Talk (June 15) at “European Congress of Analytic
Philosophy 4”, Lund, June 14-18, 2002
Ingvar Johansson
0. Explaining the
title.
The title of my talk is “Hare’s Supervenience
Revisited and Re-explicated”. It is in no sense meant as an obituary. As you
may know, Richard M. Hare died on the 29th of Januray this year.
Many times, it has been said that although the term supervenience entered modern
philosophy in 1952 with Richard M. Hare’s book The Language of Morals, the idea
of supervenience was introduced 30 years earlier in George Edward Moore’s
paper “The Conception of Intrinsic Value”. I have no qualms about such a
description, but since I find Hare more clear than Moore, I will primarily try
to explicate Hare.
Apart from issues of clarity, the essential
difference between Moore and Hare is that whereas Moore applies the idea of
supervenience to value properties, Hare applies it to evaluative words.
According to Moore, non-natural properties like ‘being good’ and ‘being beautiful’
supervene on natural properties; according to Hare, evaluative words like “good” and “beautiful” have a
supervenient character in relation to purely descriptive words.
This difference is still part and parcel of the
literature concerned with the supervenience relation. There is one group of
philosophers who prefer to discuss the supervenience relation when its relata
are entities like words, concepts, predicates or theories; and there is another
group who prefer relata like properties, relations or states of affairs.
Normally, I belong to the latter group. For
instance, I am interested in the supervenience relation as a relation that is
relevant to the mind-body problem. My re-explication of Hare, however, is
intended to be independent of what kind of relata the supervenience relation
takes. First, I will show that well-known modern definitions of supervenience
are not at all true to Hare’s original idea. Then, I will point out that Hare’s
idea of a supervenience relation, now rightly understood, is, Hare
notwithstanding, very interesting also when its relata are realistically
conceived properties.
1. Desiderata of
supervenience
Below, you can find a list of proposed
desiderata of supervenience that I have assembled from “the supervenience
literature”.
1.
Supervenient properties are determined by their base properties.
2. Supervenient properties are dependent on their base properties.
3. An entity has supervenient properties in virtue of its having base
properties.
4. Base properties underlie supervenient properties.
5. Base properties realise supervenient properties.
6.
Descriptions of base properties do not entail descriptions of supervenient
properties.
(the non-entailment requirement)
7.
Supervenient properties cannot possibly exist without being connected to base
properties.
(the existential dependence
requirement)
8. A
supervenient property may have different base properties.
(the multiple realisability
requirement)
9. If two
entities have the same base properties, then necessarily they have the same
supervenient property; or, base
property indiscernibility entails supervenient
property indiscernibility.
(the indiscernibility requirement)
10. If two
entities have different supervenient properties, then necessarily they have
different base properties; or,
supervenient property difference entails base
property difference.
(the converse indiscernibility
requirement)
Because of my interests in ontological realism,
I have chosen to formulate all the proposed desiderata with properties as the
relata of the relation. I have presented it mainly in order quickly to create
associations in your minds. However, before I take the list away, I would like
you to try to memorise especially desiderata number 6, 7 and 9. These
desiderata I call the non-entailment requirement, the existential dependence
requirement, and the indiscernibility requirement, respectively.
Note that the desiderata 9 and 10 are logically equivalent. Both of them
entail that there cannot possibly be a many-to-one
relation between a supervenient property and the corresponding base
properties. A one-to-many relation is
allowed by both of them, but it is
made into a requirement only by desideratum 8; a desideratum which represents
the so-called multiple realisability of supervenient properties.
With respect to mind-body supervenience,
desideratum 8 says that the same kind of mental property can be borne by
different kinds of neural properties. However, often when in overviews it is
spoken of “the core idea of supervenience”,
this requirement of multiple realisability is neglected. Therefore, I will
leave it out of account, too. Usually, it is only the covariation spelled out
in desiderata 9 and 10 that is called the
core idea of supervenience; it
says that base property indiscernibility entails supervenient property
indiscernibility, or, alternatively, that supervenient property difference
entails base property difference.
Since, in my opinion, this requirement is only part of Hare’s core idea, I have
given it another name. I call it the indiscernibility requirement.
As far as I can see, there can be no doubt that
Hare put forward both desideratum 6 and desideratum 9. Below, you find two
quotations from The Language of Morals.
The non-entailment requirement:
“it is not the case that there is any conjunction C of descriptive characteristics such that to say that a man has C entails that he is morally good.” (The Language of Morals, Oxford University Press: London, 1969, p. 145).
