Joe Lau
Excerpt from Ph.D. Thesis
13 August 1994
No footnotes in this version
In everyday life, we typically explain what people do by attributing mental states
such as beliefs and desires. Such mental states belong to a class of mental states that
are intentional, mental states that have content. Hoping that Johnny will win, and
believing that Johnny will win are of course rather different mental states that can lead
to very different behaviour. But they are similar in that they both have the same content
: what is being hoped for and believed is the very same thing. According to the thesis of
externalism that has been defended most notably by Hilary Putnam and Tyler Burge, not all
of the contents of our mental states are determined by our intrinsic properties. Instead,
the contents of our beliefs and desires are often determined in part by our relations to
the environment. They are, so to speak, "wide" contents that are "not in
our heads." Although externalism is accepted by most philosophers, many have argued
that mental states with wide contents must also have a kind of content wholly determined
by the intrinsic properties of the individuals who are in those states. This kind of
content is called "narrow content". The aim of this paper is to distinguish
between three rather different motivations for postulating narrow content. I argue that,
given a certain conception of narrow content that I shall explain below, none of these
three motivations succeed in establishing the existence of narrow content.
1. Three Motivations for Narrow Content
Arguments for externalism often rely on familiar thought experiments such as the
following one. Jane is an ordinary earthling who is acquainted with water in the normal
way, but like many people, she is ignorant of the chemical nature of water, such as the
fact that water is made up of H2O molecules. Nonetheless she has many beliefs
about water. For example, she believes that water quenches thirst, and that water puts out
fires. But now suppose she has a duplicate twin-Jane who has the same intrinsic
properties. Twin-Jane grows up on planet twin-earth where everything is exactly the same
as on earth, except that there is no water there. Instead twin-earth has twin-water, a
substance that has the same appearance as water. Twin-water tastes just like water, and it
quenches thirst in just the same way. But it has a totally different chemical nature,
being made of the compound XYZ instead. We are supposed to have the intuition that
twin-Jane lacks beliefs about water, and that de dicto belief ascriptions
that use the word "water" and that are true of Jane will not be true of
twin-Jane. So unlike Jane, twin-Jane lacks the belief that water quenches thirst, and she
lacks the belief that water puts out fire, etc.. Instead, what she believes is that
twin-water quenches thirst and puts out fire. Thus Jane and twin-Jane have beliefs that
differ in truth-conditions. It is supposed to follow that their beliefs differ in content,
despite the fact that they have the same intrinsic properties. If an individual I
has a mental state m with a certain content, and it is metaphysically possible for
a duplicate of I to lack a mental state with that same content, then I shall say
that m has a "wide" content.
I shall not challenge the externalist conclusion that many of our mental states have wide
contents. But notice that externalism does not imply that all intentional mental
states have wide contents. Consider my belief that something exists, and my belief that
everything is identical to itself. Although some of my possible duplicates might live in
communities that speak a somewhat different language, where for example the word
"exists" has a slightly different meaning, the claim that my duplicates could
lack those two beliefs has very little intuitive support. So if what we mean by there
being narrow content is that there are some mental states whose contents are not wide,
then I think there is indeed a prima facie case for the existence of narrow
content.
There is, however, little discussion of such examples by either the friends or foes of
narrow content. Those who defend narrow content generally do not argue that there are
particular belief predicates of the form "believes that p" which are true
of an individual and all his or her possible duplicates. Instead, their position seems to
be that every belief with wide content also has an additional kind of content that
is wholly determined by the intrinsic properties of their subjects. So proponents of
narrow content agree for example, that the content of Jane's belief that water quenches
thirst is wide. Nonetheless they insist that this very belief also has a different
kind of content that is wholly determined by her intrinsic properties. This is the narrow
content of Jane's belief, and although twin-Jane does not believe that water quenches
thirst, her belief that twin-water quenches thirst is supposed to have the same narrow
content as Jane's belief that water quenches thirst.
In this paper I shall focus on a particular class of intentional mental states - states
that have truth-conditions, and in particular beliefs and thoughts. Among the proponents
of narrow content, I think it is a common assumption that all beliefs and thoughts have
both a narrow and a wide content. For example, on Fodor's account, the truth-condition of
a belief or thought is supposed to be a function of its narrow content and some relevant
context. Brian Loar has also argued that beliefs and other intentional mental states have
both a social content and a psychological content, only the latter of which is determined
entirely by the subject's intrinsic properties. So, whatever narrow content is, I think it
is fair to say that it has to satisfy at least these two principles :
(A) Content is dualistic : every belief or thought that has a wide content also has
a narrow content.
