specificity characteristic of
the the determinable/determinate relation; but what is the ontological ground
for such a difference? In the case of token states (events, etc.), the powers of
these states serve as the requisite ground: a determinate feature is more specific than its associated determinable feature(s) in that the former has strictly
more powers than the former, on any given occasion. But as before, it is unclear
that powers have powers, and so the usual metaphysical ground for the determinable/determinate relation is not available here. Similarly for the suggestion
that M ’s power to cause E is a proper part of P ’s power to cause E. We can
make out the parthood relation as holding between token states by appeal to
one state’s having more powers than the other; but what are the metaphysical
subconstituents (abstract or concrete) of powers, such that one power can be a
proper part of another in having strictly fewer of these subconstituents? More
generally, it is unclear that powers have the sort of “fine-grained” constituents
whose relations could provide an intelligible ground for showing how problematic
36
overdetermination may be avoided, compatible with Physicalism.50
5
NRP: alive and well
In this paper I have argued for three claims pertaining to the powers-based subset
strategy.
First, I have argued, contra Kim and others, that implementation of this
strategy provides a principled basis both for taking mental and other realized
states to be ontologically and causally autonomous (as per Non-reduction, and
for taking such states to be physically acceptable, given the physical acceptability of their realizing states (as per Physicalism). Second, I have argued that,
whether they realize it or not, many NRPists are implementing a powers-based
subset strategy, according to which the preferred non-reductive realization relation, properly understood, satisfies SCCP. Third, I have argued that implementation of the strategy is not only sufficient but necessary for the viability of a
given account of non-reductive realization, insofar as a representative variety of
purportedly NRP accounts not implementing the powers-based subset strategy
(including accounts based in supervenience, token-identity, and constitution) fail
to guarantee satisfaction of one or both of the core NRP theses. These results
collectively indicate that NRP is alive and well, and so far as I can tell, represents (along with robust emergentism, which remains a live option) an attractive
means of accommodating the pretheoretic appearances of special science entities,
as appropriately dependent on, yet distinctively autonomous from, the lower-level
(ultimately fundamental physical) entities upon which they depend.
Further questions remain, of course, including: What specific relation(s) satisfying SCCP are best suited to non-reductive realization? What other dependence
relations hold between special science and (ultimately fundamental) physical entities? And last but not least: Is NRP true?
50
Another concern remains even if there is some way of making out such a metaphysical
relation between powers. For no matter how you slice up the relevant power of P so as to
gain the relevant power of M , the latter will be a power to produce a different effect than the
former. The present strategy avoids overdetermination between non-identical token powers of
realized and realizing features, then, at the price of requiring that the powers of M and P are
not powers to produce the same effect. This may not be a killing objection; perhaps it is good
enough for purposes of vindicating the pretheoretic appearances that higher-level features can
(only) cause other higher-level features. But it does depart from the usual NRPist approach to
the problem of mental causation, which aims also to vindicate the seeming fact that higher-level
mental features can, at least sometimes, produce lower-level physically acceptable effects.
37
References
Antony, Louise M. and Joseph M. Levine, 1997. “Reduction with Autonomy”.
Philosophical Perspectives, 11:83–105.
Bird, Alexander, 2007. Nature’s Metaphysics: Laws and Properties. Oxford:
Oxford University Press.
Chalmers, David, 1996. The Conscious Mind. Oxford: Oxford University Press.
Clapp, Lenny, 2001. “Disjunctive Properties: Multiple Realizations”. Journal of
Philosophy, 98:111–136.
Clarke, Randolph, 1999. “Nonreductive Physicalism and the Causal Powers of
the Mental”. Synthese, 51:295–322.
Davidson, Donald, 1970. “Mental Events”. In L. Foster and J. Swanson, editors,
Experience and Theory. Amherst: Massachusetts University Press. Reprinted
in Davidson 1980/2001.
Davidson, Donald, 1980/2001. Essays on Actions and Events. Oxford: Oxford
University Press.
Ehring, Douglas, 1996. “Mental Causation, Determinables, and Property Instances”. Nous, 30:461–480.
Ehring, Douglas E., 2003. “Part-whole Physicalism and Mental Causation”.
Synthese, 136:359–388.
Ellis, Brian, 2001. Scientific Essentialism. Cambridge: Cambridge University
Press.
Fodor, Jerry, 1974. “Special Sciences (Or, The Disunity of Science as a Working
Hypothesis)”. Synthese, 28:77–115.
Gillett, Carl, 2002. “The Dimensions of Realization: A Critique of the Standard
View”. Analysis, 62:316–323.
Horgan, Terence, 1989. “Mental Quausation”. Philosophical Perspectives 3:
Philosophy of Mind and Action Theory, 47–76.
Horgan, Terry, 1993. “From Supervenience to Superdupervenience: Meeting the
Demands of a Material World”. Mind, 102:555–586.
38
Kim, Jaegwon, 1976. “Events as Propety Exemplifications”. In M. Brand and
D. Walton, editors, Action Theory, 159–77. Dordrecht: Reidel. Reprinted in
Kim 1993b.