The indiscernibility requirement:
“Suppose that we say ’St. Francis was a good man’ It is logically impossible to say this and to maintain at the same time that there might have been another man placed in precisely the same circumstances as St. Francis, and who behaved in them in exactly the same way, but who differed from St. Francis in this respect only, that he was not a good man” (ibid.).
In a paper called exactly “Supervenience”,
presented in 1980 (published 1984 in Aristotelian
Society Supp. vol. 58, pp. 1‑16), Hare presented some second
thoughts on supervenience. In this paper, I would say, it is even more clear
than in The Language of Morals that
he regards the non-entailment requirement as being one of two requirements on
supervenience.
2. David Armstrong and
David Lewis.
Perhaps here, in this room, no one doubts that
Hare put forward a non-entailment requirement on supervenience and, by so
doing, made his term ‘supervenience’ by definition connected to
non-reductionist positions. However, at least two very prominent analytic
philosophers, David Armstrong and David Lewis, have put forward definitions of
supervenience that completely neglect Hare’s non-entailment requirement.
Of course, they are free to do so, but I think
they should have made it clear to their readers that their concepts of
supervenience differ essentially from
Hare’s concept. Armstrong’s concept, but not
Hare’s, fit reductionist positions; and Lewis’ concept is in fact neutral with respect to issues of
reductionism versus non-reductionism.
The brief presentation of Armstrong’s and
Lewis’ concepts of supervenience that I will now put forward is a summary of a
paper “Critical Notice of Armstrong’s and Lewis’ Concepts of Supervenience”
that I have recently published in SATS – Nordic Journal of Philosophy
(no. 1, 2002).
Armstrong defines supervenience for entities in
general. His definition in A World
of States of Affairs looks as follows:
“We shall say that entity Q supervenes upon entity P if and only if it is impossible that P should exist and Q not exist, where P is possible” (A World of States of Affairs, Cambridge University Press: Cambridge, 1997, p. 11),
and
“supervenience in my sense amounts to entity P entailing the existence of entity Q” (ibid.).
Two observations can easily be made.
First, all
entities that conform to Armstrong’s definition of supervenience conform to
Hare’s indiscernibility criterion,
too. If Q Armstrong-supervenes on P, then necessarily if two particulars have
the same base entity P they have the same supervenient entity Q, too.
Secondly, no
entity that conforms to Armstrong’s definition can possibly conform to
Hare’s non-entailment criterion. In
order to apply the term ‘supervenient entity’, Armstrong requires entailment
whereas Hare requires non-entailment.
Armstrong’s concept of supervenience is wholly
distinct from Hare’s concept. Their extensions cannot have one single entity in
common. That follows from the definitions, and both Armstrong and Hare stick to
their definitions when they give examples.
If it could be shown that mental
properties supervene in Armstrong’s sense
on material properties, then it would be adequate to say that mental
properties have been reduced to
material properties; since then the mental properties are entailed by material
ones. On the other hand, if it could be shown that mental properties supervene in Hare’s sense on material properties,
then it would be adequate to say that mental properties have been shown not to be reducible to material
properties. Mental properties could then not possibly be entailed by material
properties.
Next,
Lewis’ concept. He defines supervenience in the following way:
“To say that so-and-so supervenes on
such-and-such is to say that there can be no difference in respect of so-and-so
without difference in respect of such-and-such” (Papers in Metaphysics and Epistemology, Cambridge University Press:
Cambridge, 1999, p. 29).
“Supervenience means that there could be no difference of the one sort
without difference of the other sort. Clearly, this “could” indicates modality”
(On the Plurality of Worlds, Basil
Blackwell: Oxford, 1986, p. 15).
Whereas
most writers on supervenience are using formulations like ‘base entity indiscernibility entails supervenient
entity indiscernibility’, Lewis is
mostly using formulations like ‘supervenient entity difference entails base entity difference’.
He is not using the indiscernibility requirement (desideratum 9), but the
logically equivalent converse indiscernibility
requirement (desideratum 10).
As you can see, Lewis does not in his
definition take Hare’s non-entailment requirement into account. The
indiscernibility requirement, which for Hare is merely one of two requirements on supervenience, becomes for Lewis the
whole definition.
With respect to Lewis-supervenience, I want you
to note one thing. If it is shown that a certain kind of entity is
Lewis-supervenient on another kind of entity, then this fact tells neither for nor against reduction, since
if the entity in question is Lewis-supervenient it can be either a case of reductionist Armstrong-supervenience or a case of non-reductionist
Hare-supervenience.
3. Jaegwon Kim and
Double-Modal Supervenience.