(B) Narrow content is narrow : if an individual has a belief (or thought) with a
narrow content N, then every possible duplicate of that individual also a belief
(or thought) with the same narrow content N.
These two principles obviously do not explain what it is for a belief or thought to have a
narrow content. But they might be regarded as necessary conditions that a satisfactory
notion of narrow content have to satisfy. These two conditions are however rather minimal.
For example, I might stipulate that that snow is white is the narrow content of all
intentional mental states. Such a notion of narrow content would indeed satisfy principles
(A) and (B) : every belief or thought with a wide content also has a narrow content, and
the mental states of duplicates do not differ in narrow content. But of course this notion
of narrow content is quite useless since it does not explain anything. So if the
thesis that narrow content exists is to have any interest at all, one would have to show
not only that there is a coherent notion of narrow content that satisfies principles (A)
and (B). It would also have to be shown that this notion of narrow content has some
theoretical interest. In what follows, I shall discuss three different motivations that
have led philosophers to introduce narrow content. The first is that we need to introduce
narrow content to account for our first-person knowledge of our intentional mental states.
The second is that narrow content is to be understood as the internal component of a
mental state with wide content, that which is left of the mental state as we abstract away
from its relations to the environment. The third is that narrow content is needed for the
formulation of psychological laws. I think it is important to distinguish between these
three different motivations, for I think we have as yet no reason to think that there is a
single property that performs all three explanatory tasks. Furthermore, I shall argue for
a more basic conclusion : the three motivations by themselves provide no reason for
thinking that there has to be a theoretically important notion of narrow content that
satisfies both principles (A) and (B).
2. Narrow Content and Self-Knowledge
Recently there has been some discussion of externalism in connection with first-person
knowledge of our intentional mental states. By such self-knowledge, I mean knowledge of
our intentional mental states that do not rely on empirical evidence or observation of
behaviour. On the face of it, there does seem to be a prima facie difficulty in
reconciling externalism with our self-knowledge. If the contents of our thoughts are
determined in part by our relations to the environment, then one might think that in order
to know what we think, we have to find out what our environment is like. But
self-knowledge is precisely knowledge that does not come about by empirical
investigations. So it seems that we have a dilemma, that either contrary to appearance we
do not really know the contents of our own thoughts, or if we do, then externalism is
false.
However, the problematic conclusion follows only if we accept the following principle : to
know that p, one has to know that the conditions necessary for the truth of p
do indeed obtain. If our purpose is to reconcile externalism with the possibility of
self-knowledge, then it seems that narrow content is not really relevant whether we accept
this principle or not. First of all, if we reject this principle, there is then no reason
why externalism requires that, to know the contents of our own thoughts, we have to know
that the environmental conditions necessary for such thoughts do in fact obtain. On the
other hand, if the principle is accepted, it threatens to make not just self-knowledge,
but empirical knowledge in general, impossible. For example, it implies that Jane cannot
know that water puts out fire, since Jane does not know that a liquid made up of H2O
molecules can put out fire, and this has to be true if water does put out fire. But if it
turns out that general skepticism is the reason for thinking that externalism cannot be
reconciled with self-knowledge, then it is hard to see how postulating narrow content can
be of any help. If the skeptic thinks that we cannot know the wide contents of our
thoughts because we do not know what our environment is like, then surely this skeptical
challenge cannot be met by saying that thoughts have narrow content! So either way, there
does not seem to be any reason why we need to appeal to narrow content in reconciling
externalism with self-knowledge.
Brian Loar has argued that there is some related phenomena that calls for narrow
content. According to Loar, "subjective intentionality" describes the fact that
we have introspective knowledge of the intentional properties of our thoughts, and that
this is something we need narrow content to account for. To use Loar's example, when I
entertain the thought that Freud lived in Vienna, there are true judgements that I can
make with regard to its intentional properties, e.g. it is a thought about Freud, and it
is true if and only if Freud lived in Vienna. It is plausible that such judgements
constitute knowledge of the intentional properties of my occurrent thought, and that I
come to have such knowledge not by carrying out empirical inquiry of the ordinary kind.
