Externalism about Content and McKinsey-style Reasoning

Externalism about Content and McKinsey-style Reasoning
James Pryor
Harvard University
<jpryor@fas.harvard.edu>
Draft 1 — 10/1/01
WARNING: This is a first draft. I want to stress that the footnotes and citations
are just off the top of my head at this point, and may very well
misrepresent some other philosophers’ views.
I
It’s widely accepted nowadays that the contents of some of our thoughts are
externalist: we’re only able to have thoughts with those contents because we inhabit
environments of certain sorts.1 For example, to have thoughts about water it may be

1
Different philosophers mean different things when they talk about a thought’s content. I take a
thought’s content to be those of its representational properties which are essential to its being that thought.
These may or may not suffice to determine the thought’s truth-conditions. That is a substantive
philosophical question. I think that content does determine truth-conditions; but I don’t want to assume this
in the very definition of content. On some views, like Lewis’s [cite??], the thought I express by saying “My
pants are on fire” and the thought you express by saying “My pants are on fire” are the same thought.
Hence, those are views that allow thoughts with different truth-conditions to have the same content, in my
sense. Similarly, on some views, the thought I express by saying “Water puts out fires” and the thought my
twin on Twin Earth expresses by saying “Water puts out fires” are the same thought, thoughts with a single
content; it’s just that they have different truth-conditions in the different environments. I don’t count such
views as externalist views. As I understand externalism, an externalist has to say that, though my belief
and my twin’s belief may have interesting properties in common, they are nonetheless different beliefs,
with different contents.
One sometimes hears it said that “one and the same token thought can be ‘typed’ in different
ways.” It certainly is true that a single thought can exemplify many different types. The thought I express by
saying “My pants are on fire” is: a thought about fire, a first-person thought, a thought about Jim Pryor, a
false thought, and so on. But not all of these types are relevant to the thought’s identity. It is not essential to

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 2
necessary to inhabit an environment containing samples of water, which you or other
members of your community have causally interacted with.2 When you think to yourself
, the content of your thought is not available to people who live in
Twin Earthian environments that contain XYZ instead of water.
On some views, the contents of our demonstrative thoughts, and thoughts we
express using proper names, are externalist in the same way. If I see a man walking down
the street and I think to myself
, the content of this thought is not
available to someone who lives in an environment that that man never inhabited or left
traces in.
If we accept those kinds of externalism, then we have to confront a puzzle that
arises about our ability to know what we’re thinking. This puzzle was first[??]
articulated by Michael McKinsey in 1991. It goes as follows. It would seem that you can

my thought that it be false, for example. When philosophers say that my thought “can be ‘typed’ in
different ways,” they’re usually making a claim about the thought’s identity conditions. According to one
way of “typing” my thought that my pants are on fire, it would count as “the same thought” as your thought
that your pants are on fire; according to another, they would count as “different thoughts.” And both ways
of “typing” the thoughts would be correct. (Or perhaps each would be correct in certain settings.) In my
view, this ecumenical attitude towards identity conditions has little to recommend it. Certainly the mere
fact that it’s controversial what the identity conditions of our thoughts are does not by itself suffice to make
the position plausible. In any event, as I understand externalism, it does take our thoughts to have definite
identity conditions, conditions which cannot be satisfied unless one’s environment is of a certain sort.
2
It is not easy to specify what the relevant environmental features are, that externalism says are
necessary for you to think about water. According to Burge [[Other Bodies]], it would be possible to have
thoughts with the content
, even if no samples of water ever existed—but only if you or
someone else in your community had some beliefs about what water’s underlying nature is. [[Cite
Mendeleev / Ekaboron story]] If this is right, then assuming you do not yourself have beliefs about water’s
underlying nature, the way your environment would have to be, for you to think about water, is: it would
have to contain either samples of water or other people who had beliefs of certain sorts. See Brown [[??]]
for discussion. These complications will not affect the main points in my discussion, so to keep things
simple, I will work with the condition I give in the text.

