Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

The Transcendental Significance of Phenomenology

Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

PSYCHE: http://psyche.cs.monash.edu.

au/

The Transcendental Significance of Phenomenology

Stephen L. White
Department of Philosophy
Tufts University
Medford, MA 02551
stephen.white@tufts.edu
S. White 2007

PSYCHE 13/1, April 2007


KEYWORDS: agency, demonstratives, experience, meaning, phenomenology,
skepticism, transcendental arguments
Abstract: There is a well-known line of thought, associated with Donald Davidson, that
connects the notion of a perceptual givenof non-linguistic or non-conceptual experience
of the worldwith skepticism. Against this, I argue that the notion of what is given in
perception leads to skepticism only on certain interpretations. I argue, in fact, that there
must be perceptual experience such that there is something it is like to have it, or that
would provide the subject of a phenomenological analysis, if we are to block skepticism in
its most radical forms. In particular, I claim that there is a distinctive phenomenology of the
experience of agency. These phenomenological claims are conclusions of a transcendental
argument according to which our having such experience is a condition of our having a
meaningful language. Moreover, the same transcendental argument is sufficient to show
the incoherence of radical skepticism about the external world. And I argue that the proper
understanding of perceptual experienceas object involvingrenders the standard
objections to transcendental arguments ineffective.

1. Introduction
Transcendental arguments in their modern guise consist of premises describing
conditions of our having a meaningful language (having intelligible thoughts, being
subjects of experience) and an anti-skeptical conclusion. In this paper I argue that there is
a distinctive phenomenology of agency, and the argument will be transcendental in this
sense. But what exactly is the skepticism at issue? Most immediately, of course, it is
skepticism about the meaningfulness and truth of our talk about the phenomenology of
experience in general and agency in particular. Thus it entails skepticism about there
being something it is like to experience the world and to experience ourselves as agents.

PSYCHE: http://psyche.cs.monash.edu.au/

The argument, then, will be that it is a condition of our having a meaningful language that
we actually have such subjective experience of the world. And the claim that there is a
distinctive phenomenology of agency is central to the argument.
The argument for a phenomenology of agency, however, stems from the
requirement that a meaningful language be grounded in an appropriate relation to the
world. And this requirement connects the conclusion about agency to another conclusion
of a more traditional sort: that the conditions for our having a meaningful language are
intimately tied to the conditions for its being possible to justify empirical beliefs about
the external world. To put the point in different terms, what we might call meaning
skepticismskepticism about the satisfaction of the conditions of our having a
meaningful languagegoes hand in hand with traditional epistemological skepticism
about (the possibility of justifying our beliefs about) the external world. Thus the
argument that a certain kind of phenomenology is necessary for our having a meaningful
language is also an argument that it is necessary for our having justified empirical beliefs.
Moreover, I shall argue that the phenomenology in question is sufficient to counter the
fundamental skeptical argument against the possibility of such justification.
This may seem surprisingeven paradoxicalto those familiar with the claims
of Donald Davidson and others that it is the postulation of experiential intermediaries
between our empirical beliefs and their objects that leads to skepticism about the external
world. As Davidson puts it,
. . . Quine and Dummett agree on a basic principle, which is that whatever there is
to meaning must be traced back somehow to experience, the given, or patterns of
sensory stimulation, something intermediate between belief and the usual objects
our beliefs are about. Once we take this step, we open the door to skepticism, for
we must then allow that a very great manyperhaps mostof the sentences we
hold to be true may in fact be false. It is ironical. Trying to make meaning
accessible has made truth inaccessible. (Davidson 2001, pp. 144-145)
In what follows I shall criticize some of the assumptions behind Davidsons position.
What is ironical, given this fundamental disagreement, is that Davidson and I agree on
many of the basic theses in this area. Unsurprisingly, we understand them very
differently. The moral, I think, is obvious. The concepts currently invoked in these
discussions are seriously underspecified. Any useful consideration of these issues
requires that they be unpacked and made more explicit and more precise in ways that
have not yet been attempted.

2. The Transcendental Argument in Outline


What are the most fundamental conditions for our having a meaningful language?
Certainly among them is the condition that language be grounded in a connection to the
world. We can see what this means by reflecting that there could not be an infinite
backward regress of lexical definitions, of words defined in terms of other words. That is,
there could not be an infinite backward regress of merely word-to-word connections. If
we are to have a meaningful expression, such a regress must terminate in a connection to
the worldan ostensive definition, a demonstration, a word-to-world connection. In the
absence of such word-to-world connections, we would have nothing over and above a