According to Loar, we do not ordinarily conceive of such intentional properties as
extrinsic properties : "From a pre-critical perspective, knowledge of the references
of my own thoughts is privileged in a certain way, and that perspective involves no
apparent conceptions of external reference relations." Apparently, the claim is that
when I judge that, say, my occurrent thought is about Freud, I do not conceive of the
property of being a thought about Freud as an extrinsic property. But why should
this give us any reason to think that narrow content exists? Perhaps the argument is that
if I do not conceive of this intentional property as extrinsic, then it has to be the case
that the property is an intrinsic one.
But if this is Loar's argument for narrow content, then it is not a very convincing one.
First of all, as Stalnaker points out in his discussion of Loar, it is natural to conceive
of intentional properties as extrinsic properties even from a commonsensical perspective.
Although the externalist thesis is a substantial one, it is one that seems to gather ready
conviction even from a pre-theoretic perspective. More importantly, even if there is an
individual who judges that his occurrent thought has a certain intentional property, but
who does not conceive of the property as extrinsic, it still does not follow that the
intentional property so ascribed is intrinsic. One can ascribe a property without knowing
the full nature of the property, and it would be just as fallacious to argue that the
instantiation of the property of being water is independent of chemical
constitution, on the ground that one can ascribe the property without conceiving of it as
involving molecules.
On the other hand, perhaps Loar has something else in mind, when he claims that we do not
have an extrinsic conception of the intentional properties we ascribe to our occurrent
thoughts. One thing he might mean is this : my judgement that my thought is about
Freud is not a judgement that is based on my beliefs about my relation to the
environment. Thus I do not arrive at the judgement by reflecting on the nature of aboutness,
and inferring from my beliefs about my causal relations to Freud that the thought is
indeed about him. This is perhaps why Loar says that "the referential judgement is
from a first-person perspective independent of thoughts about causal reference
relations." The admittedly correct observation is that I can form direct judgements
about the intentional properties of my occurrent thought, without having to consciously
infer such judgements from my beliefs about my relations to the environment. I think
Loar's position is that there has to be some intrinsic properties about me that explain my
judgements of the intentional properties of my occurrent thought, properties which we
might take to be the narrow content of the thought. This seems to be the line of thought
behind his remarks that the "(objectively non-intentional) properties of object-level
thoughts which contribute to explaining why upon reflection they reveal themselves as
'about this and about that' can be counted as the basis of subjective intentionality ...
[which is] the disposition of thoughts to reveal themselves 'as intentional' upon
reflection."
I think there is a real and interesting issue as to what these intrinsic properties might
be. But note that if this is what narrow content comes down to, then such a notion of
narrow content will most likely fail to satisfy principle (A). That is, not all beliefs
and thoughts with wide contents will have the kind of narrow content that Loar has
envisaged. One reason is that subjective intentionality is after all a sophisticated
psychological phenomena, present only in creatures capable of reflexive judgements about
their own thoughts. Furthermore, even in our own case it is a phenomena that is restricted
to only conscious intentional mental states. Certainly an unconscious thought does not
normally have the disposition to cause us to judge of it that it has some
intentional property P! So if according to Loar narrow contents are those intrinsic
properties that are the basis of subjective intentionality, then it seems quite likely
that narrow content is possessed only by those conscious intentional states of creatures
capable of reflexive judgements. The mass of our unconscious intentional states, as well
as the intentional states of simpler thinking creatures that lack the ability for
reflexive thinking, will most likely not be states that have narrow content in this sense.
However, I think many philosophers (and perhaps Loar included!) will argue that there is
still a need to ascribe narrow content to these mental states for the other two purposes.
They would still argue, for example, that psychological laws should also apply to
unconscious mental states with wide content, and we need to ascribe narrow contents to
such states for the formulation of such laws. To distinguish between these different
motivations for narrow content, let us use "subjective content" to designate
those properties of our conscious thoughts that enable us to form correct judgements of
their intentional properties. I have argued that whatever subjective content is, it is not
something that satisfies principle (A).
