A Naturalistic Defense of Realism

A Naturalistic Defense of Realism
by Michael Devitt
The basic idea of this paper is that
arguments against realism depend on a
couple of different kinds of mistakes.
 The first mistake is in assuming that
realism is inherently bound up with
the correspondence theory of truth,
or with a reference theory of
meaning. Devitt thinks that this is a
very widespread error, and attempts
to correct it.
 The second mistake is in taking a
priori arguments more seriously than
the results of empirical science. In
other words, the second mistake is in
not adopting a thoroughgoing
Devitt’s definition of Realism
Realism: Tokens of most common
sense, and scientific, physical types
objectively exist independently of the
This definition is not in any way
idiosyncratic. It captures the basic idea
behind realism, viz., that the world exists
independently of human experience,
and that the concepts we use to refer to
the things in the world, both common
sense and scientific, really exist.
(Note: familiarize yourself with the typetoken distinction. A token is simply a
particular instance of a type. Brenda is
a token of the type “woman”, Everest is
a token of the type “mountain”, etc.)
Since we have just finished reading
Sellars, you may be surprised to see
that Devitt is not involved in choosing
between the objects of the manifest and
the theoretical objects of science: he
explicitly accepts them both.
What he specifically rejects is captured
as follows (p.2):
In insisting on the objectivity of the
world, Realists are not saying that it is
unknowable. They are saying that it is
not constituted by our knowledge, by our
epistemic values, by our capacity to
refer to it, by the synthesizing power of
the mind, by our imposition of concepts,
theories, or languages; it is not limited
by what we can believe or discover.
He then goes on to identify many
philosophers who explicitly do not
qualify as realists by this definition:
Berkeley, Kant, Dummet, Goodman,
Kuhn, Putnam. Some, like van Frassen
are common sense realists, but not
scientific realists. Others, like Sellars
(who gets no mention here) are
scientific realists, but not common sense
One of the most important things to
understand about Devitt’s definition is
that it makes no use at all of the concept
of truth. This is important. We can, if
we want to, identify a perfectly
acceptable definition of realism that
does use the word true. For example:
Most common sense, and scientific,
physical existence statements are
objectively and mind-independently true.
The problem in doing this is that it
makes people think that realism is
inherently tied up with some conception
of truth, viz., the correspondence theory
of truth. This is wrong, and Devitt goes
to some length to demonstrate its
The easiest way to understand his point
is to observe that the second definition
really adds nothing to the first. To see
this, just look at two simpler sentences
 Cats meow.
 “Cats meow” is true.
We learn nothing from the second
sentence that we did not already get
from the first. The simple use of the
word “true” does not add any
information. It is just a way of conveying
the original information by talking about
sentence, rather than the world.
Note that neither of the above
sentences is equivalent to this sentence:
 “Cats meow” is true in virtue of
corresponding to the fact that cats
This statement invokes the
correspondence notion of truth, and it
purports to tell us what is actually
involved in a particular statement being
Devitt points out that
 this statement is entirely consistent
with realism, but
 it does not imply realism, nor does
realism imply it, and in fact
 the correspondence theory of truth is
actually compatible with any
metaphysics at all.
o What does he mean by this?
Well, for example, if you are an
idealist you could still assert that
truth of a statement consists in a
correspondence relation between
that statement and the fact that
the statement represents. It is
just that you understand the fact
in ideational, rather than, say,
physical terms.
Devitt offers a kind of hand-waving
explanation for why so many very smart
people have been confused about such
an apparently simple point. He thinks
that philosophers’ metaphysical
intuitions got really messed up by the
“linguistic turn” in philosophy: too much
talk about language, too little talk about
It is important to see that Devitt is not
dismissing talk about truth, or even the
correspondence theory of truth (to which
he apparently subscribes). Rather, he
thinks that people just don’t really
understand what theories of truth are
really for. So he tells us (p. 6)
Correspondence truth is a semantic
doctrine about the pretensions of one
small part of the world [i.e. language] to
represent the rest. This doctrine is the
subject of lively debate in the philosophy
of language, philosophy of mind, and
cognitive science. Do we need to
ascribe truth conditions to sentences
and thoughts to account for their roles in
the explanation of behavior and as
guides to reality. Do we need reference
to explain truth conditions. Should we
prefer a conceptual role semantics? Or
should we, perhaps.. eliminate meaning
altogether. These are interesting
questions… but they have no immediate
bearing on scientific realism.
Why Be a Common-Sense Realist?
In this section Devitt summarizes the
history of philosophy insofar as it gives
us reasons to be doubtful about realism.
