<< . .

( 29)

. . >>



The view of substance concepts I am offering is an uncompromisingly
externalist view. What makes a thought be about a certain substance is
nothing merely in the mind, nor any mere disposition of the mind, not
even a wide disposition, but the thought™s origin “ an external
causal/historical relation between the concept and the substance (Sec-
tion 4.8). But meaning externalism has recently come under heavy at-
tack on the grounds that it leaves thinkers in no position to know
themselves either what they are thinking about or whether they are
genuinely thinking at all. And indeed, the best-known externalist theo-
ries all do seem to have this consequence. There is no necessity for an
externalist thesis to have this consequence, however, and I propose to
show how to avoid it.
What is needed to counter this entirely reasonable complaint against
meaning externalism, I believe, is first an adequate account of what
would constitute knowing what one is thinking of. If you are directly
thinking about an external object, knowing what you are thinking of
obviously cannot be done as Russell once described it, by having the
object of thought literally within or before your conscious mind. Nor,
a fortiori, can it be done by simultaneously holding your thought, or a
thought of your thought, before your conscious mind, on the one hand,

1 This chapter draws heavily on the Tenth Annual Patrick Romanell Lecture (Millikan
1998c), with the same title, delivered to the American Philosophical Association, Decem-
ber 30, 1997.

and comparing it with its object, also held before the mind, on the
other. What on earth then could knowing what one was thinking about
possibly be? Gareth Evans devoted much of The Varieties of Reference
(1982) to this question, and I will devote much of Chapter 13 to com-
paring his and my views on the matter.
A second thing that is needed to counter this complaint against ex-
ternalism, I believe, is an adequate empiricist epistemology for empiri-
cal concepts, with which one™s particular externalist position must, of
course, be compatible. The externalism I have described implies, first,
that there is no a priori guarantee against reference failure for substance
concepts. It is always a priori possible that one™s substance conception is
not in fact connected to any real substance at all. Second, there is no a
priori guarantee against reference duplication for substance concepts, no
guarantee against unknowingly referring over again to exactly the same
thing with two substance concepts.Third, there is no a priori guarantee
against equivocation in reference, against thinking of two substances as
if one, merging them together in thought. Elsewhere I have argued at
length, moreover, that any kind of externalism proposed for any kind of
empirical concept will necessarily have these three consequences (Mil-
likan 1993a, Chapter 14). If this is so, then it is clearly incumbent on the
externalist to show how evidence for the nonemptiness, nonredun-
dancy, and univocity of our empirical concepts can be gathered through
The externalist is obliged to construct an empiricist epistemology of
concepts. This epistemology, I will claim, must be different from and
prior to traditional empiricist epistemologies, which are all epistemolo-
gies either of judgment or of theories taken as wholes. This chapter is
devoted to the task of constructing such an epistemology. I will develop
it entirely independently of the theses already laid down about sub-
stance concepts, as a story about empirical concepts generally. In the
end it will be apparent, however, that only a rather special kind of the-
sis on the nature of externalist meaning would be compatible with such
an epistemology. Indeed, the only one I know of is the present one, and
a similar theory for concepts of empirical properties presented in Mil-
likan (1984, Chapters 14“17).
The complaint against meaning externalism is often put in a rather
different form, however, a form in which it is not legitimate. It is said
that meaning externalism deprives us of “incorrigible access” to our
own thoughts and that this result is untenable. What seems to be meant
is that making the transition from having a thought to correctly repre-

senting to oneself that one has that thought is problematic on an exter-
nalist view in ways it is not problematic on an internalist view. But this
is clearly mistaken. To begin with, it is far from clear that we have “in-
corrigible access” to our own thoughts, if that means that we can™t be
mistaken about them. Not only the Freudian tradition but a host of
modern experiments on cognitive dissonance attest that we certainly
can be mistaken. Nor is it any consequence of externalism that access to
knowledge of our own thoughts must be through the same channels as
access to knowledge of others™ thoughts. Externalism does not prohibit
having a different and more direct way of gathering information about
your own thoughts. Sellars, for example, held that you could interrupt
your dispositions candidly to speak your mind and turn them into dis-
positions to tell yourself what you believe (1975). Indeed, barring the
Cartesian position that mind is epistemically transparent to itself “ that
the knowing of things mental just equals the being of these things “ ex-
ternalist and internalist would seem to have exactly the same problem
of explaining how one acquires concepts of the mental “ something
that we know small children don™t have “ and how one successfully ap-
plies them to oneself. Even if what thought is about were determined
within thought itself, that would not help us with how thought reflects
on itself, how it comes to know itself as object (see also Gibbons
1996).2 Of course, the externalist cannot tell a priori that a second-
order thought, a thought about the content of a first-order thought, has
content if she cannot tell whether the first order thought has content.
But that is a different problem. It is the problem I have already promised
to address in this chapter, not something new.
Bill Lycan has reminded me of one more sort of problem that has
been posed for the externalist about correctly representing one™s
thoughts to oneself. Suppose that I have been living on earth but wake
up on twinearth one day. After a time, many suppose, what used to be
my thoughts of water will metamorphose and become thoughts of twa-
ter.Then when I think that last year I thought that twater was wet I will
be wrong, for in fact what I believed last year was that water was wet.
On the contrary, I suggest, when more than one substance has been