In fact, if introspection is anything to rely on, there is some reason to think that the
subjective content of a conscious thought is some kind of phenomenal property. Conscious
and deliberate thinking often seem to be a matter of silent speech, and we often seem
to hear words and phrases in our heads as we think. What this suggests is that there is
something which it is like to have a conscious thought, and a difference in the contents
of our conscious thoughts can correspond to a difference in what it is like to have those
thoughts. There is some plausibility to the idea that such phenomenal properties enable us
to identify the intentional properties of our conscious thoughts from a first person
perspective. Whatever the nature of such phenomenal properties might be, at least it seems
clear that unconscious occurrent thoughts do not have phenomenal properties. If this is
right, it provides some further evidence that subjective content is not a property
possessed by all mental states with wide content.
3. Narrow Content by Subtraction
So we still have not found an interesting notion of narrow content that satisfies both
principles (A) and (B). But why should we think there has got to be such a notion? I think
one motivation comes from the powerful and plausible intuition that there has got
to be a sense in which our mental states are dependent on what is happening within
us. Although externalism tells us that what we believe and desire are dependent on our
relations to the environment, it surely cannot be the case that what we believe and desire
are independent of our intrinsic nature. Believing that water quenches thirst is
clearly not like being three miles away from a burning barn, where the location of the
object is all that matters. We cannot just dump an object on twin-earth, and thereby
brings it to believe that twin-water quenches thirst. How can we deny that the internal
constitution of an object plays some role in determining its beliefs and desires?
But then it is reasonable to think of having a mental state with wide content as being
determined by two factors : an internal factor that depends only on our intrinsic
properties, and an external factor that has to do with our relations to the environment.
We might then stipulate narrow content as simply the internal component that contributes
to being in that mental state with wide content.
Here we might draw an analogy with the case of weight. Having some particular weight is of
course an extrinsic property. Nonetheless an object having the weight it has is a function
of its mass and the local gravitational field. What the proponent of narrow content aims
to do, is to find some similar way of factoring out the internal component of having a
mental state with a wide content. If every mental state with wide content can be thus
resolved into an internal and external factor, then this notion of narrow content would
indeed satisfy principle (A). Something like this seems to be what Ned Block has in mind
when he introduces narrow content as follows :
Where we have a relation, in certain types of cases we have individualistic
properties of the related entities that could be said to ground the relation. If x hits y,
y has some sort of consequent change in a bodily surface, perhaps a flattened nose, and x
has the property of say, moving his fist forward. ... There is a nonrelational aspect of
propositional attitude content, the aspect "inside the head," that corresponds
to content in the way that moving the fist corresponds to hitting. This nonrelational
aspect of content is what I am calling narrow content.
The idea is that whether the relation "x has a belief with wide content y"
obtains between an individual and a certain content does depend on the extrinsic
properties of the individual. But one might try to factor out those aspects of the
relation that is "inside the head".
However, many philosophers have pointed out that such a move is not entirely innocent.
Hilary Putnam has gone even further to argue that this project of factoring a belief is
doomed to fail. Putnam argues that "there is no one physical state or one
computational state that one has to be in to believe that there is a cat on the mat."
He considers and rejects various candidates (e.g. perceptual prototypes, conceptual roles)
for the narrow content of this belief, and argues that none is to be found. I think that
Putnam is indeed correct on this point, but it is not clear why a proponent of narrow
content should be troubled by this line of argument at all. For this criticism presupposes
that tokens of beliefs with the same wide content must all have the same narrow
content. However, there is no reason why a proponent of narrow content is committed to
this assumption. Consider again the analogy with weight : having a weight of five grams,
like having the belief that water quenches thirst, is an extrinsic property. But is there
a single intrinsic property that all objects must have in order to weigh five grams?
Objects that weigh five grams can have different masses and differ in other intrinsic
properties in all sorts of ways. Of course, they do share the intrinsic property of having
a non-zero mass. But having a non-zero mass does not contribute to explaining why an
object has the weight it does, even given the strength of the gravitational field it is
in. It is the property of having the particular mass it has that is the explanatory
internal property. This is also true of Block's example of hitting, where the internal
aspects of different instances of hittings can be different, even if the subjects and the
patients are the same.
Similarly, a proponent of narrow content can cheerfully agree that there is no single
interesting intrinsic property shared by all those who believe that there is a cat on the
mat. It is only natural that people who share this belief differ in all sorts of ways in
their computational states and internal constitution. They do share the property of having
a belief of course, and perhaps it might be argued that this is an intrinsic
property. After all, although externalist arguments show that for many contents, having a
belief with those contents depends on how one is related to the environment, they
certainly do not show that having a belief with some content is also an extrinsic
property. But even if having a belief is a intrinsic property, this is not what narrow
content is. The narrow content of a particular belief token is determined by the internal
factor that goes toward explaining why the subject has that particular belief. Two belief
tokens with the same wide content might nonetheless have different narrow content. But
this is no more problematic than there being two things having the same weight but
different mass.
