Abstract:
Any property has two sorts of causal features: “forward-looking”
ones, having to do with what its instantiation can contribute to causing,
and “backward-looking” ones, having to do with how its instantiation can
be caused. Such features of a property are essential to it, and properties
sharing all of their causal features are identical. Causal necessity is thus a
special case of metaphysical necessity. Appeals to imaginability have no
more force against this view than they do against the Kripkean view that
statements like “Gold is an element” are metaphysically necessary.
1.
There was a time, within the memory of some of us, when it was widely
accepted that there are just two kinds of necessity. There was logical
necessity, which was generally construed as including the necessity of
analytic truths. This was assumed to be something to which we have an
a priori access. And there was causal necessity, to which our only access
is empirical. Since the dominant views about causality then current were
Humean in inspiration, there was some question as to how so-called causal
necessity, the distinctive status of causal laws and their consequences,
deserved the name of “necessity” at all. But that was what it was frequently
called.
Since then the boat has been radically rocked, first by Quine and then
by Kripke. The intended effects of their attacks on the traditional view
were of course very different. Quine’s attack on the analytic-synthetic
distinction sought to contract, if not to empty, the class of truths that are
called necessary. Kripke, on the other hand, argued that the class of truths
59
CAUSAL AND
METAPHYSICAL
NECESSITY
BY
SYDNEY SHOEMAKER
Pacific Philosophical Quarterly 79 (1998) 59–77 0031–5621/98/0100–0000
© 1998 University of Southern California and Blackwell Publishers Ltd. Published by
Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and
350 Main Street, Malden, MA 02148, USA.
deserving this label is much larger than had traditionally been supposed.
And, in his most radical departure from the traditional view, he held that
many of these truths have the epistemic status of being a posteriori. One
important class of these truths included statements of identity, such as
“Hesperus is Phosphorus” and “Water is H
2
O”. Another included state-
ments about the essences of natural kinds, such as “Gold is an element”
and “Tigers are mammals”.
Such truths were characterized as being true in all possible worlds. That
might suggest that they can be called logical truths. And that seems a
radical departure from the traditional notion of logical truth. But so also,
one might claim, was the earlier extension of the notion of logical truth
to cover analytic truths such as “All bachelors are unmarried”. If such
truths are logical truths, they owe this status to the fact that their truth
is guaranteed by certain paradigmatic logical truths – say that all
unmarried men are unmarried – together with semantic facts, say that
“bachelor” is synomymous with “unmarried male”. What Kripke showed
is that the class of semantic facts that can contribute to the bestowal of
the status of necessary truth is much broader than the class of synonymies
or analytic equivalences, supposing there are such; it includes such facts
as that the term “gold” refers to a substance with a certain essential nature.
I favor the usage that restricts the term “logical truth” to what logicians
would count as such, excluding both analytic truths like “Bachelors are
unmarried” and Kripkean necessities like “Gold is an element”. So I shall
refer to the latter as metaphysical necessities, rather than as logical
necessities. But this is not to say that metaphysical necessity is a weaker
kind of necessity than logical necessity. A statement can be metaphysically
necessary without being conceptually necessary, and without being logically
necessary. This is true of “Gold is an element”. But it is not plausible to
put this by saying that there is a possible world in which gold is not an
element, or by saying that the world might have been such that gold is
not an element. We can compare this with the fact that “Bachelors are
unmarried” is not, in the strict sense, a logical truth; we would not want
to put this by saying that the world might have been such as to contain
married bachelors.
So if Kripke is right, we have a set of truths, the metaphysically
necessary ones, that are necessary in the strongest possible sense, and yet
whose epistemological status is that of being a posteriori. Such truths are
knowable, if at all, only empirically. And this epistemological status they
of course share with statements of causal necessity. What, then, of the
traditional view that causal necessity, if it deserves to be called necessity
at all, is a weaker sort of necessity – that in possible worlds jargon, the
causal laws hold in only a subclass of the metaphysically possible worlds?
I think this view is widely held by philosophers who accept the Kripkean
view that there are necessities a posteriori. But of course there is another
60 PACIFIC PHILOSOPHICAL QUARTERLY
© 1998 University of Southern California and Blackwell Publishers Ltd.
possibility. Maybe the view that causal necessity is a weaker sort of neces-
sity, or necessity only in a stretched meaning of the term, is a holdover
from the pre-Kripkean view, which took it that genuine necessities, or at
least those of the strongest sort, are knowable a priori because a priori
reducible to truths of logic. Maybe, instead, causal necessity is just a
special case of metaphysical necessity, and is necessity in the strongest
sense of the term.
1
This is a view I defended in a paper published over fifteen years ago.
2
In that paper I put forward the view that properties are individuated by
their causal features – by what contribution they make to the causal
powers of the things that have them, and also by how their instantiation
can be caused. Collectively, causal features of this sort constitute the
essence of a property. So insofar as causal laws can be construed as
describing the causal features of properties, they are necessary truths. One
way to get the conclusion that laws are necessary is to combine my view
of properties with the view of David Armstrong, Fred Dretske, and
Michael Tooley, that laws are, or assert, relations between properties.
Views similar to mine have been defended by Chris Swoyer, and by
Evan Fales.
3
But I think it is fair to say that it is definitely a minority view.
The established view, even among those who have absorbed Kripke’s
lessons, is that causal laws are contingent, not just in the sense that their
epistemic status is that of being a posteriori, but in the sense that there
are genuinely possible situations in which they do not hold.
I shall not here repeat all of the arguments with which I supported the
causal theory of properties and the necessitarian view about laws –
although I will give a version of what I now take to be the central one.
