Twin
Earth, Dry Earth, and Knowing the Width of Water
_________________________________________
Abstract: Paul Boghossian has most explicitly and in the most
careful detail argued that one can know a priori that a concept like water
is wide in the sense that possessing that concept in particular requires being
appropriately related to a specific type of external environment. I will show exactly where Boghossian’s
argument fails and exactly where empirical presuppositions must intrude into a
Twin Earth thought experiment. One
cannot know a priori that a concept like water is wide by thinking about
Twin Earth from the comfy confines of the philosopher’s armchair.
The content externalist claims that certain of our
concepts are wide in the sense that possessing those concepts in particular
requires being appropriately related to a specific type of physical or social
environment. Several authors have urged
that we can know a priori that a concept is wide, but Paul Boghossian has done
so in the most explicit detail.[1] Boghossian argues that we can know a priori
that a concept like water is wide by knowing that we can run a
successful Twin Earth (TE) thought experiment on the term that expresses
it. His strategy is to explicitly lay
out the conditions that a term like ‘water’ must meet to be TE-eligible and
argue that we can know a priori that those conditions are met. I will argue that one of those conditions,
the apriority of which Boghossian takes entirely for granted, cannot be known a
priori; and then I will show how the problem with that condition can be
generalized to show “where
exactly…empirical elements intrude into the TE experiment.” (Boghossian 1997: 277) Along the way, I will explain why Boghossian
is mistaken about what an externalist can say about his Dry Earth scenario.
1.
Boghossian on Twin Earth
Content externalism (CE) is most often motivated by
Twin Earth thought experiments.[2] These thought experiments seem as though
they can be run entirely from the armchair and yield a priori results.[3] Boghossian believes that one can know a
priori that a concept is wide by knowing that the TE-eligibility conditions for
the corresponding term are met. At
first, Boghossian claims that there are four term-specific
presuppositions involved in a successful Twin Earth experiment.[4] About the specific term ‘water’, Boghossian
writes,
First, and least controversially, water and twater
have to be thought of as distinct substances, distinct natural kinds;
otherwise, it won’t be true that Oscar’s word ‘water’ and Toscar’s word ‘water’
have distinct extensions and referents.
Second, the word ‘water’—whether on Earth or Twin Earth—must be thought
of as aiming to express a natural kind concept; otherwise, the fact that water
and twater are distinct natural kinds will not be semantically relevant. Third, Oscar and Toscar have to be thought
of as chemically indifferent, as having no views about the chemical composition
of the liquid kinds around them; other wise, they won’t end up as functional
duplicates of each other in the way that the experiment requires. Fourth, the concepts expressed by the
Earthly and Twin Earthly tokens of ‘water’ have to be thought of as atomic
concepts, not compound concepts that are compositionally built up out of other
concepts in well-defined ways. For
example, the experiment presupposes that water can’t be thought of as
capable of being defined as a tasteless, odorless, liquid that flows in the
rivers and faucets. (Boghossian
1997: 274-275)
But later, when he goes to argue that one can know a
priori that the conditions for TE-eligibility are satisfied, Boghossian draws
our attention back to only three of the conditions:
What conditions does a word have to meet if it is to
be Twin Earth-eligible? As we have
seen, it has to be a word that expresses an atomic concept. It also has to aim to name a natural
kind. Furthermore, the user of the word
must be indifferent about the essence of the kind that his word aims to name;
he must be chemically indifferent.
(277)
He argues that one can know a priori that a specific
term like ‘water’, (a) expresses an atomic concept, (b) aims to name a natural
kind, and (c) can be used by one who is indifferent about the essence of the
kind it aims to name. There are
significant worries about the apriority of at least the atomicity condition and
the aiming to name a natural kind condition (see Brown 1999). But the main worry for me, and the one which
can be generalized, has to do with the condition that was overlooked when it
came time to argue for apriority.
2. Knowing
that Water is not Twater
Boghossian admits that ‘water’ is TE-eligible only
if water and twater can be thought of as distinct, but he never argues for the
apriority of this condition. I will
argue in this section that one cannot know a priori that water and twater are
distinct.