PSYCHE 2007: VOLUME 13 ISSUE 1

PSYCHE: http://psyche.cs.monash.edu.au/

formal calculus. That is to say, we would have nothing more meaningful than an
uninterpreted formal language, formal rules of inference, and a subset of the set of wellformed formulas as axioms. Certainly we could, in an obvious sense, specify inferential
roles in such a language. But this simply shows that such a formal inferential role is not
sufficient for what we think of pre-theoretically as meaning.
What, then, constitutes the connection necessary for a meaningful language? Is it
anything more than the causal connection that virtually everyone acknowledges between
our referring expressions and their referents in the world? Intuitively, if a speakers grasp
of meaning is concerned, a causal connectionone to which the speaker need have no
accessdoes not seem an appropriate substitute for an ostensive definition or an act of
demonstration. But why? And what is the connection between such intuitions about
meaning and the kinds of transcendental arguments that concern us?
The answer is that the idea of a meaningful language is complicated by the wide
range of things that might count as meaning or contentbroad content, narrow content,
language meaning, speakers meaning, meaning within an idiolect, meaning in virtue of a
division of linguistic labor, and so forth. The notion of meaning most directly relevant to
transcendental arguments and to skepticism, however, is that in terms of which we can
rationalize a subject or ourselves. That is, it is the notion in terms of which we can do
justice to the rationality and coherence of the subjects beliefs and desires, and to the
practical and theoretical rationality of the subjects actions and speech. And this is
because it is generally (and I believe correctly) assumed that such rationality and such
coherence are constitutive of (are internally connected to) the very possibility of ascribing
intentional states. Certainly Davidson is committed to such an assumption. (Davidson,
1973, Lewis 1974) Were it impossible by the skeptics own lights to make sense of, or
ascribe such content to, his or her own thoughts or utterances, we would clearly have a
successful transcendental argument. Skepticism of the sort in question would have proved
self-defeating in the strongest possible sense.
The kind of meaning, then, that is relevant for our purposes is, to use the Fregean
term, cognitive significance. This gives us a second necessary condition on meaning. Not
only must a subjects words be grounded in something non-linguistic. In addition this
grounding must be in accordance with our commitment to doing justice to the rationality
of the subject, and hence to the cognitive significance of the subjects terms. At a
minimum this means that ascriptions must satisfy Freges constraint in the various
versions in which it is currently formulated. (Stephen Schiffers version: If x believes y to
be F and also believes y not to be F, then there are distinct modes of presentation m and
m such that x believes y to be F under m and disbelieves y to be F under m. Schiffer,
1978, p. 180.) As I have argued elsewhere, the satisfaction of these conditions is
necessary but not sufficient for doing full justice to the rationality of subjects and to the
cognitive significance of their terms (White 2007). But these are not the issues that
concern us here. I shall say, as a kind of shorthand, that the condition for a languages
being meaningful with which we are concerned is that it be grounded subject to Freges
constraint. But it is more accurate, if more cumbersome, to say that the condition requires
us to ground the language in such a way as to do full justice to the rationality of speakers.
And this requirement commits us to grounding language in such a way that we can solve
all the relevant Frege problems as they arise.
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

PSYCHE: http://psyche.cs.monash.edu.au/

We have, then, the relevant conditions for meaningfulness. And we have, as well, our
answer to the question whether language can be grounded in a merely causal connection.
The answer clearly is no. As is well known, two expressions, such as Hesperus and
Phosphorus, may be causally connected in the relevantly appropriate way to the same
object, even though they differ in cognitive significance. And if it is objected that the
causal chains themselves will be different, then it must be remembered that these are
differences to which a normal subject need not (and in general will not) have access.
Thus whatever role causal relations play, they are not themselves sufficient to ground
meaning in the sense of cognitive significance. We need, it seems, some intermediaries,
accessible to the speaker, to serve as the appropriate modes of presentation that the
satisfaction of Freges constraint requires.
In the case of Hesperus and Phosphorus, of course, there is evidently no
problem. Hesperus and Phosphorus, we might suppose, are associated with different
descriptions, both of which are fully accessible to the subject. Such descriptions might be,
for example, the first heavenly body visible in the evening and the last heavenly body
visible in the morning, respectively. But notice: The association of a description with a
proper name just postpones the real problem. This association is just one more word-toword connection, and what we required was a connection between words and the world.
It seems, indeed, that we face a dilemma: If we try to establish a direct link between
words and the worldone unmediated by linguistic-descriptive contentby appealing to
a causal connection, we violate Freges constraint. And if in response we postulate
descriptive modes of presentation, we reintroduce the backward regress of word-to-word
connections.
We can see the difficulty more clearly if we consider Freges problem in its
demonstrative guise. In Gareth Evans example, one may prepared to accept, pointing to
the bow of a ship out one window, That ship was built in Japan. Equally, one may be
unprepared to accept, That ship was built in Japan, pointing to the stern of a ship out
another window. And this is possible even though one has, without realizing it, pointed to
the bow and stern of the very same ship (Evans 1982, p. 84). The temptation again is to
introduce descriptions, but this move, as we have seen, is self-defeating. We need some
direct or demonstrative connections to ground language. And if in every case we go on to
postulate new descriptions in order to satisfy Freges constraint, we will have failed to
explain what makes the language meaningful.
Freges constraint, then, imposes a very strong requirement on any account that
purports to do justice to the meaningfulness of a speakers utterances and to the contents
of his or her beliefs. And it seems we have no choice but to see the difference in cognitive
significance of the two references to the same ship as a matter of the two different sets of
sensory experiences that accompany them.

3. The Humean Argument for Skepticism


In spite of the arguments, however, the imposition of such a constraint and the
willingness to ascribe such a role to perception will seem to some misguided. Davidson,
as we have seen, has argued that the postulation of epistemic intermediariesthe
testimony of the senses: sensation, perception, the given, experience, sense data, the
passing showleads to scepticism (Davidson 2001, pp. 141 and 144-145). Davidson
PSYCHE 2007: VOLUME 13 ISSUE 1