2 I am grateful to Keya Maitra for focusing my attention on this point. More generally, it is
dismaying how many contemporary discussions of the nature of consciousness, qualia, in-
tentional attitudes, and so forth ignore the question how one gets from the supposed pres-
ence of this or that within the mind or consciousness to propositional knowledge of that

tracked under the very same concept, the concept has become equivo-
cal. Equivocation in thought is a very common occurrence, certainly
not one that should be ruled out by either internalist or externalist.
Each semester when I acquire a new class of freshmen I go through it
again, making embarrassing mistakes because I have got Johnny and
William or Susan and Jane mixed together in my mind. If they abduct
me to twinearth I will soon have an equivocal thought of water/twater
and I will be wrong when I believe I used to think that water/twater is
wet. My thought did not used to be equivocal. The same thing may
happen right here on earth if I know Dr. Peters for some time before
meeting, unbeknownst to me, his identical twin, Dr. Peters. It is surely
a strength rather than a weakness of externalism if it accounts for this
sort of phenomenon.
What really is a problem for externalism, as I have said, is that it implies
we cannot tell by a priori means alone (1) when our thoughts are empty
of content, (2) when we are thinking double, that is, when psychologically
separate thoughts of ours bear exactly the same contents, or (3) when we
are equivocating in thought, representing two different contents with only
one thought. The externalist owes us an account of how these various
things can be discerned empirically “ an empiricist epistemology for em-
pirical concepts. Such an account will not be obliged, however, to assuage
the insistently bleak Cartesian skeptic. Given naturalist premises, no theory
of mind could do that in principle. But in Kantian spirit, we are obliged
to show how it is possible that our meanings are tested through experience.


Traditional empiricism holds, of course, that our abilities to think of ex-
ternal objects and properties are acquired with the help of experience.
From this one might think it a short step to the view that ongoing ex-
perience is also used in testing and perfecting these abilities. The exter-
nalist challenge would then be to develop a theory of the nature of em-
pirical concepts that explains how this testing and perfecting is possible,
indeed, how it manages to be highly effective and efficient. Instead, the
best-known externalist theories of how thought gets its content make a
complete mystery of this matter.To ground our meanings, they seem to
suggest, we would need to make prior judgments about causal or his-
torical relations of our thoughts to their objects. Or we would need to

judge that the conceptual roles of our thoughts matched corresponding
relations among their objects. Such demands are regressive, of course,
requiring prior grounded concepts of the same objects “ also prior
grasp of a true theory of meaning and reference for thought.
Yet I think that a better externalist theory is surprisingly close at
hand. With just a tug and a tweak, it falls right out of the central twen-
tieth-century American tradition of philosophy of science and language,
beginning with the familiar story about theoretical terms told at mid-
century, for example, by Carnap, Hempel, and Sellars.3 On this story,
theoretical terms, such as “mass” “temperature” and “atom,” acquire
their meanings, first, from the place each holds in (what can be recon-
structed as) a formal system containing postulates and rules that fix their
intratheoretical relations to other theoretical terms and, second, from
their inference relations to observation sentences, or to sentences in a
prior theory already anchored to observation. These latter rules corre-
lating theoretical with observational sentences were termed “bridge
principles.” In opposition to Carnap and earlier verificationists, Hempel
then claimed that it was not possible to separate either the intratheoretic
laws or the bridge principles of such a theory into two distinct kinds,
meaning postulates or matters of definition on one side, empirical pos-
tulates or matters of experience on the other. For example, the mean-
ing of the geologist™s term “hardness” is determined partly by the in-
tratheoretic law postulating the relation harder than as transitive, but also
by the bridge principle that if one mineral scratches another it is harder
than the other. Together these two principles imply that the relation x
scratches y is transitive, a fact that is clearly empirical, yet neither princi-
ple is more definitional of the geologist™s concept hardness, nor more an
empirical fact about hardness, than the other. “Theory formation and
concept formation go on hand in hand; neither can be carried on suc-
cessfully in isolation from the other. . . . If . . . cognitive significance can
be attributed to anything, then only to entire theoretical systems”
(Hempel 1950, 1965, p. 113). A more familiar quote is from Quine,
who takes concepts of ordinary observable objects to be like theoretical
concepts: “Statements about the external world face the tribunal of sen-
sory experience not individually but only as a corporate body” (Quine

3 In “Empiricism and the Philosophy of Mind” (1956), Sellars lists among advocates of this
view also Braithwaite, Norman Campbell, and Reichenbach. A simple exposition of the
theory is in Hempel (1966) Chapter 6.