Water is the stuff, whatever
it turns out to be, to which my uses of ‘water’ refer.[5] A successful TE thought experiment requires
that there be some possible stuff to which my Twin Earth doppelganger’s uses of
‘water’ would refer. This possible
stuff needs to have two features.
First, it must be distinct from water; this so we can ultimately claim
that my doppelganger and I express different concepts with our uses of
‘water’. Second, it must be indistinguishable
from water with respect to a certain range R of superficial properties; this so
we can maintain that my doppelganger and I are intrinsically identical. So a successful TE thought experiment can be
run on my term ‘water’ only if there can be some stuff that is both distinct
and superficially indistinguishable from water. I contend that we cannot know a priori that there is some
possible stuff that is distinct but superficially indistinguishable from whatever
it is to which I actually succeed in referring with my uses of ‘water’.
We can simply stipulate
that some possible stuff is distinct from water, whatever water turns out to be
like. Perhaps this is why Boghossian
thinks that this condition of TE-eligibility is the ‘least controversial’. But we cannot go on to stipulate as well
that the same stuff is R-indistinguishable from water. That would require knowing that water is not
individuated solely with respect to R-properties. We would have to know, for example, that water is not just any old
tasteless, odorless liquid that flows in rivers and faucets. If it were, then there could not be some
substance that was indistinguishable from water with respect to these
properties and still really distinct from water—anything that was a tasteless,
odorless, liquid that flows in the rivers and faucets would have to be
water. Since I cannot know anything
about the underlying nature of the stuff to which I actually refer with my uses
of ‘water’ without empirical investigation of my external environment, I cannot
know a priori that water is not individuated solely with respect to
R-properties. So I cannot know a priori
that there can be some stuff that is R-indistinguishable but really distinct
from water.
At
this point Boghossian will try to fall back on the atomicity condition of
TE-eligibility and argue that we can in fact know a priori that water is
not just any old tasteless, odorless liquid that flows in the rivers and
faucets because we can know a priori that ‘water’ expresses an atomic concept. The concept expressed by ‘water’, according
to the atomicity condition, cannot be “defined as a tasteless, odorless,
liquid that flows in the rivers and faucets.” Since we can know that the concept expressed by ‘water’ is
atomic, we can know a priori that water is not just any old tasteless, odorless
liquid that flows in the rivers and faucets.
The problem with this
objection is that it subtly confuses syntax and semantics. To maintain the apriority of concept
compositionality, Boghossian must have in mind a syntactic notion. So when he says that ‘water’ cannot be defined
as a tasteless, odorless liquid that flows in the rivers and faucets, he must
mean that the concept expressed by ‘water’ is not simply the syntactic
composite of the concepts expressed by “tasteless, odorless, liquid that flows
in the rivers and faucets.” It is
perfectly possible however, as I will show later, for a term to express a
syntactically atomic concept that is nevertheless semantically equivalent to
some syntactically composite concept.
It is perfectly possible for the term ‘water’ to express a syntactically
simple concept and yet refer in all possible worlds to all and only tasteless,
odorless, liquids that flow in rivers and faucets. Simply knowing that ‘water’ expresses an atomic concept does not
tell us anything about the semantic extension of the term or the individuation
of the stuff to which the term refers.
3.
Boghossian on Dry Earth
I have just argued that one cannot know a priori
that water and twater are distinct because one cannot know a priori that there
can be some stuff that is R-indistinguishable but really distinct from
water. Let me now argue for that point
in a slightly different way, using Boghossian’s Dry Earth scenario, so that I
can raise Boghossian’s objection in terms of his argument about what an
externalist cannot say about Dry Earth.