PSYCHE: http://psyche.cs.monash.edu.au/

doesnt say what argument for skepticism he has in mind, nor what that skepticism
consists in. It isnt clear, for example, whether the skepticism in question follows from
the thesis that many or most of the sentences we hold true may be false or is identical
with it. And if the latter, some might object that the so-called skepticism is merely
common sense and, in any case, nothing with which we cannot comfortably live. (I shall
have more to say about this later.)
There is, though, a form of skepticism so strong it precludes either of these
reactions. Indeed one of the sub-themes of this discussion is that Davidson and others
have seriously underestimated the significance of the skepticism that follows if we accept
a sensory given as traditionally conceived. What I shall argue, however, is that Davidson
misdiagnoses the deep source of this skepticism. A correct diagnosis, I shall claim, allows
us to see an adequate phenomenology of experience not as the source of skepticism, but
as the only possible defense we have against it.
The argument for a stronger form of epistemological skepticism than the claim
that many or most of our beliefs could be false is Humes (Hume 2001, Section XII, Part
I, p. 184; cf. Hume 1968, Book I, Part IV, Section II, p. 212). Suppose one considers
ones present conscious perceptual experiences (experiences, say, as of a room, furniture,
etc.). Such experiences, we normally assume, are caused by what they purport to be
experiences ofa room and so forth. But certainly, it seems, the same experiences (or
experiences indistinguishable from these) could be caused in completely different ways
by a dream, an evil demon, the Matrix, and the like. Thus there is a gap between the
character of our experiences and their causes. And bridging this gap requires an inference
(since closing the gap between perception and its causes couldnt itself be a matter of
perception). Because the connection is causal, however, the inference could not be a
priori. Neither, though, could it be a posteriori. For such an inference could only be
justified by a principle connecting the character of perceptual experiences with external
causes. And the legitimacy of appealing to any such principle is exactly what the Humean
skeptic questions. The conclusion is that one couldnt be rationally justified in preferring
the commonsense hypothesis (that ones experiences are caused by a real world much as
we take it to be) over any of the skeptical alternatives.
This conclusion is obviously much stronger than the proposition that many or
most of our beliefs could in fact be false. But what has not been sufficiently seriously
considered is that there is an even stronger conclusion in the offing. Suppose we accept,
for the sake of argument, Humes conception of experience as interpreted by
contemporary sense-datum theorists. What we are given, then, in visual experience (to
take the most important example) are mental entitiessense-datathat actually have the
visual propertiesshapes, colors, relative sizesthat external objects seem to have. The
sense-datum normally caused by a round table viewed at an angle, for example, will be
elliptical, and that normally caused by a white cup in red light will be red. On such a
conception, we never experience external objects directlyindeed, we could not do so in
principle. Doubts about the connection between sense-data and the world, then, could
never in principle be resolved, because all we could ever get are more sense-data.
Such a conception of experience clearly supports the Humean argument for
epistemological skepticism, since exactly the same sense-data could be caused in
radically different waysincluding all those suggested in the usual skeptical scenarios.
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

PSYCHE: http://psyche.cs.monash.edu.au/

And such scenarios provide the doubts that could never be resolved in principle. But now
put aside, for the time being, the epistemological question: whether we could ever know,
or be rationally justified in believing, that ordinary external objects really exist. Consider
a more basic question. Can we even attach any sense to the terms in our external object
vocabulary? Grant, for the moment, that we understand the idea of a pattern among all of
our sense-datapast, present, and future, actual and possible. (Of course even this
assumption would be problematic for a consistent Humean. I shall discuss this below.)
For example, what one thinks of as the table involves many predictable sense-datum
patternsincluding the one in virtue of which one would take oneself to be walking
around it. What justifies the assumption that what we mean by table is any more that the
totality of such patterns?
Suppose we stipulate that table* is to be understood in just this way. Then we
can ask what justifies the assumption that table means anything different from table*?
Obviously the two terms are experientially equivalent on the Humean or sense-datum
view of experience. Thus it seems that anything one could point to in characterizing the
meaning of table is something one could point to in characterizing the meaning of
table* and vice versa. What about a difference in inferential roles? It is tempting to
think that we could define a physical object as a cause of sense-data and hence as distinct
from any collection of them or pattern among them. But this just pushes the issue one
step back, presupposing as it does that we have a concept of causation as something more
than just another sense-datum regularity. (Hume was, of course, famously skeptical on
this point.) The same point applies if we appeal to counterfactuals to talk about the back
of the table that exists at the same time we are looking at the front (and not just if and
when we walk around to see it). And indeed, the point is a perfectly general one about
any appeal to a difference between the inferential roles of table and table*.
So far, the conclusion is merely phenomenalistic. And those who thought we
could learn to live with Humean epistemological skepticismi might suppose that we can
live with this as well. But recall: We have been assuming unproblematic access to
patterns among past, present, and future sense-data and among sense-data that are
possible as well as actual. And the same considerations apply here. We cannot supply any
content to the notion of a past or future sense-datum over and above what we understand
in the notion of the relevant pattern of sense-data in the present. (These sense-data
include, of course, those we class as memories and anticipations). Nor can we assign
meaning to the notion of a possible sense-datum over and above the relevant pattern
among those we actually experience. The upshot is what we might call analytical
solipsism of the present momentwe can attach no meaning to terms that refer to
anything other than our actual, present sense-data or their actual, present properties and
relations. And this, surely, is a form of skepticism too extreme for anyone to live with.

4. The Source of Humean Skepticism


It seems, then, that there is no problem with Davidsons claim that for a certain
conception of perceptual experienceone in which it plays the role of an epistemic
intermediary (in one sense of the term)such experience leads to skepticism. Indeed,
as we have just seen, Davidson grossly underestimates the seriousness of the skeptical
consequences. But the question is how we should generalize the result. What, in other
PSYCHE 2007: VOLUME 13 ISSUE 1

PSYCHE: http://psyche.cs.monash.edu.au/

words, is the source of the skeptical upshot? We can put the question another way by
asking what the skeptic needs for a Humean skeptical argument to go through.
Notice first that the Humean skeptic is clearly not trying to call everything into
doubt simultaneously. There is a well-defined domain and a corresponding set of
propositionsa posteriori propositions about the objective, external worldwith regard
to which it is claimed that we have no rational basis for our beliefs. And there is a
counter-domain and its corresponding set of propositions with regard to which no such
challenge is raised. (For simplicity, I shall drop the distinction between the domains and
their corresponding sets of propositions in what follows.)
What relation, then, between the domain and the counter-domain, is required to
generate Humean skepticism regarding the domain? Certainly a necessary condition is
that there should be no conditions regarding the counter-domain (where condition
regarding a domain is used to include its existence) that are sufficient for any condition
regarding the domain. (This would rule out, for example, phenomenalism.) We can put
this point by saying that Humean skepticism about a domain requires a counter-domain
that is domain neutral. In the case of Humean skepticism about the external world, there
must be an external-world-neutral counter-domainthat is, a counter-domain that can be
completely characterized in a way that is logically and conceptually neutral with regard to
the existence and character of the external world. Clearly Humes domain of conscious
experience, understood in term to sense-data, was intended to (and arguably does) have
this feature. And equally clearly, Davidsons radically externalist conception of content,
according to which the objects of a belief are the causes of that belief, does not (Davidson
2001, p. 151). This is not the end of the story, however, since, Davidsons radical
externalism cannot satisfy Freges constraint. Thus it cannot do justice to the rationality
of the subject to whom the mentality is ascribedan extremely ironic result, given the
constitutive role that rationality plays for Davidson himself in the ascription of intentional
states to others.
But, it seems equally clear that this necessary condition (call it the neutrality
condition) is not sufficient. Consider an example I have used elsewhere (White 2004a, p.
276). Imagine someone confined to a windowless room (perhaps under house arresta
latter-day Galileo). Suppose that although he cannot leave the room, he has access to the
outside world via all the usual modern media, as well as unlimited opportunities to
communicate with others outside the room and to receive them as visitors. Clearly this
example satisfies the independence condition. A posteriori propositions regarding the
world outside the room are logically independent of those regarding things inside.2 And,
since the connection between events inside the room and events outside is merely causal,
the external causes of events insidethe causes of the images on TV, the words in the
email, the speech of visitorsmight be radically different from what would normally be
supposed. Thus with regard to the causal sources outside of events inside, one could
entertain obvious analogues of most of the usual skeptical scenarios. In fact, the room and
the events within it could exist even though nothing existed outside. But does this
example really generate outside-the-room skepticism, analogous to Humean skepticism
about the external world? (From now on I shall use Humean skepticism loosely to refer
to Humean epistemological skepticism as generated by the Humean argument we saw
above.)
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