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 3
usually tell the contents of your thoughts, just by introspection. For example, you can
know just by introspection that you’re thinking that water puts out fires:
McK-1
You’re thinking a thought with the content
.
And it seems that our knowledge that externalism is true, and hence that you can’t have
thoughts about water unless you inhabit a certain sort of environment, is a deliverance of
a priori philosophical reflection on Twin Earth thought-experiments. So it seems that you
can know a priori:
McK-2
You couldn’t have that thought unless your environment is a certain
way, e.g., some same samples of water must have sometime existed
and you or other members of your community must have causally
interacted with them.
But if that’s right, then it seems like you should be able to put those two pieces of
knowledge together, and conclude:
McK-3
Your environment is the relevant way, e.g., some samples of water
have sometime existed…
So it looks like introspection and philosophical reflection would be enough to enable you
to know that you inhabit the relevant sort of environment. And that is a puzzling result.
Prima facie, it seems quite counter-intuitive that you should know those sorts of facts
about your environment just on the basis of these kinds of reflection.
Let’s be clear about what the source of the puzzle is. On the one hand, we have a
certain argument: the argument from McK-1 and McK-2 to McK-3. There’s nothing
wrong with that argument, per se. In fact, it’s a sound argument. Its conclusion is true and
we know it to be true. And in itself, there doesn’t seem to be anything objectionable about
believing that conclusion on the basis of those premises. This only becomes puzzling
because of the kind of justification you happen to have for those premises. You seem to
be able to know McK-1 by introspection, and you seem to be able to know McK-2 a
priori, so it looks like you’re in a position to know the conclusion McK-3 just on the
basis of introspection and a priori reflection. That is what seems puzzling.3

3
Some philosophers doubt that philosophy is a purely a priori enterprise. This might engender

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 4
The puzzle is sometimes formulated as a problem about whether externalism
would enable us to have “a priori knowledge” of what our environment is like.4 Authors
who speak this way are counting introspection as a kind of “a priori knowledge.” I think
that’s quite misleading, and encourages certain confusions. We will be disentangling
some of those confusions later in this paper. For clarity, I will not count introspection as a
kind of a priori knowledge. I will use the more general expression “knowledge by
reflection” to talk about things you know either by introspection or by a priori reasoning,
or by a combination of the two. So in my terms, the puzzle says that you can know
certain facts about your environment just on the basis of reflection, and that’s what seems
incredible.
The puzzle is sometimes formulated as a problem about whether we have certain
kinds of epistemic authority about our environment, authority which we would have
thought we had only about our own mental states. But it is controversial and unclear what
kinds of authority we have about our own mental states. And I don’t think the current
puzzle really depends on any specific assumptions about that. It’d already be very
puzzling if we could have knowledge about our environment from the reflective sources
we’re discussing—even if no special epistemic authority came with that knowledge.
If that’s right, then it wouldn’t be a satisfying resolution of the puzzle merely to
say that introspection is perceptually defeasible, e.g. by evidence that my environment
has never contained any samples of water. Nor do I think the puzzle requires thinking of
introspection as being indefeasible. Let’s suppose for the sake of argument that my
introspective awareness that I’m thinking of water is defeasible. How would that help? It
might be thought to take the sting out of saying that we can know about our environment

doubts about whether our justification for believing McK-2 is purely a priori. But even if we accepted that
philosophy isn’t a priori, it would still be puzzling if you could come to know that your environment
contains water, just by a combination of introspection and philosophy. Even if doing philosophy requires
you to have certain kinds of empirical evidence, presumably it doesn’t require you to have perceived water,
or anything like that.
4
[[Authors who speak this way]]

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 5
just on the basis of introspection and a priori reasoning. For now that knowledge will be
defeasible in the ways we think knowledge about our environment ought to be. But I
don’t think it really does resolve the puzzle; at least, not by itself. For as I said, it already
seems puzzling to say that we can know about our environment from reflective sources
like introspection and a priori reasoning. That is already puzzling, without the need for
any further assumptions about defeasibility.
There are a number of ways one can respond to this puzzle.
One view endorses the conclusion that you can know things like McK-3 by
reflection alone. That is one way of construing Putnam’s argument in Ch. 1 of Reason,
Truth, and History. Putnam argues that people who have always been brains in vats can’t
refer to or think about vats. So, in order for us to have thoughts about vats, our
environment has to be a certain way: it has to be such that we haven’t always been brains
in vats. As it happens, we can tell by introspection that we do have thoughts about vats.
(In fact, such thoughts are necessary, to be entertaining the skeptical hypotheses we are
entertaining.) So it follows that we haven’t always been brains in vats.5
Other philosophers regard it as unacceptable to say that we could have knowledge
of things like McK-3 by reflection alone. Intuitively, those seem to be facts that can only
be known by empirical investigation. So the puzzling McKinsey-style reasoning has to be
blocked in some way. One popular way to block it is to be an incompatibilist about
externalism and our ability to know the contents of our thoughts by reflection alone. The
incompatibilist says that it can’t be true both that a given thought has an externalist

5
Those who take this stance towards McKinsey puzzle: Warfield?? Sawyer??
This is only one interpretation of Putnam’s argument; there are also other interpretations… [[Some
interpretations require extra premises about what brains in a vat do refer to and think about, when they use
the word “vat”; or what I would refer to with “vat” if I were a vat.]]
Some versions of Putnam’s argument employ, not introspective knowledge of what one is
thinking, but rather disquotational knowledge about one’s own language. This raises special difficulties that
we cannot pursue here.

