What’s Wrong with McKinsey-style Reasoning?

What’s Wrong with McKinsey-style Reasoning?
James Pryor
NYU
<jim.pryor@nyu.edu>
12/5/2006
forthcoming in Internalism and Externalism in Semantics and Epistemology, ed. by
Sanford Goldberg
I

It’s widely accepted nowadays that some of our thoughts are externalist. What does that
mean? I understand a thought’s content to be those of its representational aspects that
one experiences as representational and are essential to its being that thought. And we
can understand a thought to be externalist when subjects who are internally the same can
differ with respect to whether they’re thinking the thought’s content.1
Our thoughts about water are widely thought to be externalist. So too are our
demonstrative thoughts.

1
I leave it open whether a thought’s content suffices to determine the thought’s truth-conditions. On
a view like Lewis 1979’s, when I think to myself “My pants are on fire,” and you think to yourself “My
pants are on fire,” we’re thinking the same content though our thoughts have different truth-conditions.
Segal 1989 proposes a similar view about demonstrative thoughts. I don’t count extra-contentual semantic
differences of these sorts as externalist. Neither do I count as externalist views that say that Earthling
thoughts are about H2O, but only contingently so; those very same thoughts could have been about XYZ
instead. An externalist has to say there are essential differences in the contents of some duplicates’
thoughts.

My gloss on “externalist” employs the notion of subjects being “internally the same.” It isn’t
straightforward what that amounts to. It’d be nice if our definition of externalism didn’t presuppose any
commitments about materialism. But it’s not obvious that we can characterize “how you are internally” in
terms of phenomenology either. Some philosophers (Dretske 1995 Ch 5; Tye 1995 §5.4 and 2000 §3.5) say
that sensory phenomenology can itself be externalist. Our definitions ought to permit such a view.
Additionally, I think it can also be part of your phenomenology that you are now occurrently thinking such-
and-such—where that content may intuitively be an externalist one. For these reasons, I find it difficult to
specify what it is for subjects to be “internally the same,” in a way that preserves the familiar
classifications. I’ll just ask you to use whatever rough and ready understanding you have of this.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 2
If we accept any kind of externalism, then we confront a puzzle that arises about
our ability to tell what we’re thinking. This comes from what’s commonly known as
McKinsey’s Argument.2 It rests on two ideas.
First, we think you can tell the contents of your thoughts just by introspection. For
example, you can tell just by introspection that you’re thinking that water puts out fires:
McK-1
You’re thinking a thought with the content Water puts out fires.
As we proceed, it will be handy to have a way of identifying thoughts, without
prejudging anything about how they’re individuated, or indeed, whether they’re even
externalist. One way to identify a melody is to produce it: you can say “Hey, you know
that jingle [here you hum: la-la-lalalaa…]?” Analogously, I think you can identify a
thought content by thinking or entertaining it. You can say: “the thought content [here
you think the relevant content, perhaps by rehearsing to yourself the sentence ‘Water
puts out fires’].” I’m using the notation:
the content Water puts out fires
to express this way of demonstrating a thought content, by thinking it.3
The second idea driving McKinsey’s Puzzle is that you can also tell that some of
your thoughts are externalist, purely by armchair philosophical reflection. And it would
seem that if a given thought is externalist, then it’s only available to be had by subjects in
certain sorts of environments (more on this in a moment). Hence, it looks like you can
establish a priori something of the form:
McK-2
If you’re thinking a thought with the content Water puts out fires,
then…[here some claim about your environment, e.g., it actually does
or did contain samples of water].

2
After McKinsey 1991; see also Brueckner 1986 and Brown 1995.

3
Notice that this is not a description; it doesn’t just mean “whatever content I think by saying these
words to myself.” It’s a way of genuinely apprehending the content, and thereby enabling yourself to refer
to it. This will be important later.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 3
Putting the two together, it looks like you can conclude, purely on the basis of
introspection and a priori philosophy:
McK-3
Your environment is the relevant way [e.g., it actually does or did
contain samples of water].
And while few doubt that we do know things like McK-3, it’s extraordinary that we
should be able to know them purely on basis of introspection and a priori philosophy.
That’s the puzzle. It’s counter-intuitive that you should be able to tell what your
environment is like just on the basis of this kind of reasoning.4
In discussing this puzzle, some authors refer to our supposed introspective
knowledge of McK-1 and our supposed philosophical knowledge of McK-2 as all being
“a priori.”5 The name doesn’t matter much. But there are important epistemic differences
between introspection and the usual paradigms of the a priori. Calling them all “a priori”
can encourage confusions; some of which we’ll be disentangling later. Hence, I will
reserve the name “a priori” for justification that comes from logical understanding,
philosophical reasoning, and so on. Things like I am now thinking about a prime
number I’ll say instead are justified through your introspective experience or awareness
of your occurrent mental life.6 I’ll use the umbrella term “by reflection” to cover the lot.
So in my terminology, the surprising result posed by McKinsey’s Puzzle is that you can
tell what your environment is like purely by reflection.
One response to the McKinsey Puzzle endorses the result that you can establish
things like McK-3 by reflection alone. That is one way of construing Putnam’s

4
As we’ll discuss later, the puzzle isn’t confined to your knowing or having some special authority
about the presence of water. It’s already counter-intuitive that you should be able in this way to acquire any
justification at all to believe there really is water in your environment.