Kim, Jaegwon, 1989. “The Myth of Nonreductive Materialism”. Proceedings and
Addresses of the American Philosophical Association, 63:31–47. Reprinted in
Kim 1993b.
Kim, Jaegwon, 1992. “Multiple Realization and the Metaphysics of Reduction”.
Philosophy and Phenomenological Research, 52:1–26. Reprinted in Kim 1993b.
Kim, Jaegwon, 1993a. “The Non-Reductivist’s Troubles with Mental Causation”.
In John Heil and Alfred Mele, editors, Mental Causation, 189–210. Oxford:
Oxford University Press. Reprinted in Kim 1993b.
Kim, Jaegwon, 1993b. Supervenience and Mind: Selected Philosophical Essays.
Cambridge: Cambridge University Press.
Kim, Jaegwon, 1998. Mind in a Physical World. Cambridge: MIT Press.
Kim, Jaegwon, 2006. “Emergence: Core ideas and issues”. Synthese, 151:547–
559.
Macdonald, C. and Graham F. Macdonald, 1995. “How to be Psychologically
Relevant”. In Philosophy of Psychology: Debates on Psychological Explanation,
update. Oxford University Press.
MacDonald, Cynthia and Graham MacDonald, 1986. “Mental Causes and Explanation of Action”. In L. Stevenson, R. Squires, and J. Haldane, editors,
Mind, Causation, and Action, update. Oxford: Basil Blackwell.
McLaughlin, Brian, 1992. “The Rise and Fall of British Emergentism”. In
Ansgar Beckerman, Hans Flohr, and Jaegwon Kim, editors, Emergence or Reduction? Essays on the Prospects of Non-reductive Physicalism, 49–93. Berlin:
De Gruyter.
Melnyk, Andrew, 2003. A Physicalist Manifesto: Thoroughly Modern Materialism. New York: Cambridge University Press.
Melnyk, Andrew, 2006. “Realization-based Formulations of Physicalism”. Philosophical Studies, 131:127–155.
39
Melnyk, Andrew, 2008. “Can Physicalism be Non-reductive?” Philosophy Compass, 3:128–1296.
Moore, G. E., 1903. Principia Ethica . Cambridge: Cambridge University Press.
Papineau, David, 1993. Philosophical Naturalism. Oxford: Basil Blackwell.
Paul, L. A., 2002. “Logical Parts”. Nous, 36:578–96.
Pereboom, Derk, 2002. “Robust Non-reductive Materialism”. Journal of Philosophy, 99:499–531.
Pereboom, Derk, in progress. Consciousness and the Prospects of Physicalism.
Pereboom, Derk and Hilary Kornblith, 1991. “The Metaphysics of Irreducibility”.
Philosophical Studies, 63:125–45.
Polger, Thomas W. and Lawrence Shapiro, 2008. “Understanding the Dimensions
of Realization”. Journal of Philosophy, 105.
Putnam, Hilary, 1967. “Psychological Predicates”. In Art, Mind, and Religion,
37–48. Pittsburgh: University of Pittsburgh Press.
Robb, David, 1997. “The Properties of Mental Causation”. Philosophical Quarterly, 47:178–94.
Shoemaker, Sydney, 1980. “Causality and Properties”. In Peter van Inwagen,
editor, Time and Cause, 109–35. Dordrecht: D. Reidel.
Shoemaker, Sydney, 2001. “Realization and Mental Causation”. In Physicalism
and its Discontents, 74–98. Cambridge: Cambridge University Press.
Shoemaker, Sydney, 2007. Physical Realization. Oxford University Press.
Swoyer, Chris, 1982. “The Nature of Natural Laws”. Australasian journal of
Philosophy, 60:203–223.
Wilson, Jessica, 1999. “How Superduper Does a Physicalist Supervenience need
to Be?” The Philosophical Quarterly, 49:33–52.
Wilson, Jessica, 2002. “Causal Powers, Forces, and Superdupervenience”. Grazer
Philosophische-Studien, 63:53–78.
Wilson, Jessica, 2005. “Supervenience-based Formulations of Physicalism”. Nous,
39:426–59.
40
Wilson, Jessica, 2006. “On Characterizing the Physical”. Philosophical Studies,
131:61–99.
Wilson, Jessica, 2009. “Determination, Realization, and Mental Causation”.
Philosophical Studies, 145:149–169.
Wilson, Jessica, forthcominga. “Non-reductive Physicalism and Degrees of Freedom”. British Journal for the Philosophy of Science.
Wilson, Jessica, forthcomingb. “What is Hume’s Dictum, and Why Believe It?”
Philosophy and Phenomenological Research.
Wilson, Jessica, in progressa. “Hume’s Dictum and Natural Modality: Counterfactuals”.
Wilson, Jessica, in progressb. “Metaphysical Emergence: Weak and Strong”.
Yablo, Stephen, 1992. “Mental Causation”. The Philosophical Review, 101:245–
280.
Yoshimi, Jeffrey, 2007. “Supervenience, Determination, and Dependence”. Pacific Philosophical Quarterly, 88:114–133.
41