Here is my argument: Dry
Earth is a situation where there appears to be a tasteless odorless liquid
flowing in the rivers and faucets, but there is really nothing there—it is a
complex hallucination. Still, the
people on Dry Earth use ‘water’ in all the same narrowly described ways as we
do. That is, they use ‘water’ to talk about
the tasteless, odorless liquid that they believe flows in their rivers and faucets. The residents of Dry Earth used some
description like “the tasteless, odorless liquid that flows in the rivers and
faucets” to fix the reference of their term ‘water’; but there was nothing in
their environment for that description to pick out. If we imagine, as the story is told, that Dry Earthlings
nevertheless continue to use the expression ‘water’ in all the same narrowly
described ways as we do, then we must admit that their uses of ‘water’ are in
fact meaningful and do express some concept.
I contend that the term ‘water’ as it is used on Dry Earth must express
a concept that refers in any possible world to all and only tasteless, odorless
liquids that flow in rivers and faucets.[6] The description that was intended merely as
a reference-fixing device has in fact, and perhaps unbeknownst to the Dry
Earthlings, become the de facto definition or semantic analysis of the
term. The stuff to which Dry Earthlings
actually succeed in referring with their uses of ‘water’ is individuated solely
with respect to a certain range R of superficial properties. There cannot be a stuff that is
R-indistinguishable but really distinct from the stuff to which Dry Earthlings
actually succeed in referring with their uses of ‘water’.
Now if we happen to be in, and
to have always been in, a Dry Earth scenario, then our term ‘water’ actually
refers to a stuff that is individuated solely with respect to a certain range
of superficial properties. So if the
actual world, contrary to what believe, turns out to be like Dry Earth, then
there cannot be a substance that is superficially indistinguishable but
substantially distinct from water.
Since I cannot know a priori whether the actual world is like Dry Earth,
I cannot know a priori whether there can be stuff that is superficially
indistinguishable but substantially distinct from water.
Boghossian would certainly
object to my contention that the Dry Earthlings uses of ‘water’ turn out to
refer to all and only tasteless, odorless liquids that flow in rivers and
faucets. He himself believes that this
is the right thing to say about the Dry Earth scenario, but he argues that the
content externalist cannot say it:
“...the Twin Earth
externalist is committed...to holding that ‘water’ expresses an atomic concept under conditions where it
has a non-empty extension...That is one of the presuppositions of the Twin
Earth thought experiment. But, then,
how can the very same word, with the very same functional role, express an
atomic concept under one set of external conditions and a compound,
decompositional concept under another set of external conditions? A concept’s compositionality is exclusively
a function of its internal ‘syntax’, and can’t be contingent upon external
circumstances in the way that the present proposal would require.” (Boghossian 1997: 281)
Boghossian’s argument here can be fairly laid out as
follows:
P1: If ‘water’ has the same functional role in two different
external circumstances, then it cannot express an atomic concept in one but a
compound concept in the other.
P2: The
externalist is committed to saying that ‘water’ expresses an atomic concept on
Earth.
C1: So the externalist must say that ‘water’ also expresses an
atomic concept on Dry Earth.
P1 is supposed to follow from the idea that a
concept’s compositionality is solely a function of its internal ‘syntax’. There is an assumption at work here that the
functional role of the term that expresses a concept determines the internal
‘syntax’ of the concept (whatever exactly that is supposed to be). Boghossian is assuming that if two terms
have the same functional role, then the concepts they express have the same
internal syntax. That is what is needed
to move from same functional role to same compositionality, as in P1. It is not at all obvious to me that the
assumption here is so unproblematic that it can go without argument, as it
does. It can seem relatively
unproblematic to hold that internal syntax (of concepts, understood as elements
of a language of thought) determines functional role, but Boghossian is
implicitly relying on the converse of that.
And that seems much more problematic, especially in the present
context. For, an externalist may well
want to say that ‘syntactic’ features of concepts are determined in part by the
social/physical environment to which one is appropriately related.
There
are problems, I think, with justifying P1; but I am willing to concede it. Let’s say that C1 above is true. We do not yet have an objection to my claim
that ‘water’ on Dry Earth refers in any possible world to all and only
tasteless, odorless liquids that flow in rivers and faucets. What is needed to complete Boghossian’s
objection is the following:
P3: If
on Dry Earth ‘water’ refers to all and only tasteless, odorless liquids that
flow in rivers and faucets, then ‘water’ expresses a compound concept on Dry
Earth.