PSYCHE: http://psyche.cs.monash.edu.au/

Intuitively the answer is no. And seeing why will help us see what has gone wrong with
Davidsons argument that the postulation of anything that would ground or justify a
phenomenological description leads to skepticism. First, as we saw, the Humean
argument that there must be an inference from the character of our perceptual experience
to our beliefs about the external world presupposes that we cannot perceive external
objects directlyi.e., without there being an inference in any relevant sense. (Of course,
the question of what counts as a relevant sense will have to be examined, and I shall do so
below.)
With regard to outside-a-room skepticism, however, this presupposition is,
arguably, false. Imagine, for example, that the person who never leaves the windowless
room does so because he is the younger and more resourceful (but equally phobic)
brother of Howard Hughes. Such a person would have, in addition to the resources
already described, the money and access to scientific talent necessary to create prosthetic
perceptual devices (some of which exist, some of which are currently in development)
literally to look through the walls. Alternatively, the instruments already at his disposal
might become virtually transparent with use. In such a case it might be completely
appropriate to say that he is seeing not, say, images on a screen or any other features of
an instrument, but external objects themselves. (And at the limit, it isnt clear that this
alternative is really distinct from the first one. I shall say more about this below.) And, of
course, combinations of these possibilities might give someone with almost limitless
resources a view of the world unlike, and far superior to, that of any normal subject.
Finally, of course, there is nothing in principle to prevent someone in such a situation
from cutting a hole in the wall, orthe most radical solution of allfacing down his
phobias and walking out the door.
These possibilities block any application of the Humean skeptical argument to
generate outside-a-room skepticism. The Humean argument presupposes that any
empirical belief about the external domain must (logically, conceptually) be based on an
inference from what is given in perceptual experience. And if that presupposition is false,
then the argument that any such inference would be unjustified is irrelevant. But there is
another way in which the Humean argument as applied to the case of house arrest fails.
For the Humean argues that no a posteriori inference could get us from our sense
experience as given from the inside to a conclusion about the external source of that
experience. But why not? The answer is that such an inference would have to be justified
by a principle connecting the subjective character of the perceptual experience with
external objects, and any such principle would be among the things taken already called
into question by the skeptic. Hence the appeal to such a principle would beg the question.
But now notice that the subject confined to a room can do experiments to establish inside
the room generalizations about causal connections not only between objects inside the
room, but between objects inside and outside and objects all of which are outside. Can the
Humean skeptic try to disallow this? Can the skeptic claim that we have evidence only
for causal priniciples as they apply within the room? It seems not. What could it be
claimed is special about the distinction between what goes on inside the room and what
happens outside? Suppose it is alleged that the spatial distinction matters. Then
experiments could be performed inside the room that would support the claim that some
causal generalizations established inside are spatially invariant over at least reasonable
distances. And for some events, the distance between events inside and outside the room
PSYCHE 2007: VOLUME 13 ISSUE 1

PSYCHE: http://psyche.cs.monash.edu.au/

will be less than the distance between events inside. If it is alleged that it is a matter of
the materials with which the room is constructed, we can bring those materials inside the
room and experiment with transmissions through themand so forth.
Finally, there is one more way in which the Humean argument goes wrongand
it goes wrong in the original context as well as in this one. It is an assumption of the
Humean that the inference from the character of ones perceptual experience to
conclusions about the external world could not be a priori, since we are talking about
empirical knowledge. But this is open to question in either context, since the fact that the
issue is empirical knowledge does not obviously rule out the possibility of a
transcendental argument for anti-skepticism about the external world. Such an argument
might deliver either the conclusion that some of our beliefs about the external world must
be justified or that some are self-justifying or in no need of the justification that the
Humean assumes their rationality requires.
Could it be objected that this treatment of outside-a-room skepticism is questionbegging? In the first reply we took it as given, apparently without justification, that
because we have unproblematic access to things inside the room, we can suppose that our
access to things outside is sometimes, or at least can be, direct. In the second we seemed
to take it as given that because we have unproblematic access to causal relations within
the room, we can suppose we have such access to causal relations connecting things
inside and outside. And in the third reply we apparently supposed that transcendental
arguments could justify the truth of a realist conception of the external world, or that an
anti-realist conception is more than just another version of skepticism. (Barry Stroud has
questioned both these assumptions.3 I shall return to the issue of transcendental
arguments later.)
What this possible objection points up is that the Humean skeptics argument
must take place against the background of certain theories. I shall argue in particular that
it must take place against a background that includes a philosophical theory of perception
and a philosophical theory of meaning. But how is this claim justified? After all, it flies in
the face of another apparent assumption of Stroudsthat we cannot defeat the skeptic by
appeal to a theory of meaning. Strouds objection to such an appeal to demonstrate the
incoherence of the skeptics claims seems to be that we are as confident in our
understanding of what the skeptic is saying as we could be in any theory that entails its
unintelligibility. (Stroud is explicit about this where verificationism is concerned, but the
point seems to be a general one. Stroud 1984, p. 205). Clearly there is something to this
objection. And in any case, why is it these background theories that are required?
The answer to the last question is easy. The Humean argument says that there
must be an inference from the character of our perceptual experience to the character of
the external world and that neither an a priori nor an a posteriori inference could ever be
justified. And since we can take it (because the Humean does) that a legitimate a priori
inference will be grounded in meaning broadly conceived, the Humean clearly owes us
an account of it if the need arises. (And it will in the context of the possibility of a
transcendental argument for anti-skepticism.) Similarly, since it is claimed that there must
be an inference to the nature of things in the external world, the Humean is committed to
the claim that we cannot in principle perceive objects in the external world directly (i.e.,
in a way that needs no inferential support). Thus the Humean owes us an account of
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