Essentially this all boils down to the
‘gap’ between the external world and its
representation in the human mind, and
the perennial question how we could
ever really know that we are
representing this reality correctly.
The traditional failed attempts to bridge
this gap have been:
 Cartesian foundationalism, which
attempts to close the gap by building
all of our empirical knowledge on a
indubitable foundation.
 Lockeian empiricism, which holds
that there must be a resemblance
relation between the mind and
 Berkelian idealism, which actually
eliminates the gap by claiming that
the real world is not mind
independent at all.
 Kant’s a priori intuitions, according to
which the common sense world is
known because it is constituted by
our a priori intuitions, but the real
world is in principle unknowable.
 Machian phenomenalism, which, in a
Berkelian vein, attempts to translate
physical object language into
language about sense data (i.e., by
showing that a sentence like “The
cat is purring” has the same truth
conditions as a very complex
statement about catty purry sense
 Contemporary constructivism, which
is essentially Kantian without the
universality of the a priori intuitions.
All of these kinds of gap closures have
failed. Devitt groups them all as having
made the same basic kind of mistake,
and that is to have taken what are
essentially a priori speculations more
seriously than empirical research.
In other words:
Reflecting from the comfort of armchairs
[philosophers] decide what knowledge
must be like, and from this infer what the
world must be like.
The mistake, then is that philosophers
proceed as follows
 A priori epistemology → a priori
Instead of
 Empirical metaphysics → empirical
On Devitt’s view philosophy has simply
failed to produce a plausible
metaphysics because it begins with a
priori speculation about the nature of
knowledge. But if we are naturalists,
then we simply begin with the ontology
of successful science, and develop our
epistemology on that basis.
From this point on, Devitt takes some
contemporary philosophers (Dummett
and Kuhn) to task for developing highly
implausible anti-realist views on the
basis of a priori theories of meaning.
Again, the problem is the direction of the
inference. On a naturalistic view,
theories of meaning should not precede
our metaphysics. Rather, a theory of
meaning should be the result of
scientific inquiry, which begins with the
basic ontology of science.
Starting from scratch
Cartesian methodology requires us to
take absolutely nothing for granted. But
what we learned from Descartes and
others is that when you take absolutely
nothing for granted, you are condemned
to skepticism. The question, then, is not
whether you should take anything for
granted, but what to take for granted,
and the naturalistic answer is that you
take for granted is our best science.
This, of course, does not mean that our
best science is beyond doubt. Rather, it
means that you have to be given a
reason to doubt it. And, the fact that it is
not beyond doubt is not such a reason.
Why be a scientific realist?
There are two basic arguments for
scientific realism, i.e., realism about the
postulated entities of science, both of
which Devitt endorses.
Argument 1
We ought to believe in unobservables
because supposing them to exist allows
us to give good explanations of the
behavior and characteristics of observed
Argument 2
We ought to believe in unobservables,
because science using observables has
been successful, and the existence of
the observables provides the best
explanation of the success of the
science that uses them.
(Put differently: If unobservables don’t
really exist, then why are the scientific
theories that make use of them so
successful? Anti-realism leaves this
entirely unexplained)
These arguments both lean heavily on
the concept of explanation. They
depend on something like the following
 If theory T provides a good
explanation of phenomenon P, then
the basic entities of T should be
supposed to exist.
This sounds good on the surface,
perhaps a little too good. The problem
is that to provide a good explanation, at
least on one definition, just is to reveal
the underlying causal structure of the
world. But on this definition, whether
scientific theories give good
explanations is actually the very point at
A more careful way of expressing the
idea in both of the above arguments is
to observe that scientific theories have
been highly successful at achieving the
predictive goals of scientific inquiry.
Surely it is reasonable to ask for an
explanation of this predictive success,
and the best explanation we have is that
the entities really exist and that they
really behave as our theories say they
do. Formulated as an alternative
 For any theory T that consistently
provides accurate predictions of
previously unknown phenomena, the
best explanation of this predictive
success is that the basic entities of T
The problem with a claim like this is that
it’s really not clear what kind of
explanation it is. For Devitt, as a
naturalist, it would evidently have to be
a scientific explanation, i.e., subject to
the same predictive criteria as any other
scientific theory. This means that,
among other things, it should generate
testable predictions itself. But, about
the only sort of explanation generate by
T is that it will continue to meet with
predictive success in the future. This
certainly lacks the kind of precision we
normally expect of scientific theories,
but, given the nature of the claim,
perhaps it is all we require.