5
See, for example, McKinsey 1991, Boghossian 1997, and McLaughlin and Tye 1998. See also
Kitcher 1980 §V.

6
I’ll avoid any substantial assumptions about what that amounts to. BonJour 1998, pp. 7ff also
denies that introspective justification is a priori.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 4
Argument in Chapter 1 of Reason, Truth, and History. Putnam argues that people who
have always been brains in vats can’t refer to or think about those vats. (At least, not by
using the word ‘vat.’) So, in order for us to have thoughts about vats (using ‘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.7
Other philosophers resist that result. They say we can only establish things like
McK-3 through empirical investigation. So the puzzling reasoning has to be blocked
somehow. One way to block it is to go Incompatibilist about externalism and your
ability to know the contents of your thoughts by reflection alone. The Incompatibilist
says it can’t be true both that a thought has an externalist content and that you’re able to
tell your thought has that content, just on the basis of introspection.8 Some
Incompatibilists take the incompatibility to discredit externalism. Others take it to
discredit your ability to tell what externalist thoughts you’re having by reflection alone.
I think these are all over-reactions to the puzzle. We can find a more sober
response that steers between them. We can tell what we’re thinking by reflection alone,
even when what we’re thinking is externalist. However, this doesn’t give us a route to
reflective knowledge—or even reflective justified belief—about what our environment is
like. So I will argue.
As it happens, I’m not sure that reflective knowledge of our environment is to be
avoided at all costs. After all, some epistemologists maintain that thoughts like My senses

7
This is only one interpretation of Putnam’s argument. Some interpretations employ further
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 brain in a vat. Other interpretations focus on disquotational
knowledge about one’s language, rather than introspective knowledge of what one is thinking.

Sawyer 1998, Warfield 1995 and 1998, and Tymoczko 1989 defend similar responses to the
McKinsey Puzzle.

8
This is McKinsey’s, Brueckner’s, and Brown’s own response to the puzzle (Brown takes it back in
her 2004). See also Boghossian 1997. There are also other arguments for Incompatibilism, besides
McKinsey’s; two notable sources are Woodfield 1982 and Boghossian 1989.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 5
are reliable are justified a priori (albeit defeasibly). Here’s another possible route to
reflective knowledge of your environment: (i) notice, on the basis of introspection, that
you’re having an experience as of hands, and that you have no evidence that your senses
are misleading you; (ii) apply your favorite a priori epistemology of perception to get the
conclusion So I am justified in believing that I have hands; (iii) help yourself to a
defeasible but rational ampliative inference from I am justified in believing P to P.
Voilà: now you’ve got a purely reflective (albeit roundabout) justification to believe you
have hands. It may even suffice for knowledge.9 This merits careful discussion—
especially step (iii)—but it’s not obvious that the reasoning is illegitimate. Neither does it
trade on any assumptions peculiar to externalism. So my examination of the McKinsey
argument won’t take it for granted that reflective knowledge of our environment is flat-
out impossible. I just want to get the details straight. As these turn out, I think the
McKinsey argument will not give us a route to such knowledge.

II

Let’s attend to the bits we left unspecified at the end of McK-2. We supposed that, if a
given thought is externalist, then it’d only be available to be had by subjects in certain
sorts of environments. Is that supposition right?
It’ll be useful to think about an example drawn from the history of chemistry.10
Mendeleev presented his first periodic table in 1869. At that time, and then again
more carefully in 1871, he postulated the existence of four missing elements. He called
one of them ‘ekaboron.’ In 1879, Lars Fredrick Nilson, unaware of Mendeleev’s
predictions, spectrographically identified a new element in some Scandinavian minerals.
He managed to chemically isolate an oxide of this element; and he dubbed the element
‘scandium.’ This turned out to be Mendeleev’s ekaboron. At least, it occupied