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 6
content and that we are able to know our thought has that content, just on the basis of
introspection.6
Some philosophers take this incompatibility to discredit externalism. Others take
it to discredit our capacity for reflective self-knowledge. They say we can have direct,
introspective knowledge only of narrow features of our thoughts, features that supervene
on what we’re like intrinsically.
All of these responses strike me as over-reactions. The most sober response would
be a story that steered between them. It would allow us to have introspective knowledge
of the contents of our thoughts, even if those contents are externalist; but it would deny
that we can know what our external environment is like by introspection and a priori
reasoning alone. To steer this middle path, we need to find some way, other than the
incompatibilist’s way, to block the McKinsey-style reasoning.

II
The McKinsey-style reasoning is a form of modus ponens. So one way to block
that reasoning would be to articulate and defend a limiting principle on when modus
ponens reasoning is legitimate. Several of the views we’ll be considering try to do just
this.
One way to limit modus ponens reasoning is to deny Closure: that is, to say that
we can know McK-1 and McK-2, and know that these entail McK-3, but deny that this
entails we’re in a position to know McK-3. Some accounts of knowledge do deny
Closure in this way. For example, perhaps you’re in a position to rule out all the
epistemic possibilities that are relevant alternatives to McK-1 and McK-2, but when
we’re considering McK-3, more epistemic possibilities are relevant, and you’re not in a
position to rule out those additional possibilities.7

6
This is McKinsey’s own response to the puzzle [cite]. Other people who say this.
There are also other arguments for Incompatibilism besides McKinsey’s. The other arguments
have to do with “slow switching” thought experiments. I will have to discuss these elsewhere.
7
[[Dretske]]. Nozick’s account of knowledge gives us a different story about why there can be

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 7
However, denying Closure is not so popular these days. Even among
epistemologists who employ the framework of “relevant alternatives,” many nowadays
would rather keep Closure. They say that in any given context of knowledge-attribution,
there is a single range of epistemic possibilities that will count as relevant. If we keep that
range fixed, then whenever a subject’s evidence is good enough to know the premises of a
modus ponens argument, it will also be good enough to know the argument’s conclusion.8
In any event, I want to focus our attention on ways to limit modus ponens
reasoning that do not require us to deny Closure.
Crispin Wright and Martin Davies have formulated one such limitation. Consider
the following argument. You’re at the zoo, and you see a striped horse-like animal in the
pen in front of you. The sign on the pen says “Zebra.” All of this seems to justify you in
believing:
ZEBRA-1
That animal is a zebra.
A little reflection convinces you that if the animal is a zebra, it isn’t a mule, and a fortiori
it isn’t a cleverly-disguised mule. Hence, you know:
ZEBRA-2
If that animal is a zebra, it isn’t a cleverly-disguised mule.
Putting these together, you conclude:
ZEBRA-3
That animal isn’t a cleverly-disguised mule.
Now, Wright and Davies do not want to raise any doubts about Closure. They allow that
you can know both ZEBRA-1 and ZEBRA-3 to be true. What they doubt, however, is that
the reasoning I just sketched could be what gives you justification for believing ZEBRA-3.
Rather, they think that you would already need justification for ZEBRA-3 to be in place,
before you could be justified in believing ZEBRA-1 on the grounds I described. This is
because those grounds aren’t really enough, by themselves, to justify you in believing
ZEBRA-1. It’s only insofar as they’re supplemented by some antecedent or independent
justification for believing ZEBRA-3 that they can support ZEBRA-1. Hence, the ZEBRA-

failures of Closure.
8
See [[Stine 1976]], [[Cohen 1988]], [[DeRose 1995]], and [[Lewis 1996]].