I can think of no strong objection to this way of understanding narrow content, where
narrow content is primarily a property of mental state tokens and not of state types. But
I think the following point is worth noting. Recall that narrow content is supposed to
satisfy two principles :
(A) Content is dualistic : every belief or thought that has a wide content also has
a narrow content.
(B) Narrow content is narrow : if an individual has a belief (or thought) with a
narrow content N, then every possible duplicate of that individual also a belief
(or thought) with the same narrow content N.
I have argued that subjective content is not quite narrow content because it does not
satisfy principle (A). On the current proposal where each belief token with wide content
is to be factored into an internal and external component, principle (A) is indeed
satisfied if narrow content is identified with the internal component. But we do not know
whether this notion of narrow content satisfies principle (B) or not. We do know for
example that Jane's belief that water quenches thirst is determined by a combination of
internal and external factors, and the narrow content of that belief token is whatever
intrinsic property it is that in part explains why Jane has that particular belief. Call
this property P. Whatever P is, by definition it is an intrinsic property
that is shared by twin-Jane. But it does not follow that P also has to be the
narrow component of twin-Jane's belief that twin-water quenches thirst. It is true that
twin-Jane's belief is also a function of two factors. But we have as yet no reason to
think that the internal factor of this belief also has to be the property P. Or for
that matter, we have no reason to think that P has to be the narrow component of any
belief-token of twin-Jane. If the narrow content of Jane's belief is to be identified with
the property P or is something that is determined by P, the present account
of narrow content leaves it open that none of the belief tokens of twin-Jane has the same
narrow content as Jane's belief that water quenches thirst. So narrow content might turn
out not to be narrow! ( Perhaps an analogy might help. Imagine Jane whistling and walking
about the room with Matthew nearby. Matthew finds the whistling extremely annoying but he
does not mind her pacing to and fro. On twin-earth, twin-Jane has the same intrinsic
properties and also whistles and walks about the same way. But twin-Matthew finds her
pacing unbearable instead even though he has no trouble with her whistling. Now Jane has
the extrinsic property of irritating Matthew which twin-Jane lacks, and twin-Jane has the
property of irritating twin-Matthew that Jane lacks. Obviously, the internal component of
Jane's irritation of Matthew is different from that of twin-Jane's irritation of
twin-Matthew. )
Notice that the problem is not due to the fact that we have not specified what kind
of property P is supposed to be. For example, if a conceptual role theory is
correct, then we know that property P has to do with the internal conceptual role
of Jane's belief. But still it does not follow that the conceptual role that goes toward
explaining why Jane believes that water quenches thirst is the same conceptual role that
goes toward explaining why twin-Jane believes that twin-water quenches thirst.
Perhaps one might ask, why is it important that these two belief tokens should have the
same narrow content? In a way, this is not really important, if all that one aims to do is
to identify the internal and external factors that having the belief tokens consist in.
The success of such a project does not require that narrow content have to satisfy
principle (B). However, one of the main motivations for introducing narrow content in the
first place is that it is needed for psychological explanation. On this line of thought,
externalism shows that the mental states of duplicates need not share the same wide
content. It is then argued that this makes the wide contents of mental states unsuitable
for psychological explanations, and that it is their narrow contents that psychological
explanations should appeal to. On the present proposal, given an individual I who
is in a mental state m with wide content W, the narrow content of m
is stipulated as the internal component that explains why I is in state m
with wide content W. But as we have seen, it just does not follow from such a
stipulation that the mental states of a duplicate of I will have the same narrow
content. Of course, I have no argument that they will be different either. But I think it
is common ground that if narrow content is relevant to psychological explanation, then it
has to satisfy principle (B). So, without a more detailed account of how we are to factor
a mental state into an internal and external component, it is not clear whether this
present notion of narrow content is of any use in providing psychological explanations.