9
Compare “The Explainer” in Hawthorne 2002.

10
For background, see http://web.lemoyne.edu/~giunta/EA/CONTENTS.HTML and
http://homepage.mac.com/dtrapp/periodic.f/periodicity.html.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 6
ekaboron’s position in the periodic table; and Mendeleev had closely predicted many of
its properties, such as its atomic weight, its valence, and the density of its oxide.
Now, in reality, Mendeleev thought elements were individuated by their atomic
weight rather than by what we call their “atomic number.” Protons weren’t discovered
until 1918. But it will simplify our discussion to pretend that Mendeleev did stipulate
how many protons ekaboron has (namely, 21) in advance. Additionally, I will assume
that ekaboron was scandium, that is, that Mendeleev was able to think and talk about this
element before it was discovered in nature and chemically isolated. That assumption may
well be challenged; and I’m not certain it’s true. But for this discussion we’ll assume it.
Finally, let’s imagine that Mendeleev engaged in correspondence with another chemist in
1871, discoursing at length about ekaboron but not telling her what the element’s
fundamental individuating properties were.
We now have three different subjects thinking about scandium. First, there’s
Mendeleev, who stipulates that with ‘ekaboron’ he’s referring to the element with 21
protons. Second, there’s Mendeleev’s friend, who acquires competence with the name
‘ekaboron,’ but who doesn’t know how the substance is chemically individuated. Third,
there’s the discoverer Nilson, who independently encounters the element in nature. Each
of these subjects says to himself, ‘Ekaboron [scandium] is a silvery metal.’
How many different thought contents will we have?
Different externalists will answer that question differently. Some will say that
since all the thoughts concern a single substance, there is only a single content. Some will
say that since the substance is cognitively presented to the subjects in three different
ways, there are three contents. Some might argue that Mendeleev and Nilson think
different contents, but Mendeleev’s correspondent acquires the ability to think the same
contents that Mendeleev thinks; so altogether there are only two contents. I’d like our
inquiry to apply to all these views, so I hope to avoid taking sides here as much as
possible. That forces us to proceed carefully in formulating the McKinsey reasoning.
Consider next a Sort-of-Twin Earth. The Mendeleev there is internally different:
he’s introduced ‘ekaboron’ to refer to the chemically similar element with atomic
number 39. But he says all the same things about it in his correspondence; and his friend
is an internal duplicate of the Earthly friend. There’s also an internal duplicate of Nilson

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 7
there, who’s happened upon a different metal, yttrium, that turns out to have atomic
number 39 and to correspond to the twin-Mendeleev’s predictions.
Now how many thought contents do we have? Everyone will say Mendeleev and
his twin are thinking different thoughts; after all, they’re not even internal duplicates.
Externalists will want to say that Nilson and his twin are thinking different thoughts, too,
despite being internal duplicates. Some externalists will want to say that Mendeleev’s and
twin-Mendeleev’s correspondents are also thinking different thoughts.
Are the different thoughts had by Nilson and his twin only available to subjects
whose environments contain samples of the respective elements? That depends on
whether their thoughts are the same as Mendeleev’s and twin-Mendeleev’s thoughts. For
plausibly Mendeleev’s thoughts don’t constitutively depend on the environmental
presence of the element he’s postulating. Mendeleev would have had the same thoughts
even if Earth turned out to contain no traces of scandium. So if Nilson’s thought has the
same content as Mendeleev’s thought, then it’s not true that that content is only thinkable
by subjects whose environments really contain scandium. What may be true is that the
content is not thinkable in the way Nilson thinks it unless one has genuinely encountered
scandium. But it’s tricky to say what these “ways of thinking” amount to, without taking
sides about issues that externalists disagree about.
And what about the correspondents? Even if we agree that their thoughts differ,
and so are externalist, it’s plausible that, like Mendeleev, they too would have had the
same (deferential) thoughts even if their environment had contained no traces of
scandium. What their thoughts seem constitutively to depend on is not samples of
scandium but rather communication with Mendeleev and his twin. But here too matters
are sticky, since we don’t want to take sides on whether their thoughts do or don’t have
the same contents as Mendeleev’s and Nilson’s thoughts. The other guys’ thoughts don’t
require the existence of any correspondence.
Here’s my attempt to finesse these issues. Let’s focus on Nilson and his twin.
They each reason in the following way:
(1)
I’m now thinking the content Scandium is a silvery metal.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 8
(2)
I’m understanding this concept scandium to be governed by my ostensive
introduction, rather than by deference to any authority; and I’m not
making any definite assumptions about how it’s chemically individuated.
(3)
Anyone who’s able to think the content Scandium is a silvery metal in the
way described in (2) must inhabit an environment that does or did contain
samples of scandium.
(4)
So my environment does or did contain samples of scandium.
Note that (1)–(4) specify a form of argument. When Nilson thinks it through, he thinks
premises about scandium. When his twin thinks it through, he thinks premises about
yttrium. It’s prima facie plausible that these subjects should know—or at least be
defeasibly justified in believing—premises (1) and (2) just on the basis of introspective
reflection. And it’s prima facie plausible that they should also know—or at least be
defeasibly justified in believing—premise (3) on the basis of a priori, Putnam- and
Kripke-style philosophical reasoning. The apparent result is that they acquire some
wholly reflective justification to believe (4). That’s surprising. (It remains surprising even
if it’s allowed that their justification to believe the premises, and hence the conclusion, is
empirically defeasible.)
Mendeleev’s correspondent would instead reason to a different conclusion:
perhaps, “So my environment does or did contain other subjects.”11