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 8
argument can not do anything to enhance ZEBRA-3’s epistemic credentials for you.
Wright and Davies put this by saying that the justification you have for the argument’s
premises does not “transmit” into a reason to believe the argument’s conclusion.
This suggests the following limiting principle on modus ponens reasoning: a piece
of modus ponens reasoning cannot give you justification for believing its conclusion
when you need to be antecedently justified in believing that conclusion, to be justified in
believing the premises in the way you do. Reasoning that violates this constraint exhibits
a “failure of transmission,” even if it doesn’t exhibit any failure of Closure. Wright and
Davies think that the McKinsey-style reasoning suffers from just this kind of defect.9
Their reasons for thinking this are different. Wright thinks it has to do with the
specific kinds of thought-contents that McKinsey’s argument is dealing with, and the
specific kinds of grounds we have for believing we have thoughts with those contents.
Davies thinks it has to do with a more general phenomenon. He thinks that whenever you
argue for a conclusion by appeal to some premise P, you are presupposing that “there is
such a thought as P” for you to appeal to, in a sense of “presupposing” that is sufficient to
generate failures of transmission.
I agree with Wright and Davies that transmission-failure is a genuine epistemic
phenomenon, and that it deserves close study. I think the ZEBRA-reasoning [??] described
above provides a good illustration of the phenomenon. However, I also think that Wright
and Davies are inclined to see transmission-failure in too many corners.
For example, they’ve levied that charge against Moore’s famous “proof”:
MOORE-1 Here is one hand, and here is another.
MOORE-2 If I have hands, then the external world exists.
MOORE-3 So, the external world exists.
But whether one should count Moore’s proof as exhibiting transmission-failure depends
on what epistemology of perception one accepts. Wright and Davies think you are
justified in believing particular perceptual beliefs like MOORE-1 only insofar as you’re

9
Cite: [[Wright 1985, Davies 1998, Wright 2000, Davies 2000, Wright Rutgers forthcoming]]

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 9
already justified or entitled to accept general background assumptions like MOORE-3.
[[Soften their commitment here??]] I’ve argued elsewhere for an epistemology of
perception that denies this. On my view, for your perceptual experiences to justify you in
believing things like MOORE-1, you don’t need to have any antecedent justification for
believing that the external world exists, or that you’re not a brain in a vat, or that your
senses are reliable, or anything else of that sort. It’s enough if you lack reasons to believe
that you are a brain in a vat, that your senses are unreliable, and so on. Because your
perceptual justification for believing MOORE-1 does not require you to have antecedent
justification for believing MOORE-3, I do not think that Moore’s proof exhibits
transmission-failure—or, for that matter, any other epistemic vice.10
That dispute in the epistemology of perception is relevant to our present
discussion because I think it reveals that certain kinds of defeasibility are not enough, by
themselves, to make an argument guilty of transmission-failure. Wright and Davies tend
to move too quickly from:
(i)
Your justification for believing the premises of such-and-such an
argument would be defeated by evidence that not-C.
to:
(ii)
Your justification for believing the premises rests on the tacit assumption
that C, so you need some antecedent justification or entitlement to believe
that assumption.
I have argued elsewhere that (i) does not by itself entail (ii). In my view, the MOORE-
argument is one important case where (i) is true but (ii) is not; and it is only when we
have the kind of epistemic dependence described in (ii) that a charge of transmission-
failure will be appropriate.
So if Wright and Davies are going to make the charge of transmission-failure
stick against the McKinsey-style reasoning, they’ll have to show that our justification for
believing the premises of McKinsey’s argument requires us to have antecedent

10
It may have various dialectical weaknesses, but I think that’s a separate issue. See “Skeptic and
Dogmatist” and “Is Moore’s Argument…?” for discussion.

---

Externalism about Content and McKinsey-style Reasoning (10/1/01)
Page 10
justification for believing certain assumptions about our environment. Wright and Davies
think we will ordinarily have that antecedent justification. It need not be justification that
we had to earn or acquire. Rather, we might for some reason have a default entitlement to
believe those assumptions. We need not be able to articulate any explicit argument in
their support. But Wright and Davies will say we do need some antecedent justification
or entitlement for those assumptions to be in place, if we’re to be justified in believing
the premises of McKinsey’s argument.11
Now, if our justification for the premises of McKinsey’s argument does require us
to have antecedent justification for certain assumptions about our environment, then I
agree that Wright and Davies’ charge of transmission-failure might be appropriate. But it
will take a good deal of argument to show that our justification for believing McK-1 and
McK-2 rests upon environmental assumptions. What’s more, I think that if we do show
that, then we’ll have shown that our justification for those premises is not purely
reflective, in the way the McKinsey-style reasoning requires it to be. That will already be
enough to block the McKinsey-style reasoning, transmission-failure or no.12
So for the time being, I want to set aside Wright and Davies’ response to
McKinsey’s puzzle, and look at other possible solutions. (In the end, we’ll see that I do
[??] think there is some transmission-failure going on. But not in every case. And this is
just something that falls out of my solution to McKinsey’s puzzle. It doesn’t play any
role in my account of what’s wrong with the McKinsey-style reasoning.)

III
Another way in which modus ponens reasoning can go awry comes up in a kind
of example that Harman discussed in his book Thought.13 [[Is this the same example he
uses??]]

11
In addition to Wright and Davies, check also Brewer?? and Sawyer??
12
[[See Beebee]]
13
[cite] Harman says the cases were first discussed by Kripke.

---