Boghossian
could then combine this with C1 above to yield:
C2: So
an externalist cannot say that on Dry Earth ‘water’ refers to all and only
tasteless, odorless liquids that flow in rivers and faucets.
Here again, however, is
Boghossian’s subtle confusion of syntax and semantics. Just because a certain term is semantically
co-extensional with a complex description does not mean that the concept
expressed by the term is a mere syntactic composite of the concepts expressed
by the constituents of the description.
It is possible to imagine, for example, that someone expresses an atomic
concept with her uses of the term ‘bachelor’ even though that term is
co-extensional with the description ‘unmarried (adult, human) male.’ It is conceivable that someone could acquire
the concept bachelor first, and then acquire the concepts male
and unmarried in part through her grasp of the concept bachelor,
with the help of her grasp of concepts like spinster and husband
. One could learn the concept unmarried
as applying to anything that is either a bachelor or a spinster; and one could
learn the concept (adult, human) male as applying to anything that is
either a bachelor or a husband. Thus we
can imagine that bachelor is an atomic concept for someone, and not
merely a compound of the concepts unmarried and male, even though
it refers to all and only unmarried males.
A term can be semantically co-extensional with a description without
expressing a concept that is the syntactic composite of the concepts expressed
by the description.
The upshot then is
this: An externalist can say both that
‘water’ on Dry Earth expresses a syntactically atomic concept and that the
semantic extension of that concept is all and only tasteless, odorless, potable
liquids. So Boghossian is mistaken when he claims that an externalist cannot
say that ‘water’ on Dry Earth refers to all and only tasteless, odorless
liquids that flow in rivers and faucets.
My argument remains intact—if the actual world turns out to be like Dry
Earth, then our uses of ‘water’ refer to a stuff that is individuated solely
with respect to a certain range of superficial properties; and since I cannot
know a priori whether the actual world is like Dry Earth, I cannot know a
priori whether there can be some stuff that is superficially indistinguishable
but substantially distinct from the stuff to which we actually succeed in
referring with our uses of ‘water’.
4.
Conclusion
I have up to now been arguing specifically that one
cannot know that water is wide on the basis of a TE thought experiment, because
one cannot know a priori that twater is not water.[7] The problem with knowing a priori that
twater is not water generalizes: One
cannot know a priori that any concept is wide on the basis of a TE thought
experiment, because one cannot know a priori whether there can be some stuff
that is distinct but superficially indistinguishable from the stuff to which
some term actually succeeds in referring.
Imagine that a subject S
uses a term T to express a concept C and that S explicitly associates a range R
of properties with c-stuff (i.e., the stuff to which C refers). To show that C is wide, a successful TE
thought experiment must describe a possible subject S* who meets two
conditions. First, S* must be
intrinsically identical to S and use T in all the same narrowly described ways
as S. Second, S*’s uses of T must have
a different extension from S’s uses of T.
These two conditions can be jointly satisfied only if there can be some
stuff that is distinct but R-indistinguishable from c-stuff. That is, one can know a priori that C is
wide only if one can know a priori that c-stuff is not individuated solely with
respect to its R-properties.
But one cannot know a priori
that some stuff is not individuated solely with respect to a specified range R
of properties. Even if one explicitly
intends a term to express a concept that refers to some stuff that is
individuated more precisely or more finely than R-stuff, it can still turn out
that one expresses a concept that refers to all and only R-stuff, as is the
case in the Dry Earth scenario. On an
externalist account, concepts are individuated in part by specific features of
one’s external environment; and so the extension of a term may turn out to be
quite different from what one narrowly intends or believes. To know that there really is more to being
c-stuff than merely having R-properties requires knowing something about the
stuff to which T actually refers. Given
that T expresses a wide concept, what T actually succeeds in referring to
depends on some relatively specific features of our actual environment. Since one cannot know a priori anything very
specific about one’s external environment, one cannot know a priori that there
is more to being c-stuff than merely having R-properties. One cannot know a priori that c-stuff is not
individuated solely with respect to R-properties; and so one cannot know a
priori that C is wide on the basis of a TE thought experiment.