1953). In Quine, this thesis is again tied to the rejection of a clean dis-
tinction between changing your meanings and changing your empirical
beliefs, between the analytic and the synthetic.
One result of this midcentury doctrine was a disastrous semantic
holism. The meaning of each empirical concept was taken to be deter-
mined only through its position in a wide inference network contain-
ing numerous other concepts, indeed, perhaps all of one™s concepts. But
another result was the implicit emergence of the first genuinely empiri-
cist epistemology of concepts. If empirical meanings cannot be disen-
tangled from empirical theories, then if theories can be some more and
some less adequate so can concepts, and the adequacy of an empirical
concept will be tested through ongoing experience, not a priori. But
because of the holism, this empiricist epistemology of concepts is not
useful against current attacks on meaning externalism. It suggests that
we would need to arrive at the end of Peircean inquiry before know-
ing whether any of our concepts are adequate, hence before knowing
whether we are genuinely thinking about anything at all. We need, I
suggest, to take the baby from the bath, keeping the thesis that mean-
ings are tested empirically but discarding the holism. Another look at
the classical theory of theories suggests how this can be done.
The paradigm bridge principle bridging from observation to theory
was taken to correspond to an operation either determining a theoret-
ical property or measuring its numerical value. As a theory matures, it
was supposed, typically it accumulates more and more operational
bridge principles of this sort. For example, one would expect geologists
to accumulate more ways of determining hardness than by scratching, as
there are numerous different ways to measure temperature, distance, vol-
ume, pressure, and mass. Given the Hempelian position, none of these
operations will be more central than others in determining the mean-
ing of the theoretical concepts they collectively define. Take, then, any
such set of operational bridge principles helping to define a single the-
oretical proposition and consider it in isolation from the rest of its en-
compassing theory. Consider, for example, all the known ways of deter-
mining that a certain thing™s temperature is 40° centigrade. Surely the
convergence of all these ways to the same result when applied to the
same physical object attests to the reality of the property temperature 40°
centigrade quite independently of our knowing any intratheoretical laws
about temperature. The objectivity of concepts such as temperature or
mass or length is strongly evidenced quite separately from theories em-
ploying these concepts. Or better, one™s readiness to judge the same

proposition true on multiple observational bases itself constitutes a sort
of minitheory, namely, the theory that if p then p, for exactly that one
proposition. The minitheory is confirmed when a variety of empirical
methods consistently converge on this single result.
It will be objected perhaps that new operations determining inter-
esting physical properties typically are known to determine them only
through the application of theory. But surely, here it does help to distin-
guish the context of discovery from the context of justification. How a
measuring method is discovered and how it is explained are neither of
them relevant to confirming its accuracy.That the measure is good, that
it correlates with other measures, is fully compatible with failure of the
theory that predicts and attempts to explain this fact.
For nontheoretical concepts, the ability to make the same perceptual
judgment from different perspectives, using different sensory modalities,
under different mediating circumstances such as different lighting and
acoustic conditions, offers similar evidence for the objectivity of the con-
cepts employed in these judgments. Holism is easily avoided in the epis-
temology of concepts so long as there exist empirical propositions, each
of which can be judged by a variety of independent methods not em-
ploying prior empirical concepts, making possible an independent test of
the concepts contained in that particular proposition. It is not the job of
empirical concepts to help predict experience.We do not predict our ex-
perience. We predict what we will read off our experience, namely, that
since p then p. We do not predict the appearances of things. We could
not possibly do so, for we cannot predict the ever-changing conditions
under which we observe them.We predict only the truth of distal facts.4


It is important in this context not to entangle the epistemology of con-
cepts with the epistemology of judgment. Consider, for example, the
following passage from Wilfrid Sellars:

. . . if [having the concept of green] presupposes knowing in what circum-
stances to view an object to ascertain its color, then, since one can scarcely de-
termine what the circumstances are without noticing that certain objects have

4 On this particular point, Quine (1960) seems to have had it the right way around: “Our
prediction is that the ensuing close range stimulations will be of the sort that vigorously
elicit verdicts of stonehood. Prediction is in effect the conjectural anticipation of further
sensory experience for a forgone conclusion” (p. 19).

certain perceptible characteristics “ including colors “ it would seem that one
couldn™t form the concept of [such things as] being green . . . unless he already
had them. It just won™t do to reply that . . . it is sufficient to respond when one
is in point of fact in standard conditions to green objects with the vocable
“This is green.” Not only must the conditions be of the sort that is appropri-
ate for determining the color of an object by looking, the subject must know
that conditions of this sort are appropriate . . . one can have the concept of
green only by having a whole battery of concepts of which it is one element.”
[Sellars 1956, p. 275.]