PSYCHE: http://psyche.cs.monash.edu.au/

perception from which this follows if, as it will be, such an account is necessary to clarify
the nature of Humean skepticism itselfand in particular to clarify its nature by
distinguishing it from outside-a-room skepticism. The choice of the two background
theories, then, is dictated by the structure of the Humean argument itself. And, of course,
arguments and theses from other areas, such as metaphysics or philosophy of mind, will
be relevant to the extent they are necessary to make good on the deliverances of the
background theories to which the Humean is committed directly.
Thus the appeal to such background theories, the theory of meaning in particular,
is nothing like the appeal, to which Stroud objects, to such controversial theories as
verificationism. Rather, it is an appeal to the theory of meaning (and perception) not in
the first instance to refute the Humean skeptic, but to clarify the presuppositions of the
Humes argument. More specifically, it is intended to elicit the necessary and sufficient
conditions for such an argument to go through. Indeed, this inquiry into the
presuppositions of the Humean argument is so natural, it is something of a mystery why it
has not been pursued more systematically. I can only speculate that theorists have been
inhibited, either by Strouds argument for the irrelevance of theories of meaning, the fact
that the issues cut across the major divisions within contemporary philosophy, the fact
that it requires serious consideration of such unfashionable topics as sense-data and
phenomenology, or, more likely, some combination of the three.
Suppose that, on the assumption that we have such background theories on the
table, we have specified the domain and counter-domain, and specified our access to the
counter-domain as unproblematic. Then necessary and sufficient conditions for the
Humean skeptical argument to go through regarding the elements of the domain can be
stated as follows.
Regarding the elements of the domain (objects, facts, events, etc.), the Humean
argument for Humean skepticism applies, if and only if, it is a consequence of the
theory of meaning that there are no conditions regarding the counter-domain that
are sufficient (logically, conceptually) for any condition regarding the domain, and
it is a consequence of the theory of perception that none of the elements of the
domain can be perceived or observed directly.
Consider skepticism about the external world against the background of the following
assumptions. The theory of perception says that all we can in principle (as a matter of
conceptual necessity) perceive or be given directly are sense-data. The theory of meaning
says that there is no a priori inference from sense-data to elements of the external world
(as there would be, for example, if phenomenalism were true). Given these assumptions,
the Humean argument applies, and arguably it goes through. (Again we are putting aside
transcendental arguments for the time being).
With regard to the question whether there is a Humean argument for outside-aroom skepticism, the answer is more complicated. This potential form of skepticism was
defined by saying that inside the room we have unproblematic access to ordinary objects.
But, as we have seen, it is not immediately apparent what this entails regarding the
question whether objects outside the room are, in principle, directly perceivable. As we
saw above, the initial argument that there is no such thing as outside-a-room
skepticismthat the Humean skeptical argument cannot be made to applywas

PSYCHE 2007: VOLUME 13 ISSUE 1

10

PSYCHE: http://psyche.cs.monash.edu.au/

inconclusive. For it seemed a possible objection that to suppose objects outside the room
could be perceived directly, even in principle, begged the question against the would-be
skeptic.
This, however, is where the more abstract and general characterization of the
necessary and sufficient conditions for Humean skepticism pays off. For what
philosophical theory of perception could the would-be skeptic appeal to? If the
background theory of perception said that all we could perceive directly were sense-data,
then the Humean argument we have been considering would show (arguably) that we
could not have unproblematic access to ordinary objects inside the room. And this
violates one of the conditions to which the would-be skeptic is committed. For the wouldbe skeptic wants to claim that we can make sense of outside-a-room skepticism on the
analogue of Humean skepticism about the external world. Thus the proponent of
outside-a-room skepticism is committed to a background theory of perception that
gives us unproblematic access to ordinary things and events within the room, and the
Humean argument shows that this means direct perceptual access to some such objects.
However, a perceptual theory that gives us direct perceptual access to ordinary objects
within the room will give us, at least potentially, perceptually direct access to objects
outside. And, as we have seen, this is all we need to defeat the Humean argument in the
context of house arrest.
What is more relevant in the broader context than the commitments of the wouldbe skeptic about the world outside the room is that we think that (sometimes) we have
access to ordinary external objects unmediated by anything like sense-data. And the point
of the comparison between Humean skepticism and a range of (possible) analogues is to
see what is essential to Humean (or as I shall call it later philosophical or
transcendental) skepticism, and whether in fact it arises naturally out of our ordinary
epistemic practices. If a Humean skeptical argument required a sense-datum theory of
perception among its theoretical background conditions, as some seem to assume, it
would be easily blocked (Searle 1983, pp. 71-76). (As we shall see shortly, however, it
doesnt.) If, as Stroud and the other new Humeans seem to hold, there are no
controversial theoretical presuppositions that could conceivably be rejected, then it seems
that such skepticism is indeed a fact of life. Settling this issue will be crucial later to
distinguishing between philosophical and what I shall call real skepticism.
External-world skepticism based on a sense-datum theory is genuine
philosophical skepticism because for any hypothesis regarding the external world, there
is, in principle, no way of obtaining evidence for or against it. Thus, as Hume claims,
there can be no rational justification for preferring one hypothesis about the external
world over another. Outside-a-room skepticism isnt Humean or philosophical skepticism
because there are perfectly reasonable ways of gathering evidence for and against the
various hypotheses that might be entertained about the character of the world outside.
But, are there other cases of genuine philosophical or Humean skepticism that
dont involve the sense-datum theory? The answer is yes, and by now the reason should
be obvious. Consider any form of scepticism, and suppose that it draws a principled
distinction between elements in the domain and elements in the counter-domain. Suppose
also that it proceeds on the basis of a theory of perception according to which that
distinction entails that we have unproblematic access to (including direct perception of)
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