My main concern here is to clarify this view by exploring some of the
sources of resistance to it.
2.
The source of resistance that most immediately leaps to mind lies in the
fact that we can easily imagine what it would be like to experience a world
in which the laws are different. We can imagine conducting crucial experi-
ments and having them come out differently than they do here. And as
Hume reminded us, we can imagine bread failing to nourish, water failing
to suffocate, and so on.
But, of course, in the sense in which we can imagine these things we
can imagine analyzing gold and finding that it is not an element, analyzing
water and finding that it is not H
2
O, dissecting tigers and finding that they
are reptiles, and so on. Most philosophers are now persuaded, by Kripke’s
arguments, that such imaginings are no real threat to the claim that it is
CAUSAL AND METAPHYSICAL NECESSITY 61
© 1998 University of Southern California and Blackwell Publishers Ltd.
necessary that gold is an element, that water is H
2
O, and that tigers are
mammals. Perhaps one wants to redescribe a bit what one actually
succeeds in imagining; e.g., let it be analyzing what looks like gold, or
what passes the layman’s and jeweler’s tests for being gold, and finding
that it is not an element – or analyzing what passes the ordinary
observational tests for being water and finding that it is not H
2
O. Then
one can grant that what one actually does imagine is possible, but say
that it is wrong to describe the possibility as that of gold not being an
element or water not being H
2
O. Some would prefer to stick to the original
description of what is imagined, but deny that imaginability establishes
possibility. In any event, the same resources are available for someone
who wants to maintain, in the face of all that we can imagine or seem to
imagine, that causal laws are metaphysically necessary. If one’s preferred
strategy for dealing with the Kripkean examples is to challenge the claim
that imaginability is proof or evidence of possibility, one can employ the
same strategy here. If one’s preferred strategy is to allow that imaginability
is at least evidence of possibility, but to claim that what we can really
imagine in these cases is not what we initially take it to be, that strategy
too is available here.
4
Let the law be that strychnine in a certain dosage
is fatal to human beings. We can grant it to be imaginable that ingesting
vast amounts what passes certain tests for being strychnine should fail to
be fatal to what passes certain tests for being a human being, but deny
that this amounts to imagining a human being surviving the ingestion of
that much strychnine.
But of course there was more to Kripke’s argument than deploying
such strategies to ward off challenges to his necessity claims that are based
on imaginability or seeming imaginability. He had direct arguments for
the necessity of identity propositions. He had a compelling case for the
view that names and other singular terms are rigid designators. And, what
is most pertinent here, he had a view of the semantics of natural kind
terms that both implied that some statements expressed by the use of such
terms are necessarily true and made it intelligible that such statements
have the epistemological status of being a posteriori. It was part of that
view that the semantic intentions underlying our use of natural kind terms
are such that the underlying make-up of a natural kind, the properties
responsible for the phenomenal features of the paradigm exemplars of the
kind, is essential to it. So, for example, if the paradigm exemplars of gold
are instances of an element having atomic number 79, it will be necessary
that gold is such an element. This explains why something lacking that
make-up can’t be gold, whatever its phenomenal features. And it therefore
explains why the imaginability of something having the phenomenal
features characteristic of gold but lacking that physical make-up is no
threat to the claim that it is necessary that gold has that make-up.
My claim will be that the causal theory of properties can provide a
62 PACIFIC PHILOSOPHICAL QUARTERLY
© 1998 University of Southern California and Blackwell Publishers Ltd.
parallel explanation of why the imaginability or seeming imaginability of
worlds in which the causal laws are different is no threat to the claim that
the causal laws that hold have the status of being metaphysically
necessary. Corresponding to Kripke’s claim that our semantic intuitions
are such that the underlying natures of natural kinds are essential to them
will be the claim that the causal features of properties are essential to
them. Just as we can refer to a natural kind, and identify instances of it,
without knowing what its essential underlying nature is, so we can refer
to a property, and pick out instances of it, without knowing what most
of its causal features are. Our ability to pick out instances of a property
perceptually, or with the use of instruments, will of course depend on
exercises of the causal powers which the property contributes to
bestowing. For example, it may involve the power to produce experiences
having a certain phenomenal character. But having the power to produce
such experiences will not in general be sufficient for having the property
in question; so the imaginability of something having such a power but
lacking any property having certain other causal features is compatible
with those other causal features being essential to the property in question.
3.
Before pursuing further the parallels between Kripke’s claims and the
claim that causal laws are metaphysically necessary, and before examining
further the imaginability or seeming imaginability of the causal laws being
other than they are, I must say more about what is, and what is not,
asserted by the causal theory of properties I favor.
In expounding my theory in my paper “Causality and Properties” I
made use of the notion of a “conditional power”: something has a
conditional power if it is such that it would have a certain power if it had
such and such properties, where the possession of those properties is not
itself sufficient to bestow that power. E.g., to use the example I used there,
something that is knife-shaped has, among others, the conditional power
of being able to cut butter if it is made of wood or steel. And I claimed
there that properties can be identified with “clusters” of conditional powers,
where the members of the cluster are causally unified in certain ways.
I now reject that formulation. One reason for rejecting it can be found
in a postscript to that paper. I there consider a case offered by Richard
Boyd, which purports to be a case in which two different properties bestow
the same cluster of causal powers. We are to suppose that X is a compound
of substances A and B and Y a compound of substances C and D, where
A, B, C and D are all different substances, and that it is a consequence
of the laws of nature that X and Y behave exactly alike in all possible
circumstances – so being made of X and being made of Y bestow exactly
CAUSAL AND METAPHYSICAL NECESSITY 63
© 1998 University of Southern California and Blackwell Publishers Ltd.