Furthermore, there is certainly no a priori reason to think that such a notion of
narrow content has an important role to play in psychology, even if it satisfies principle
(B). In the case of weight, its internal component does turn out to be an important
physical quantity. But it is worth noting that if we have a scientific theory that makes
use of some extrinsic property or relation, even if we can factor out the relevant
internal components of instances of such properties and relations, there is in general no a
priori guarantee that such components are of any interest to that theory.
This is particularly clear in cases where we are dealing with a complex system, where the
theoretical concepts of interest are mainly concepts of extrinsic properties or relations
of dependence. Consider extrinsic properties in economics such as having a particular
price, or having some particular level of income. There are economic generalizations about
such properties and their like, and of course we do not want to say that the intrinsic
properties of an object is completely irrelevant to its instantiation of such properties.
It is not clear how we might factor out the internal contribution that an object makes to
its having the price it has. Nonetheless in the formulation of economic generalizations,
we would not expect to appeal to those intrinsic properties thus factored out, or to
appeal to the concept of narrow price. Why should we think that it is different in the
case of psychology? One might hold that there are generalizations that relate different
mental states with wide contents wide, without thereby committing to the view that there
are generalizations to be discovered by factoring such intentional mental states into
their internal and external components. So why should we think that psychological
explanations have to make use of a notion of narrow content that satisfies principles (A)
and (B)? This is the issue to which we now turn.
3. Narrow Content and Psychological Explanation
Arguments for the need of narrow content in psychology originally came as reactions to
externalist arguments such as those based on the twin-earth thought-experiments. But why
do these thought-experiments show that mental states that have wide contents also have
narrow contents? What externalism shows is that intentional mental properties, such as
having a belief with some particular wide content, are extrinsic properties that do not
supervene on a believer's intrinsic properties. But it does not follow from their being
extrinsic properties that they are irrelevant to psychological explanation. Consider for
example David Lewis's proposal that to causally explain an event is to provide information
about its causal history. This seems to be a plausible proposal of what it is to provide a
causal explanation. We have a good causal, psychological explanation of an event e
to the extent that the information provided about the causal history of e is
psychological, and relevant to the purpose at hand. But then it would seem that extrinsic
properties such as having a certain mental state with such and such a wide content can be
relevant. I can provide information about the causal history of an event e by
telling you that the history includes a mental state of attitude type A with a wide
content C, and by receiving this information it allows you to rule out other
possible casual histories of e. Thus I might causally explain why Jane went out of
her office, by saying that she wanted to drink some water from the faucet outside. My
explanation would of course be wrong if the causal history of Jane's action does not
include a desire for water, but if it is does, then surely this is a case of successful
psychological explanation. So, if mental states with wide contents can be relevant to
psychological explanations, why do we have to postulate that they have an additional kind
of content for explanatory purposes?
Of course, it is not unreasonable to think that when it comes to more detailed scientific
explanations in psychology, we would also have to appeal to psychological states and
processes that are determined only by the subject's intrinsic properties. But the fact
that intrinsic properties are relevant to psychological explanations do not show that we
need to assign narrow contents to mental states with wide contents. Consider the computer
model of the mind, which I assume provides a faithful picture of much of cognitive
science. On this conception, a major task of psychology is to identify the computational
processes and representations that are causally involved in our mental lives. On the
semantic or informational level, we might try to find out what information is made
available and processed by those computational mechanisms, what properties of the organism
and the environment are mentally represented. But we might also study these computational
algorithms and representations at a formal level, abstracting away from semantic or
informational properties. Finally, we might also try to discover how these representations
and algorithms are related to the underlying biological tissues, such as our neural
structures. One would have thought that at the formal level, the same computational
algorithms and representations are to be found in for example Jane, twin-Jane and their
duplicates. It is surely the task of psychology to discover these narrowly individuated
computational properties. These properties are therefore good candidates for providing
psychological explanations that do not appeal to extrinsic properties. Furthermore, such
computational explanations do not proceed by attributing narrow contents to mental states
that have wide contents.
Of course, explanations do come at different levels, and one might think that there is a
need for psychological theorizing at a level more general than particular computational
mechanisms. Instead of considering Jane and twin-Jane, we might consider perhaps
cousin-Jane on twin-earth whose mental representations and processes are rather different.