11

PSYCHE: http://psyche.cs.monash.edu.au/

some of the elements of the counter-domain and no possibility of direct perception, even
in principle, to elements of the domain. Then assuming that there are no conceptually
sufficient conditions regarding the counter-domain for conditions regarding the domain,
we have an example that yields philosophical skepticism. And this is true regardless of
what the theory of perception is like in respects other than those specified. Moreover, as
we shall see, there are many such theories of perception, of varying degrees of
plausibility.
The sense-datum theory, then, is sufficient for generating philosophical
skepticism, but not necessary. And we can see immediately another advantage of the
general characterization of Humean or philosophical skepticism. This characterization
explains why differences in the assumptions about what counts as a directly perceivable
or sense-datum property are irrelevant where philosophical skepticism is concerned. For
it is a characteristic of all the variations of the sense-datum theory (that dont define
external objects in terms of sense-data) that what are perceived (given etc.) directly are
the sense-data, however rich they may be conceived as being. Thus it is a characteristic of
all such theories that we perceive external objects only in virtue of our perception of
(access to) sense-datahence indirectly. Of course, some theories of perception are more
plausible than others. And there is something perennially appealing about a theory like
Berkeleys according to which the properties given directly in (visual) experience are
identical (or very tightly connected) to the properties of the image on the subjects retina.
(This appeal is, in my view, completely spurious. I shall have more to say about this
below.)
For an example of Humean skepticism that does not involve or presuppose the
sense-datum theory, consider our access to the past. It is natural to think that what we are
given or perceive directly is in the present and that things in the past are unperceivable
(directly) in principle, or as a matter of conceptual necessity. Now consider a theory of
perception according to which we have direct perceptual access to ordinary objects and to
many ordinary properties (properties of external objects, not sense-data, etc.). In fact, as
we saw above in the analogous case of outside-a-room scepticism, we have to make
this assumption if we are to have unproblematic access (in principle) to things in the
counter-domain (the present). In such a case, and in contrast to the case of house arrest,
the Humean argument does apply and does, apparently, go through. There is a gap
between what we can in principle be given in perception and any belief about the past.
And because in principle we could never be given anything about the past directly, the
gap is one that no justified a posteriori inference could bridge. For anything to which we
could appeal in support of a hypothesis about the connection between the past and present
will simply give us more information about the present. Thus, unless our conception of
the past is a form of idealism analogous to phenomenalism regarding the external world,
skepticism follows. And, as we have seen, pushed to the limit, the skepticism in question
is meaning skepticism.
For another example, consider the problem of other minds. Imagine that we have
unproblematic access to all of the kinds of external objects that we normally think we do
(objects about which we think we can, therefore, at least in principle, form justified
beliefs), except other minds. The assumption that one cannot, on conceptual grounds,
experience in the relevant sense anothers painthat anything that could be experienced
PSYCHE 2007: VOLUME 13 ISSUE 1

12

PSYCHE: http://psyche.cs.monash.edu.au/

from the inside as pain would be ones ownis very widely accepted. And the idea that
one could make an analogical inference to a justified belief about what another person is
feeling on the basis of ones knowledge of what one feels oneself in similar
circumstances is not generally considered convincing. It is, at least, unclear what could
ground such an inference since one case seems to provide a weak basis for an inductive
argument.
If this were the strongest version of the problem of other minds, of course, it
would not support Humean skepticism. Even very weak inductive support for a
proposition is some relevant evidence. And what characterizes Humean skepticism is that
nothing could count as evidence for or against any hypothesis about conditions in the
relevant domain. But this way of putting the problem with the argument from analogy is
inadequate. For our so-called knowledge of our own mental states is not knowledge of
something that could in principle be had by anyone else. This is because of the
assumption, made by proponents of the argument, that our knowledge of our own mental
states is nothing more than what we have in virtue of experiencing them. Alternatively we
can say, for example, that all that we can mean in using pain is what we can mean in
virtue of knowing when we have it. But if this is the case, then it is a matter of conceptual
necessity that there is nothing we could mean in ascribing a pain to someone else. We
have no conception of pain as something independent of what it is like to have it, hence
nothing independent of its occurring in us. If this is the case, the inductive inference is
not simply weak, but meaningless, as is any analogical inference or inference to the best
explanation. Thus we have Humean skepticism about other minds. And if so, we have
another example of the kind we wanteda Humean skeptical argument that has no
connection with any commitment to a sense-datum theory. Similar arguments could be
made for a wide range of other domains, including causation, the future, modalities,
meaning, and, on some conceptions of what is perceivable, theoretical entities.
In each case what generates Humean skepticism is not the sense-datum theory, but
the unobservability in principle of the elements in the domain (assuming the neutrality
condition holds). Thus there is a sense in which philosophical skepticism is the result of
epistemic intermediaries if these are understood broadlyso broadly that some type of
thing counts as an epistemic intermediary between a subject and a domain just in case the
following holds: Elements of the domain are in principle not directly perceivable by the
subject, and that subject perceives those elements indirectly in virtue of perceiving (being
given, having access to) things of the type in question directly. But, as we have seen,
epistemic intermediaries understood so broadly come in many different forms.
Understood in this sense, arguably (and not implausibly), present events (some of them)
have this status for us relative to events in the past and future; actual states of affairs have
it relative to possibilities, conjunctive relations have it relative to causal relations, and
bodies have it relative to other minds. Thus abolishing sense-data, the given, perceptual
experience, and the like is virtually no use in helping us understand and deal with
philosophical skepticism.
Of course it might be suggested that if we deal with philosophical skepticism
about the external world, we can simply learn to live with everything else. But this is a
mistake. For we dont really get the external world simply by eschewing the given. What
we get is the present moment shorn of all its modal characteristics, causal relations and
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