III
The McKinsey-style reasoning is a form of modus ponens. So one way to block it
would be to articulate and defend constraints on when modus ponens reasoning is
legitimate. Some responses to McKinsey’s Puzzle take that form.
One move is to deny Closure. Perhaps Nilson is able to rule out the relevant
alternatives to each of (1)–(3), but more possibilities are relevant alternatives to (4), and
he’s not in a position to rule them out. Maybe that means he’s able to know (1) through

11
For this version of the argument, see Burge 1982, Brown 1995, McLaughlin and Tye 1998 pp.
312ff, Falvey 2000, and Brown 2001.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 9
(3), without being able to know (4). Maybe. But denying Closure is not so popular these
days. Even among epistemologists who employ the framework of “relevant alternatives,”
most would rather keep some form of Closure.12 In any event, I want to restrict our
attention to responses to McKinsey’s Puzzle that don’t require us to deny Closure.
Crispin Wright and Martin Davies have formulated a different sort of constraint,
in their discussions of “transmission-failure.”13 Their analyses of the McKinsey argument
reward careful consideration; and they interact in interesting ways with the diagnosis I’ll
be giving. But I’ll have to reserve discussion of them for another occasion.
In order to avail ourselves of the McKinsey-style reasoning, we need to have
reflective justification to believe all its premises simultaneously. Sometimes modus
ponens reasoning fails because in acquiring justification to believe one of the premises,
one loses justification to believe another. For example, I may start out thinking, quite
reasonably, that you are not a rap star. The only time I ever heard you rap was when we
did karaoke together. So I justifiably think If you’re a rap star, then so am I. But then
you confront me with evidence of your covert musical career. I’m not now justified in
affirming the antecedent and concluding that I’m a rap star too. This is because in
acquiring justification to believe You are a rap star, I lost my grounds for believing the
conditional If you’re a rap star, then so am I.14
We should take care that no such funny business is going on with the McKinsey-
style reasoning. At the moment, there is no special reason to think it is. In the McKinsey-
style reasoning, your justification to believe conditionals like:
McK-2
If you’re thinking a thought with the content Water puts out fires,
then your environment is so-and-so.

12
For discussion, see Stine 1976, Brueckner 1985, Cohen 1988, Vogel 1990, DeRose 1995 esp. §10,
Klein 1995, and Hawthorne 2005.

13
See especially Wright 2000 and 2003, and Davies 1998, 2000, 2003a and 2003b.


14
Compare Harman 1973 Chapter 9; Ginet 1980; and Sorenson 1988.

---

Externalism about Content and McKinsey-style Reasoning (12/5/2006)
Page 10
comes from reflecting on a priori thought-experiments. Why should learning that the
antecedents of these conditionals are true—that you are thinking thoughts with the
indicated contents—defeat or undermine your grounds for believing the conditional? At
the moment, we have no ground for saying it would. But we will return to this possibility
later.
For the time being, then, the way looks clear for you to legitimately combine your
justification to believe premises (1) through (3), and thereby acquire justification to
believe the surprising conclusion (4).
IV
But is your justification for believing all the premises really wholly reflective? We need
to think about this more carefully.
So far we’ve been thinking about Nilson and his twin. They inhabitant different
environments—one where scandium is distributed in certain Scandinavian minerals, the
other where yttrium is so distributed—and as a result they end up thinking different
thoughts. But their environments have it in common that there’s some substance of the
appropriate sort that they’re interacting with. Let’s call any environment of that sort a
hospitable environment. Nilson will also have duplicates whose environments are
inhospitable. These will be places like Boghossian’s Dry Earth.15 Nilson’s unlucky
duplicate there will just be hallucinating handling some minerals, isolating a metallic
oxide, and so on. From the inside, everything will seem to him just as it seems to the real
Nilson.16 But outside, there’s no substance there for his thoughts to latch onto. What will
the Nilson in that inhospitable environment be thinking, when he says to himself,
‘Scandium is a silvery metal’?
There are a variety of answers one might give here.
Perhaps he’s thinking superficial descriptive thoughts: thoughts true in case some
new metallic element he’s just identified in such-and-such minerals is a silvery metal.

15
See Burge 1982, 114ff; and Boghossian 1997.

16
Though see fn. 1 for some difficulty with this.

---