Twin Earth thought
experiments require that there be some possible stuff that is superficially
indistinguishable but nevertheless distinct from that to which one actually
succeeds in referring with her uses of a certain term. This is where empirical elements intrude
into the TE thought experiment. For
this requires knowing something about the nature of the stuff to which one
actually succeeds in referring with a certain term, and that requires some
empirical investigation of one’s actual environment. So it is that we cannot know that a concept like water is
wide simply by thinking about Twin Earth from the comfy confines of the
philosopher’s armchair.
REFERNECES
Boghossian,
P. 1997. What the externalist can know
a priori. Proceedings of the
Aristotelian
Society 97: 161-75.
Reprinted in (Wright 1998). Page
references for (Boghossian 1997) are to that version.
Brown,
J. 1995. The incompatibility of
anti-individualism and privileged access.
Analysis 55: 149-56.
-- 1999. Boghossian on externalism and privileged access. Analysis 59: 52-59.
--2001. Anti-individualism and agnosticism. Analysis 61: 213-24.
Brueckner,
A. 1992. What an anti-individualist knows a priori. Analysis 52: 111-18.
-- 2002.
Anti-individualism and analyticity.
Analysis 62: 87-91.
Falvey,
K. 2000. The compatibility of
anti-individualism and privileged access.
Analysis 60:
137-42.
Gertler,
B. 2004. We can’t know a priori that H2O exists. But can we know that
water does?
Analysis 64: .
Goldberg,
S. 2003. On our alleged a priori
knowledge that water exists. Analysis
63: 8--41.
McKinsey,
M. 1991. Anti-individualism and
privileged access.” Analysis 51:
9-16.
Nuccetelli,
S. 1999. What anti-individualists cannot know a priori. Analysis 59: 48-51.
-- 2003. New
Essays on Semantic Externalism and Self-Knowledge. Cambridge: The
MIT Press.
Putnam,
H. 1975. The meaning of ‘meaning’. In Philosophical Papers Vol. II.
Cambridge: Cambridge University Press.
Sawyer,
S. 2003. Sufficient absences. Analysis 63: 202-8.
Warfield,
T. 1998. “A priori knowledge of the
world: knowing the world by knowing our
own minds.” Philosophical
Studies 92: 127-147.
Wright, Smith, McDonald (eds.) 1998. Knowing Our Own Minds. Oxford: Oxford University
Press.
[1] Michael McKinsey (1991) essentially argues that one can know a priori (i.e., non-empirically) that a concept is wide on the basis of one’s introspective knowledge that she possesses the concept and the fact that possessing that concept conceptually entails some substantive proposition about one’s external environment. For criticisms of McKinsey see Brueckner 1992, Nuccetelli 1999, Goldberg 2003, and most of the essays in Nuccetelli 2003.
Jessica Brown (1995) and (2001) essentially argues that one can know a priori that a certain concept is wide by knowing that one is agnostic about that concept’s application. For criticisms of Brown see Falvey 2000 and Brueckner 2002.
[2] See of course Putnam 1975 and Burge 1988.
[3] Boghossian (1997) and Warfield (1998) are the most explicit about this.
[4] A fifth condition is that intension determines extension, but this is not specific to any particular term.
[5] Note that we cannot presuppose the empirical information that water is H2O. The typical version of the TE thought experiment presupposes that water is H2O and then stipulates that twater is a substance superficially indistinguishable from water but with a different underlying chemical constitution XYZ. This way of describing the situation is no available if we are trying to generate a priori knowledge of the width of water. It should also be noted that it could even turn out that there is not in fact any water in my actual environment. Still, if ‘water’ is a meaningful atomic concept, then it has some extension across some possible worlds.
[6] There is no such liquid on Dry Earth, but I am claiming that the Dry Earthian term ‘water’ would refer to water (i.e., H2O) on Earth and twater (i.e., XYZ) on Twin Earth.