Perhaps the sentences of her language of thought, if there is one, have a somewhat
different syntax, and that the rules that operate on those representations are different
from that of Jane as well. One might argue, surely one should not preclude a priori
that there might be a level of psychological explanation that applies to both Jane
and cousin-Jane? But the mental states of Jane and cousin-Jane have different wide
contents, and it is supposed to be the case that their computational representations and
algorithms are rather different even at the formal level. Presumably we do want a theory
of psychology that captures the generalizations that apply to both of them. Would this not
be a case for attributing narrow contents to their mental states?
However, I think this line of reasoning takes the idea of a "level" more
seriously than necessary. There is no single level that is the formal level, since
algorithms and representations can be classified by their formal properties at different
levels of abstraction and in ways that cross-cut each other. Formally different
computational algorithms can for example be classified according to whether they are
deterministic or not, and different systems of representations can be classified together
according to the formal features of their grammars. One might provide generalizations
about the properties of a class of algorithms and representations in terms of such and
similar features, and this might prove to be of some importance in psychological
theorizing. As it stands, the case of Jane and cousin-Jane is under-described. The fact
that there are some respects in which their computational architectures are
formally different does not preclude there being formal generalizations that apply to them
both.
Similarly, although two individuals can differ in the wide contents of their thoughts, I
think it will be a mistake to conclude from this fact alone that there cannot be (wide)
intentional generalizations that apply to them both. ( So even if there are no interesting
formal generalizations that apply to both Jane and cousin-Jane, it does not follow that
there are no intentional ones that do. ) However, some such assumption seems to have been
made by Fodor's former self in his previous arguments for narrow content. The line of
thought is that (a) psychology should provide intentional generalizations that subsume the
intentional mental states of both Jane and twin-Jane, but (b) this would not be possible
unless their intentional mental states have narrow content. If I understand Fodor
correctly, I think he no longer accepts (a), even though he still subscribes to (b). We
shall come to Fodor's current view shortly, but for the time being, let us focus on (b)
instead. What might be the reason for thinking that the same intentional generalization
cannot apply to both Jane and twin-Jane? Some hints are to be found in this revealing
passage :
What the Putnam/Burge examples show is that the broad, folk-theoretic
notions of semantic property exhibit a previously unnoticed relativization to context.
Narrow content wants to generalize over the contexts to which broad content relativizes,
hence permitting psychological laws which hold without respect to context.
Abstracting from context sensitivity is a standard way of achieving scientific
generalization. We could have done physics with weight, but the price would be
context sensitivity in the laws of mechanics. Mass generalizes over the contexts to
which weight relativizes and is the preferred parameter for precisely that reason. Such
precedents would motivate narrow content even if metaphysical arguments for supervenience
didn't.
The idea seems to be that if scientific laws subsume states or objects by their
extrinsic properties, properties that are "context-sensitive", then such laws
will fail to hold across contexts where the extrinsic properties differ. We want the same
intentional psychological laws to apply to both Jane and her duplicates, but the worry is
that if intentional laws subsume mental states by their wide content, then they might
apply to Jane but not twin-Jane because of differences in the wide content of their mental
states. If there is to be an intentional law that subsumes both Jane's belief that water
quenches thirst and also twin-Jane's belief that twin-water quenches thirst, it has to be
the case that these two beliefs have some other content in common that does not depend on
the environment.
But if this really was Fodor's worry, then it seems to be based on a simple confusion. It
is one thing to claim that psychological laws subsume mental states by extrinsic
properties whose instantiation depends on the context. It is surely quite another to say
that the laws formulated in terms of such properties are themselves context-dependent and
have limited validity. The second does not follow from the first at all. In fact this
should be clear from Fodor's own comparison with classical mechanics. Classical mechanics
tells us that if an object with initial velocity u moves with a constant rate of
acceleration a, then after a period of time t it will have travelled a
distance d given by the formula d = ut + ½(at2).
Like weight, the initial velocity and rate of acceleration of an object are its extrinsic
properties, and are thus "context-sensitive" in Fodor's sense. Yet this
particular law of motion specified in terms of such properties is surely not
context-dependent in any interesting sense. It holds with respect to contexts where
objects may vary in their initial velocities and rates of acceleration in all sorts of
ways. There is no fear of loss of generalization because it subsumes objects by their
extrinsic properties, nor should we say that in additional to an object's initial velocity
and rate of acceleration it also has a narrow velocity and a narrow acceleration!