Now I will leave Armstrong and Lewis for quite
another strand of definitions of supervenience; the kind of definitions that
usually figures in conferences and anthologies dedicated specifically to the
supervenience relation. As a typcial representative, I have chosen Jaegwon Kim.
Mostly, in this tradition, the proposed definitions consider the supervenience
relation to be a relation that relates sets
of something. Kim, usually, has as relata of the supervenience relation
sets of properties.
If Lewis’ definition is reformulated to fit
this tradition, it would look as written down in my definition of the concept
of “single-modal supervenience” below. It will soon become clear why I
want to stress that it contains only
one modal operator:
Single-modal supervenience: a set of properties S supervenes
on a set of properties B,
if and only if,
it is necessary that: any two
individuals x1 and x2 that have the same
properties in B have the same
property in S.
Before I continue, I would like to add that
Armstrong and Lewis seem never to have bothered themselves with the question
how, in fact, Hare defined his concept of supervenience. But Kim has, and I
guess that is one reason why he has never put forward such simple definitions
as those of Armstrong and Lewis.
In Kim’s latest book, Mind in a Physical World (1998, p. 9) he is using a concept of
supervenience, or strong
supervenience, that can be given the following general form:
Double-modal supervenience: a set of properties S supervenes on a set of properties B,
if and only if,
it is necessary that: (for any Sj,
if xn has Sj, then there exists
a base property Bi such
that xn has Bi, and it
is necessary that:
(any x that has Bi has Sj)).
Kim’s definition, in contradistinction to
Lewis’ definition, contains two modal operators. Since both operators can be
interpreted as being either some kind of logical
necessity or some kind of nomological
necessity, we have in effect four different definitions:
(a) nomological-nomological supervenience: a set of properties S
supervenes on a set of properties B, if and only if,
it is nomologically necessary that:
(for any Sj, if xn has Sj, then there exists
a base property Bi such
that xn has Bi, and it
is nomologically necessary that:
(any x that has Bi has Sj)).
(b) logical-logical supervenience: a set of properties S supervenes on
a set of
properties B, if and only if,
it is logically necessary that: (for
any Sj, if xn has Sj, then there exists
a base property Bi such
that xn has Bi, and it
is logically necessary that:
(any x that has Bi has Sj)).
(c) nomological-logical supervenience: a set of properties S supervenes
on a set of properties B, if and only if,
it is nomologically necessary that:
(for any Sj, if xn has Sj, then there exists
a base property Bi such
that xn has Bi, and it
is logically necessary that:
(any x that has Bi has Sj)).
In relation to these first three definitions, Hare would, I think, have said the following. Nomological-nomological supervenience is too weak, whereas both logical-logical and nomological-logical supervenience are too strong. The first concept is too weak since every supervenience relation should essentially contain some kind of logical relationship, and the latter two concepts are too strong because the covariation that is spoken of (any x that has Bi has Sj) should not in itself be this logical relationship. Hare is in fact very strict on this last point (see his paper “Supervenience”, pp. 9-11) .
What then about the fourth definition?
(d) logical-nomological supervenience: a set of properties S supervenes
on a set of properties B, if and only if,
it is logically necessary that: (for
any Sj, if xn has Sj, then there exists
a base property Bi such
that xn has Bi, and it
is nomologically necessary that:
(any x that has Bi has Sj)).
Logical-nomological supervenience cannot from a Harean point of view be dispensed with as easily as the other three double-modal definitions can. Of course, ‘nomological’ – here – has to be understood in such a way that it can qualify more than laws described by the natural sciences. In order to fit Hare’s example of supervenience, it has to be allowed to qualify descriptions of connections between values and value-making characteristics, too. But as far as I am concerned such a use of ‘nomological’ is no problem.
The problem, as I see it, is that not even this more complex definition is true to Hare’s simple categorical non-entailment requirement. That requirement has been replaced by a kind of hypothetical nomological requirement that is weaker than Hare’s requirement.
4. Triple-modal
supervenience
Having noted this, however, we can in a very
simple way amend Kim’s definition. We can just add Hare’s categorical non-entailment requirement to Kim’s double-modal
definition and then investigate what such a more complex definition may
contain. Obviously, the new definition will contain three modal operators. It looks as follows:
Triple-modal
supervenience: a
set of properties S supervenes on a set of properties B,
if and only if,
(i) for any Bi, it is
logically possible that: there exists an x such that Bix and,
for all Sj, ¬Sjx,
and
(ii) it is logically necessary that: (for any Sj, if xn
has Sj, then there exists a base property
Bi such that xn
has Bi, and it is
nomologically necessary that: (any x that has Bi has Sj)).