13

PSYCHE: http://psyche.cs.monash.edu.au/

potentialities, meaning of any kind, and other minds. And recall that where philosophical
(Humean) skepticism is concerned, the epistemic deficit is not merely a matter of
certainty. Rather it is the lack of any rational justification for preferring any hypothesis
about the problematic domain over any other. Recall, moreover, that philosophical or
Humean skepticism has its immediate counterpart in meaning skepticism regarding the
terms in which we purport to describe the elements of the domain in question. Even, then,
if we completely repudiate any notion of a sensory given, we are left with a form of
skepticism with whose truth or unanswerability we could not even pretend to be
reconciled.
Thus far I have sketched the outlines of a transcendental argument that our having
a meaningful language (even one of such ordinary objects as tables and chairs) requires
our having perceptual experience of the kind Davidson evidently wants to deny. For we
need, it seems, perceptual experience as opposed to linguistic-descriptive content in order
to deal with the demonstrative versions of Freges problem. And, as we have seen,
Davidsons attempt to associate such experience with philosophical skepticism fails.
What we have instead is a clear association between such skepticism and epistemic
intermediaries in the broad sense defined above, including, as a special case, sense-data.
That is, there is a clear association between skepticism and things perceived (given)
directly in virtue of which the elements of a problematic domain are, as a matter of
conceptual necessity, perceived indirectly at best. What Davidson never provides is an
argument for the assumption that all sensory experience must have this character. In the
next section I shall provide a more detailed version of the transcendental argument, in the
course of which I shall argue that this assumption is false.

5. The Transcendental Argument


As we saw in the outline of the transcendental argument (Section 2), linguisticdescriptive content is not sufficient, either for the characterization of a meaningful
language, or for the specification of the contents (in terms of cognitive significance) of a
subjects beliefs. There must, in addition, be content in virtue of which the subjects
language is grounded subject to Freges constraint. More explicitly, there must be
connections between words and the world that are direct in the sense that they are
unmediated by linguistic-descriptive contentand which, nonetheless, provide the modes
of presentation that the satisfaction of Freges constraint requires. And this result, as we
have seen, stems from the impossibility of an infinite regress of linguistic descriptions,
the need to distinguish a meaningful language from a merely formal calculus, and the
requirement that we deal with the Frege problems in their demonstrative versions.
But to say that our access to external objects is (sometimes) direct in this sense
(i.e., unmediated by linguistic description) is still not enough. Such a requirement is fully
compatible with a theory like Humes or a sense-datum theory according to which our
words get their meanings through association with collections of, and patterns among,
sense-data. What rules out such theories is not the regress argument, the inadequacy of
formal inferential roles in grounding meaning, or the need for non-linguistic modes of
presentation in cases of demonstrative access to external objects. Such theories are ruled
out, as we have seen, by a different argumentthat because they preclude the direct

PSYCHE 2007: VOLUME 13 ISSUE 1

14

PSYCHE: http://psyche.cs.monash.edu.au/

perception of external objects in principle, they make the Humean skeptical argument and
its meaning-skeptical counterpart unanswerable.
It follows, then, that our access to external objects must, in at least some cases, be
direct in another sense. It must, as we have seen, be non-inferential. But, again, our
access must be direct in this sense while being perspectival. Equivalently, our access
must be direct, while doing full justice to all the relevant distinctions in content where our
intentional states are concerned and where content is understood as cognitive
significance. This means it must be direct while allowing us to do full justice to the
rationality of the subjects of intentional states and thereby satisfying Freges constraint.
And with the sense-datum solution ruled out, it is unclear what an account of such access
could look like.
To see what such access might amount to, consider two more definitions of
directness that, while ostensibly different, are, when properly understood, arguably
equivalent: (1) Non-descriptive access to some element of a domain in virtue of an
intentional state (i.e., the availability to the subject at the personal level of that element in
virtue of an intentional state) is direct just in case the subject has such access and does
not have it in virtue of having access to anything else. (2) Non-descriptive access to some
element of a domain in virtue of an intentional state is direct just in case if the object
doesnt exist, there is no complete content. We can see that these come to the same thing,
since if our access (to an external object, for example) is direct in sense (1), then there is
nothing (at the intentional and personal level) to which we have access and in virtue of
which we have access to the object. The object itself, we might say, is an ineliminable
part of the intentional content that provides or constitutes the subjects access to the
object (See Evans 1982, pp. 64-73). Thus, if the object doesnt exist, the intentional
content will necessarily be incomplete. And if the access is direct in sense (2) so that if
the object doesnt exist there is no complete sense, then there couldnt be an intentional
item to which we have access and which is such that in virtue of this access we have
access to the external object in question. For such an intentional item would complete an
intentional content, assuming it was complete in other respects, even if the external object
itself didnt exist.
Not only does it seem that there is a natural understanding according to which (1)
and (2) are equivalent, it is also natural to suppose that they are both equivalent to the
requirement that our access to the objects in question be potentially (or, in the case of
external objects at least sometimes actually) non-inferential. For if our access is direct in
the sense of being non-inferential, then there is nothing given to us in virtue of which the
object is givenif there were, then there would be room for an inference, and if so, one
would be required. And if there is nothing in virtue of which the object is given, then
there is no room for an inference and, arguably, no need for one.
But what does it mean for access to be direct but perspectival? Or, to put the point
in a different and, I think, more useful way, what kind of account could we give of such
access? We can see immediately, it seems, that it will not be a reductive account in any
ordinary sense. For if we try to characterize such access in causal terms (e.g., causal
terms involving physical or functional states, or even states of a mental substance) we fail
to capture one of its essential features given its defining role in satisfying Freges
constraintits accessibility to the subject to whose rationality we are committed to doing
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