The same point can be made with regard to Fodor's worry that the same intentional laws
will fail to apply to duplicates if they subsume mental states by their wide content. As
many authors have argued, intentional generalizations are generalizations that quantify
over contents. This includes Fodor himself who repeatedly points out that the formulation
of such generalizations do not mention particular contents at all. In Psychosemantics,
he writes,
[P]sychological theories typically achieve generality by quantifying
over the objects of the attitudes. In consequence, many of the most powerful
psychological generalizations don't care about content per se; what they care about is
only relations of identity and difference of content.
But if intentional generalizations do not "care" about particular
contents, why can't we have the same intentional generalization subsuming the different
intentional states of the twins? One such generalization that Fodor mentions in the first
chapter of Psychosemantics is this : "if someone believes that Fa, then
ceteris paribus, that person believes $x(Fx)."
Let us assume that this is indeed a true generalization. One would have thought that this
will be a generalization that applies to both Jane and twin-Jane. Jane believes that water
quenches thirst; twin-Jane believes that twin-water quenches thirst. Presumably the
generalization predicts that they both believe that something quenches thirst. If as Fodor
says, psychological generalizations do not mention particular contents but quantify over
them, there is then no reason to believe that such generalizations cannot subsume the
different intentional states of Jane and twin-Jane, and maybe that of cousin-Jane also.
I should perhaps explain why I have chosen Fodor as my target here even though he has
recently come to the conclusion that perhaps psychology can make do without narrow
content. The reason is that despite his change of position, he still seems to hold that
twin earth cases show that broad intentional laws will miss generalizations. In his Jean
Nicod Lectures, he writes, "Twin cases say : if you insist that computationally
implemented intentional laws be broad, you will miss generalizations in virtue of which my
psychology is the same as that of my computationally identical twin." But why should
this not be a problem for psychology? If I understand Fodor's response correctly, his
answer is that it is alright even if psychology misses such generalizations. This is
because (i) psychology is a special science whose intentional laws are ceteris paribus
laws, and (ii) twin earth contexts are ones where the ceteris paribus clauses of
intentional laws fail, and (iii) a special science need not provide generalizations that
deal with contexts in which ceteris paribus clauses fail. However, if the point I
have just made is correct, there is no reason to think that twin-earth cases pose a
problem for the generality of intentional laws in the first place. Although Jane and
twin-Jane differs in the wide contents of their mental states, there is a
straightforward sense in which they have the same psychology : their intentional mental
states are subsumed by the same set of intentional laws. Not only that, as suggested
earlier the computational states and processes that we find in both of them, as
characterized in formal terms, are exactly the same. One does not therefore have to
postulate narrow content then, to satisfy the desire for psychological explanations that
appeal to only intrinsic properties, or for intentional laws that subsume the mental
states of individuals on both earth and twin-earth.
4. Conclusion
The main point I have been arguing in this paper is that there is no straightforward
inference from externalism to narrow content. I do not want to deny however, that perhaps
some of our mental states do have contents that are determined by our intrinsic
properties. As suggested in the beginning of this paper, it might be that externalism is
not true of some of our beliefs. Furthermore, I have not addressed the issue concerning
externalism with regard to the contents of perceptual experiences. Finally, I think it is
not unreasonable to think that some of our computational representations also have
contents that are determined by our intrinsic properties. Perhaps there are subpersonal
states that can be said to represent the distribution of light as detected by our retinal
cells. It is also reasonable to think that in any complicated computational system such as
ours, there are states which monitor other states of the internal environment, such as
representing whether a certain module has successfully carried out some operation. If it
makes sense to ascribe content to such states at all, then these would be good candidates
of states whose contents do supervene on their subject's intrinsic properties. What I have
been trying to argue against, however, is the view that there has to be a useful notion of
narrow content that satisfies principles (A) and (B). I think we have good reasons for
thinking that subjective content does not satisfy principle (A), and I pointed out that
narrow content introduced as the internal component of a wide mental state need not
satisfy principle (B). Finally, considerations based on psychological explanations do not
motivate such a notion of narrow content either. Of course, I have not shown that
there cannot be a theoretically important notion of narrow content that does satisfy these
two principles. But I think this is as it should be. Narrow content is a theoretical
concept, and there is no better argument for the need of such a concept other than by
embedding it within a theory and show how it bears fruit. What I have been trying to
resist, are relatively a priori arguments intending to show that the project has to
proceed in some particular way.