Sellars™ basic concern here is not that one couldn™t in point of fact learn
to respond discriminatively to green objects with the vocable “This is
green” in standard conditions without already having a battery of con-
cepts. Rather, it is that one couldn™t know that anything was green
without this. His concern is to ensure that suitable observation judg-
ments indeed express “knowledge” in the sense that they can be “placed
in the logical space of reasons, of justifying and being able to justify
what one says” (Sellars 1956, p. 299). But there is no cause to suppose
that the process of fashioning and honing adequate concepts presup-
poses the ability to justify the judgments that use these concepts.
Knowing about the conditions under which one™s perceptual and cog-
nitive systems will work properly is not required for learning how to
use them properly any more than knowing about the atmospheric con-
ditions required for breathing properly is required for breathing prop-
erly (compare Section 4.3). The epistemology of concepts is prior to
and not the same as the epistemology of judgments. Nor is it a crite-
rion of adequacy for an epistemology of concepts or, more broadly, for
a theory of mind, that it can lever a person out of skepticism. There is
no compulsion to suppose that human minds are so built that we can™t
possibly fall into epistemological black holes that can™t subsequently be
reasoned out of. The question we need to answer about our concepts
is how we do it, how we make them clear (cf. how we focus our eyes)
not how we can know that we have succeeded in making them clear.
Recognizing when one™s thoughts are clear is not making a judgment
about one™s thought, nor is it knowing how to justify one™s thought.5
Sellars™ final conclusion that “one can have the concept of green only
by having a whole battery of concepts” does not follow from the
premises he offers.

5 In Chapter 13, I will argue, likewise, that knowing what one is thinking of is not making
a judgment about one™s thought.

The mechanisms of perceptual constancy that enable us to perceive,
for example, the same color, shape, voice, or moving object as being the
same one through diverse proximal stimulations, diverse intervening
media, and various kinds of distortions and static, exemplify our ability
to make the same perceptual judgment in a variety of ways. So does our
ability to use different senses to confirm the same judgment perceptu-
ally. Given a variety of ways to observe the same state of affairs, none of
these methods is definitional of the concepts employed, just as on a
Hempelian view, no bridge principles leading from observation into
theory are more definitional of the theory™s concepts than others. None
of our ways of making the same judgment is distinguished as a or the
infallible method of judging its content. Each is but a practical ability,
more or less reliable, to identify the perceptually presented situation
correctly. Each relies either on historically normal conditions obtaining
for correct use of one™s perceptual mechanisms, or on historically nor-
mal conditions for observing that these normal conditions obtain. Or
better, being more careful, the most usual reliance is on normal condi-
tions obtaining for the support of successful epistemic action. One
knows how, physically, to maneuver oneself into conditions normal for
making accurate perceptual judgments of a given kind.
Highly consistent convergence of independent methods to the same
judgments serves as strong testimony to the objective univocal sources
of these judgments. For example, I check my perception by moving in
relation to the object, by employing others of my senses, by manipulat-
ing the object, to confirm a constant result. In so doing I not only ver-
ify my original result, but also confirm the more general abilities that
constitute, in part, the subject and predicate concepts on which my
judgment rests. I confirm them again when I find that another person
has arrived at the same judgment as I, another way of making judg-
ments being to believe what one is told (Chapter 6). Emptiness in em-
pirical concepts shows up characteristically in lack of variety in the per-
spectives from which they can be applied. Equivocation shows up in the
emergence of contradictions systematically correlated with perspectives
taken. Redundancy shows up, just as Leibniz said, with the accumula-
tion of coincident properties and the absence of contrary ones.
In Word and Object, Quine defines stimulus meaning as having two
parts, “affirmative” and “negative,” and he claims that “[t]he affirmative
and negative stimulus meanings of a sentence (for a given speaker at a
given time) are mutually exclusive” (Quine 1960, p. 33). This exclusiv-
ity results naturally from the fact that “stimulus meaning” is defined by

reference to overt affirmations and denials of a sentence coupled with
the assumption that a speaker won™t affirm and deny the same sentence
at the same time. Quine also remarks that “many stimulations may be
expected to belong to neither” the affirmative or negative stimulus
meaning. Notice that this “neither” category will, technically, cover
cases of total confusion as well as more ordinary “can™t tell” cases such
as those Quine explicitly includes as “poor glimpses.” The effect is that
Quine overlooks the most interesting cases relevant to concept forma-
tion, namely those in which contradiction bypasses theory and appears
directly at the level of observation. These are the cases having the most
leverage for testing meanings, but they are invisible given Quine™s tools

<< . .

( 29)

. . >>

Copyright Design by: Sunlight webdesign