I think that this explication of supervenience
takes account of the whole of Hare’s idea of supervenience. It contains the
indiscernibility requirement as well as the non-entailment requirement. But it
contains something more. A third requirement on superveneince can be derived
from this definition. Let me show.
First observation, conjunct (ii) in the definiens entails statement (iia):
(iia) it is logically necessary that: for any Sj, if xn has Sj, then there exists a base property Bi such that xn has Bi.
This sub-requirement is a kind of existential dependence requirement. It says that every supervenient property is for its instantiation dependent on the instantiation of some other property, a base property. And it means that even desideratum 7 of my desiderata list (the existential dependence requirement that I asked you to remember) is satisfied by this definition.
Second observation, since conjunct (ii) entails statement (iia) the following is true:
(ii) Ù (i) |¾
(it is logically necessary that: for any Sj, if xn has Sj, then there exists a base property Bi such that xn has Bi) and
(it is logically possible that: (there exists an x different from xn such that Bix and, for all Sj, ¬Sjx).
(S is a variable for sets, Sj is a variable for elements in the set S; B is a variable for sets, Bi is a variable for elements in the set B; x is a varibale for individuals/particulars, x1, x2, and xn are names of definite particulars.)
What is this? Before answering, let me point out a third observation: conjunct (i), which I have so far called the non-entailment requirement, can equally well be called an existential independence requirement. It says that it is logically possible for the base properties to be instantiated independently of the supervenient properties.
This means that we have from the triple-modal definition of supervenience derived both a dependence requirement and an independence requirement. In other words, we have derived something that fits the label a requirement of one-sided existential dependence. A supervenient property is for its existence one-sidedly existentially dependent on base properties.
I think that in several philosophical respects this is a very important notion that is unduly neglected. However, there might be an implicit reason behind this neglect. The requirement (iia) represents no connection between determinate elements of the sets S and B. It says that each element of S is for its existence dependent on the existence of some element of B. In other words, each element of S is for its existence dependent on the disjunction of elements of B. Therefore, it might be called a “disjunctive dependence”.
Now, that seems to be an odd notion when speaking of realistically conceived properties. I want to make it perfectly clear that, like other realists, I do not believe in the existence of disjunctive properties. And, surely, if there are no disjunctive properties it is odd to think that there can be any disjunctive property dependencies. However, there is another option available to the realist.
If one regards the concepts of the sets S and B of the definition as merely convenient ways of representing two realistically conceived natural kinds (the elements being their variations), then one can claim that a supervenient natural kind S may be one-sidedly existentially dependent on a subvenient base natural kind B. Similarly, if one regards the concepts of the sets S and B as representing two realistically conceived property-determinables (the elements being their determinates), one can claim that a supervenient determinable (e.g. being a mental property) may be one-sidedly existentially dependent on a subvenient base property-determinable (e.g. being a material property).
Let me now summarise: Triple-modal property supervenience entails one-sided existential dependence. It entails desideratum 7, i.e. the requirement that supervenient properties cannot possibly exist without being connected to base properties. In my opinion, this is no mystery. This desideratum is implicitly in work both in Hare’s and in Moore’s writings. Hare simply takes it for granted that evaluational statements are put forward only in contexts where descriptive statements are around. Similarly, Moore just takes it for granted that moral goodness is a property that is ascribed to, or withdrawn from, entities that already have natural properties.
I take this triple-modal definition as being the true explication of Hare’s whole idea of supervenience. It takes into account his two explicit requirements as well as his implicit requirement.
However, my concluding comments that now follow are not at all dependent on whether or not this triple-modal definition really is an explication of Hare’s concept of supervenience. The mere fact that the definition is both well-formed and meaningful is in itself of philosophical interest.
With respect to the issue of mind-body supervenience, the first requirement in the definition states that zombies are logically possible, and the second requirement entails that ghosts and angels are logically impossible. This means that a claim to the effect that mental properties supervene on properties of matter contains a claim that nonreductive materialism is true. The definition shows that, whether or not such a claim is true, it is at least both well-formed and meaningful.
With respect to
the issue of supervenient non-natural moral properties, the first
requirement in the definition states that a world wholly without moral
properties is logically possible, and the second requirement entails that
free-floating moral properties are logically impossible. This means that a
claim to the effect that moral properties supervene on natural properties
contains a claim that there is a non-reductive axiology. The definition shows
that, whether or not such a claim is true, it is at least both well-formed and
meaningful.
This idea of a triple-modal supervenience,
which I also take to be Hare’s idea of supervenience, is an irreducibly
non-reductionist idea.