15

PSYCHE: http://psyche.cs.monash.edu.au/

justice. The account, then, will be in terms of states that are both subjective and
intentional. But, again, what could such an account look like?
To begin to answer this question, I want to make a provisional assumption. The
access we wantaccess that is direct but perspectivalis a matter of our perceptual
states. This assumption is natural given what is apparently indisputable: that language is
grounded in our demonstrative access to objects. For in our example of the demonstrative
version of Freges problem , the Evans ship example, the difference in the two modes of
presentation of the same ship seemed clearly perceptual in nature. But to say that the
assumption is provisional is not to say simply that it is one for which, as yet, no argument
has been given. Rather it is to say that the assumption is, in this context, a partial truth,
and that in a broader context, and in something closer to the final analysis, it will appear
in a different light.
Of course, to say that our access to the ship in the Evans case is a matter of our
two different perceptual experiences (one of the bow, one of the stern) just pushes the
problem one step back. What account can we give of the nature of the perceptual
experiences themselves? The two (apparently) different perceptual contents cannot be
understood solely in terms of linguistic-descriptive content, since this would not take us
outside the circle of word-to-word connections and hence couldnt ground language. Nor
could the difference between the perceptions be a matter of different causal chains, since
(given that these are not necessarily, and not generally, available to the subject) this
wouldnt serve to ground language subject to Freges constraint. This seems to leave as
the only other alternative a difference in the apparent visual properties available to the
subject in the two experiences. But we have, as yet, no way of understanding this,
compatible with the satisfaction of Freges constraint, except in terms of sense-data. And,
as we have seen, far from grounding language, the sense-datum theory of perception
leads to the most radical form of meaning skepticism.
To see the existence of an alternative, consider the things that the two different
perceptual experiences allow us to doin particular the basic actions they allow us to
perform. The visual perception of the bow, for example, allows one to point to it, trace its
outline, move toward or away from it, move in relation to it so as to get a better viewto
produce a better line of sight, or to obscure the line of sight (perhaps so one cannot be
seen by a subject located on the bow), etc. We can think, then, of a very large package of
basic action possibilities associated with the view of the bow. And this will be a different
package or set from the one associated with the view of the stern. Where the project of
grounding language is concerned, we could express the underlying intuition in terms of
the familiar slogan (though one which I believe has elicited more lip service than genuine
appreciation) Meaning is use.
In spite of the intuitive plausibility of the move to basic-action possibilities in the
context of grounding language, however, it seems anything but intuitive where perceptual
experience is concerned. (There is no slogan Perception is use, nor does it answer to
any currently felt need.) And to reduce perception to agency, or (worse) the possibility of
agency, seems to leave out precisely what is essential to perceptionthat it is experience,
that there is something it is like to have it. What I am suggesting, however, is neither a
reduction of the perceptual to the agential, nor an analysis or an explanation of the former
in terms of the latter. As will become clear, this analysis of direct access could just as
PSYCHE 2007: VOLUME 13 ISSUE 1

16

PSYCHE: http://psyche.cs.monash.edu.au/

easily be called a perceptual analysis of agency as an agential analysis of perception. That


is to say, the analysis treats neither perception nor agency as more primitive than the
other. At the moment, however, I want to focus on perception and make three points. The
first is a point about the intimacy of the connection between perception and action. It
involves an argument designed to show that, contrary to what we might suppose, there is
no logical or conceptual gap between our conscious perceptual awareness and our
corresponding agential capacities. Needless to say, the argument is designed to make the
agential approach to direct perception less unintuitive. The second point is made with the
same aim, but proceeds from the opposite side. That is, I argue briefly for the inadequacy
of the sense-datum conception of visual perception (the source of virtually everyones
pre-theoretical intuitions, regardless of their theoretical commitments). And I do so on
phenomenological grounds. The intention is to counter the claim that the treatment of
perception Im putting forward is motivated by exclusively theoretical considerations and
ignores our ordinary intuitions. The third point is a more explicit explanation of the
theoretical payoff of this treatment of direct perception.
On the first point, consider an argument I have used elsewherethat the notion of
perfect blindsight is incoherent (White 2004b, pp. 215-216). As is well known, subjects
with certain kinds of brain lesions report blind areas in their visual fields. When pressed
to answer questions about visual features of external images falling in such areas,
however, their rates of success are considerably better than chance. Such subjects, then,
will have many of the agential capacities cited in connection with Evans ship. There is
no reason to suppose that they are limited to describing the various kinds of shapes and
motions they detect. They could (almost certainly) point in the appropriate direction,
trace a rough trajectory of a movement, trace the rough shape of a stationary image, and
so forth. In spite of this, taking their testimony at face value, they have no visual
experience. Thus it might be thought that this is exactly the example to show the
conceptual independence of perceptual experience and agential capacities. But now
consider a thought experiment. Imagine that the agential capacities of a blindsight subject
get better and better until they match those of a normal subject. Imagine, that is, that this
so-called blindsight subject can do all the same things, answer all the same questions, and
do so with the same confidence and spontaneity as a normal subject. Suppose in
particular that whereas the guesses of actual blindsight subjects have to be elicited by
others, beliefs about the part of the world hidden in the blind area occur
spontaneously and in the same profusion as they do for the normal subject. And suppose
that far from seeming alien, or as though they come from nowhere, these beliefs are as
well integrated with the subjects practical capacities as those of any subject with normal
sight. Suppose, in fact, that the occurrence of such beliefs is not different in any relevant
way for the parts of the world inside and outside the blind area. Can we still suppose
there is any sense to the idea that regarding that area the subject is blind?
Imagine looking at your computer keyboard with your eyes closed. