<< . .

( 29)

. . >>

or the informational contents of two thoughts are really the same.

3 If it were not transparent to mind, then that one cannot think a contradiction about a
thing while thinking of it under just one mode of presentation could not be criterial of
sameness of sense, nor could it be assumed that uninformative identities are never false
identities. (If Frege™s position is not that identity of sense is transparent to mind, certainly
this is what many have thought his position to be. The purpose here is not, of course,
Frege exegesis but clarification of where certain incompatibilities of position lie.)


I have been arguing in this book for an externalist and representation-
alist position on thought, but it is crucial that we not rely on a lan-
guage-of-thought model of mental representation. This misleading
model offers an image of thought as having two levels of sameness
marking, on one of which sameness of content for concepts is not an
empirical issue but a matter of mere inner form. It pictures errors about
identity as impossible on this level and, correlatively, pictures the differ-
ence between valid and invalid mediate inference as though it were dis-
tinctly marked from within. But on externalist grounds, no distinction
can be clearly marked in thought between valid inference that relies on
false identity premises and invalid inference. There is no analytic/syn-
thetic distinction for identity as grasped in thought.
It follows, for example, that the Quinean distinction between tau-
tologies and merely analytic sentences and, similarly, the distinction cus-
tomarily drawn between inferences valid due to logical form and those
whose validity depends on the meanings of nonlogical terms, cannot be
drawn for thought. In a similar vein, we have a linguistic convention
that any adjective grammatically marked as comparative “ in English, for
example, by adding the suffix “-er” “ expresses a relation that is transi-
tive. This gives the appearance that thoughts expressed in the argument
form “A is φer than B and B is φer than C, therefore A is φer than C”
are valid a priori. But, as Hemple showed us clearly, using the geologist™s
“harder than” as his example (Section 7.2), particular applications of this
form are not valid a priori. They are valid only in case the English lan-
guage happens to conform, in particular cases, to the convention, and
conformity cannot be guaranteed, exactly because the transitivity of an
empirical relation cannot be known a priori. The linguistic form of the
argument makes it valid by linguistic convention, but there can be no
guarantee that the convention is manifested in particular cases. Similarly,
Whitehead claimed that it is always an empirical matter, in the particu-
lar case, that one plus one equals two. One raindrop plus one raindrop
sometimes equals one raindrop, and one quart of water plus one quart
of alcohol equals less than two quarts. The convention is to use num-
bers and numerical operators only where number theory applies, but
that it applies in particular cases is known empirically.
Putting the inevitably aposteriori nature of our grasp of sameness in
John Campbell™s terms (1987/88), there is no such thing as completely
“manifest sameness of reference.” The mental sentence image causes us

to overlook the most central fact about cognition, namely, that its most
difficult job is to get the empirical identities right, to create a coherent,
nonredundant and nonequivocal mental representational system. With-
out such a system, or something sufficiently close, there can be no con-
ceptual thinking at all.
There can be no representation of sameness in thought without
sameness marking, and there are no substance concepts without repre-
sentings of sameness. Indeed, what substance concepts are initially for is
grasping, which requires somehow marking, sameness in substances. But
since there also is no difference between marking sameness and fixing
identity beliefs, it follows that there are no representations of substances
that are free from the possibility of empirical error. It is always contin-
gent that a substance concept represents univocally, or represent at all.
We must proceed very carefully here, however. It does not follow that
substance concepts somehow make claims, or that they are “theories,” or
that they really are identity judgments in disguise.
There are lots of ways to do things right rather than wrong without
making claims or holding theories. You don™t make claims when you
stand up to walk just because it™s possible you could trip and fall. Simi-
larly, you don™t make claims when you develop substance concepts or
when you mark identities in thought. Erroneous identification is not
failure on the level of know-that but failure on the level of know-how.
It is failure in an activity. Standing back from a failed activity it is often
possible to explain its failure by pointing to some proposition that, had
it only been true, would have prevented the failure. Had that wrinkle in
the rug not been there, I would not have tripped. It doesn™t follow that
my attempt to walk involved a judgment that no wrinkles were in the
rug. Similarly, when representations carrying information about differ-
ent substances are wrongly coidentified, it is true that had they carried
information about the same they would not have been coidentified
wrongly. It does not follow that the act of coidentifying, or of identity
marking, is a judgment that they carry information about the same. It is
not a judgment about sameness of content.
Rather than substance concepts being implicit judgments or theo-
ries, it is better to say that, as distinguished from an identity sentence or
assertion, there is no such thing as an identity judgment. It is not the job
of an identity sentence to induce a belief. Its job is to induce an act of
coidentifying.True, an identity sentence has a grammar superficially like
that of ordinary subject-predicate fact stating sentences. It has tradition-
ally been recognized, however, that it does not have a logical subject

and a logical predicate. Only if we insist on modeling thought on lan-
guage should there be a temptation to assimilate what an identity sen-
tence produces “ an act of coidentifying “ to what a subject-predicate
sentence characteristically produces “ namely, an intentional attitude.
Grasping an identity is not remotely like harboring an intentional at-
titude. Similarly, mistaking an identity is not harboring a false belief. It
is an error of its own kind. Misidentifying is not, in central cases, an in-
nocent act of false judgment, but an act that tends to muddy the very
content of the thought involved, corrupting the inner representational
system. The development and maintenance of relatively clear and dis-
tinct ideas is a substantive ongoing activity. Descartes was quite right
that not all ideas are clear and distinct, indeed, that some are materially
false. He went astray only in failing to see how much more than mere
armchair reflection is involved in the activity of clarifying our ideas
(Chapter 7).


Where Frege drew one distinction, between sense or mode of presen-
tation and reference, there seem really to be two distinctions, or anyway
two phenomena, confused together. First, there is the distinction be-
tween concept and conception, that is, between designating conceptual
abilities by their ends only and designating them by their means (or by
certain of their means) as well. Second, there is the possibility of having
more than one concept of the same, these concepts being separate and
not marked as of the same.
This second possibility gives rise to an embarrassment in terminol-
ogy. I have spoken of two different ways to understand the notion “same
ability” hence “same concept,” depending on whether “the same” means
the same end achieved or whether it means the same end achieved by
the same means (Sections 1.9 and 6.3). But it appears now that there is
also a third way to interpret the notion “same concept.” If concepts can
be called “the same” when they involve abilities to identify the same,
thus allowing you and me to have many of “the same” concepts, then it
will also be possible for me to have two different tokens of the “same
concept” because I have failed to coidentify two concepts of the same
thing. “The same” concept token, is identified neither by its end alone
nor by its end plus its means. If a person harbors two different tokens
of the same concept these will, of course, be governed by different con-
ceptions, by different means, but that is not what makes them two.

Similarly, someone might be said to have two separate abilities to
achieve the same end, where the point is not merely that they know
several ways to achieve that end. Suppose, for example, that before it
was known that scarlet fever is the same disease as rheumatic fever but
in a different form, there was a doctor who knew how to cure scarlet
fever by one means and knew how to cure rheumatic fever by another.
Such a doctor might be said to have two different abilities to cure strep
infections, two tokens of the same ability.
In sum, we should distinguish, first, among concept tokens, and sec-
ond, among two different sorts of concept types, types distinguished by
ends and types distinguished by ends plus means. It is also possible to
classify concepts more abstractly, according merely to some of their
means. But here we should proceed very cautiously
One of the things we have put in place of Fregean modes of presen-
tation is conceptions. I have described conceptions as the “means” by
which the thinker knows how to reidentify a substance. Is the word
“means” in this usage a singular noun, a plural noun or a mass noun?
Can we, for example, sensibly speak of “some of the means” of a con-
cept? In Section 11.5, I argued that in the case of recognitional abilities,
the means actually or possibly used to reidentify a substance cannot be
divided into discrete countable “ways of identifying.” “Means,” in this
context, seems to be a mass noun. Conceptions don™t for the most part
divide into discrete parts. The individual recognitional abilities that any
two people have have, allowing them to reidentify the same substances,
are very unlikely to be identical in means, nor are they likely to be de-
scribable in a readily available way.
On the other hand, some features of these means may be describable.
For example, I might be able to recognize Xavier by sight but not by
his voice on the phone, or recognize lemons by smell but not by their
name in Russian, or recognize Jon Jones by the fact that he is the one
chairing the meeting but not by scanning his face. Of course, that I rec-
ognize Xavior by sight does not tell how I recognize him by sight, nor
does my recognizing Jon Jones by his chairing the meeting tell how I
recognize which person is chairing. Mentioning certain features of the
conceptions I use is not telling the whole story about how I recognize,
even the whole story on some specific occasion. Still, it is possible,
sometimes, to describe conceptions by aspects of their means in this
rough sort of way, and this allows us to give certain kinds of psycho-
logical explanations.

Where a person™s reidentifying abilities have proved fallible, descrip-
tions of intentional attitudes referring only to the objects and properties
these concern, that is, purely referential descriptions of them, can be
misleading. Where it matters to psychological explanation, the normal
assumption is that relevant identities have been correctly recognized. If
they haven™t, or easily might not have been, we describe thoughts by
reference to rough aspects of the conceptions involved.Thus we explain
why Paul didn™t speak to the woman he admires, even though she was
present and he very much wanted to, by the fact that he didn™t know
what she looks like. Similarly, we may refer to identifying knowledge
that forms part of a person™s conception of an object, or make reference
to a name by which they recognize an object. In this way we may move
back and forth between purely “transparent” descriptions of a person™s
intentional attitudes and somewhat more “opaque” ones. Consider, for
example, a case with which I opened this book. It would certainly be
misleading to say, without further explanation, that someone at the Yale
Alumni Association headquarters wished to know whether I knew
where I was, but not misleading to say that they wished to know
whether Mrs. Donald P. Shankweiler knew where Ruth Garrett Mil-
likan was. From the fact that the conceptions governing the two differ-
ent concepts of me could be described individually in this manner it
does not follow, however, that the entire conception governing either
concept could be described. Nor does it follow that there could be a
description of an intentional attitude that was entirely opaque, making
no transparent reference to any objects or properties outside of the
thinker.4 Even knowing someone merely by their name is not having a
conception of them describable in completely opaque terms. “The abil-
ity to recognize the name X” is a transparent description of that ability.
So-called opaque descriptions of conceptions and intentional attitudes
are never more than semiopague.
Besides semiopaque descriptions of various means supporting real
substance concepts, there also can be, of course, semiopaque descrip-
tions of various “would-be” means but that fail to support real substance
concepts. That is, unbeknownst to their possessors, they are not the
means for any real recognition abilities. These substance concepts are
“empty” or, more accurately, they are not substance concepts at all. An

4 Various different ways of describing intentional attitudes are discussed in Millikan (1984,
Chapter 13).

ability that is not an ability to do anything is not an ability at all. Empty
substance concepts result from failures of the mechanisms designed to
develop substance concepts. They are “concepts” only in that their bio-
logical purpose was to have been concepts. Nor should they be con-
fused with concepts we merely pretend to have, such as the concept we
pretend to have of Santa Clause after we are grown. Semiopaque de-
scriptions can also be given of pretend concepts, of course, but that is
another matter.
Opaque descriptions characterize aspects of conceptions, that is, as-
pects of ways of identifying substances. They do not describe ways of
thinking of substances. “Via a definite description” and “via a proper
name” are not ways of thinking of things. Nor, of course, is “by recog-
nizing her face” or “by recognizing her voice” a way of thinking of a
person. A neo-Fregean tradition has it that perceiving a person is a sort
of way of thinking of them, indeed, it is supposed, an indexical way of
thinking of them “ thinking of them via an indexical mode of presen-
tation. This, I believe, is a serious confusion. Suppose that you see Alice
and track her perceptually, picking up information about her as you
proceed but without recognizing her as Alice. Then for the moment
you have two concept tokens of Alice that you have not coidentified.
Your current tracking ability supports a “naive Strawson mode of pre-
sentation,” if we may still use that terminology, which is separate from
the naive Strawson mode in which your prior knowledge of Alice is
stored. As we saw in Section 11.2, however, this mode is no Fregean
mode of presentation. Furthermore, once you have recognized Alice, no
distinction of “modes” of any kind remains.
To be sure, your current perception of Alice yields, in part, a special
kind of information about Alice, namely, information about her current
spatial relation to you. But that it yields information about her relation
to you has nothing to do with indexicality. This is a subject I have al-
ready addressed at length elsewhere, and will not pursue further here.
There are, I have claimed, no mental indexicals at all, least of all any (so-
called) essential indexicals (Millikan in press a).

Knowing What I™m Thinking Of 1

. . . for it is scarcely conceivable that we can make a judgment or entertain a
supposition without knowing what it is we are judging or supposing
about. . . . the meaning we attach to our words must be something with
which we are acquainted . . . [but] Julius Caesar is not himself before our
(Russell, The Problems of Philosophy, p. 58)

The difficulty with Russell™s Principle has always been to explain what it means.
(Gareth Evans, The Varieties of Reference, p. 89)


In Chapter 7, I offered an answer to the question: How do we know
when we are thinking of a substance, and thinking of it unequivocally
and nonredundantly? But I did not answer the question, equally urgent:
What, on an externalist account, could possibly constitute that one
knows what substance one is thinking about? In this chapter, I will try
to answer that question.
I will agree with Evans that grasping the identity of the object of
one™s thought requires having a concept of that object. I have already
agreed with him, throughout this book, that a (substance) concept is, in
part, an ability to reidentify its object. But abilities, I have said, can be

1 Parts of this chapter are revised from “On unclear and indistinct ideas” (Millikan 1994),
which appeared in Philosophical Perspectives, 8, Logic and Language, edited by James E.
Tomberlin (copyright by Ridgeview Publishing Co., Alascadero, CA). Reprinted by per-
mission of Ridgeview Publishing Company, with the kind permission of Ridgeview Pub-
lishing Company, and from “Knowing What I™m Thinking of ” (Millikan 1993c), reprinted
by courtesy of the Editor of the Aristotelian Society © 1993.

better or worse (Section 4.3). Especially, one can know how to do a
thing only under very restricted conditions or under a great variety of
conditions. Knowing what one is judging about is thus a matter of de-
gree. One can come to know better what one is judging about.
Also, as I have emphasized (Chapter 4), one can know how to do a
thing but still fail.The conditions required for successful exercise of one™s
ability may be absent, nor need one be aware of this absence. Russell and
Evans to the contrary, it is not uncommon to be mistaken about the ob-
ject of one™s thought on particular occasions. That is, even though you
do have an ability to identify the object of your thought, hence do know
what you are thinking of, you can still make mistakes about the object
of your thought. Similarly, having the ability to walk will not prevent
you from sometimes tripping. If not soon corrected, however, mistaking
the identity of an object of thought produces equivocation in thought,
hence the beginning, at least, of change in the object of thought.
In Chapter 14, I will examine “Russell™s principle” in another light,
asking whether there are other kinds of mental representation, the iden-
tities of whose intentional objects remain unknown to the thinker. This
will turn out to be the same as the question whether there are non-
conceptual mental representations.


To inquire whether it is possible to make a judgment or think about
something without knowing what one is thinking of we first need to
understand what it would be to know what one is thinking of. Exter-
nalism concerning mental content clearly implies that we cannot “know
what we are thinking about” in the strictest Russellian way. In Russell™s
view, what can be “thought about,” in the strictest sense is only what is
within or directly before the conscious mind. On a representationalist
view on the other hand, what is within the mind when one thinks of
an object is a representation of the object, not the object itself. Or if the
object should happen to be “in” the mind, for example, if it is itself a
mental representation, still it is not by being in the mind that it be-
comes an object of thought.Thinking of one™s thoughts cannot be sup-
posed to be thinking of or knowing in some completely different sense
than thinking about the empirical world. What thinking of something con-
sists in cannot be supposed to change with the object of thought. If one thinks
about one™s representations, this must be by means of other representa-
tions. Representations do not represent themselves. Similarly, on a rep-

resentationalist view, what the mind is “aware of ” when it successfully
represents an object is the object represented, not the vehicle in the
mind that represents the object. But if that is so, against Russell, Julius
Caesar may indeed be “before our minds” in the only sense that any-
thing can be “before our minds.” What is in our minds and what is be-
fore our minds must be sharply distinguished. We must not confuse the
vehicle of thought with its content (Chapter 8).
Nor can we interpret Russell™s dictum to mean, say, that I cannot
make a judgment about Alice unless I also judge that my judgment is
about Alice. Knowing that I am thinking of Alice is surely posterior
rather than prior to thinking of Alice. I cannot know that I am think-
ing of Alice unless I first think of Alice, any more than I can know that
I am hungry unless I am first hungry. Nor is knowing that I am think-
ing of Alice necessitated by my thinking of Alice. Knowing that requires
judging that, and judging that I am thinking of Alice requires the ca-
pacity to think about thoughts. But this is a capacity there is no reason
to suppose every thinker must have.There is evidence, for example, that
children don™t have this capacity until well after they acquire fluent
Again, consider what it would be to know that I was thinking of Al-
ice. Barring Russell™s view of thought as direct confrontation of mind
with object, this knowing could not involve directly comparing my
thought with Alice. Rather, I would have to think of my thinking and
I would have to think of Alice and perhaps also of the relation that
made the one a thought of the other. In any event, I would surely have
to think of Alice. But if thinking of Alice involves knowing that I am
thinking of Alice, and this requires thinking of Alice again, we have a
regress. It is not regressive (though I believe it is surely false) to claim
that it is necessary to have the capacity or the disposition, whenever my
thought turns to Alice, to think that I am thinking about Alice. But it is
not possible that actualizing this capacity should be constitutive of having
thoughts about Alice.


Gareth Evans was an externalist and he believed, nonetheless, that there
was a way of explaining “what Russell™s Principle means” that makes it
come out not only sensible but true (Evans 1982). A central move in
Evans™ analysis was interpreting “knowing what one is judging about,”
at crucial junctures, not as a kind of knowing that (as we have so far

been interpreting it) but as a kind of knowing how. As Evans understood
it, knowing what one is thinking of is having some sort of “ability” or
“ he conflated all these “ “capacity” or “disposition” or “knowing how.”
Evans held that knowing what one is judging about is “a capacity to
distinguish the object of [one™s] judgment from all other things” (1982,

<< . .

( 29)

. . >>

Copyright Design by: Sunlight webdesign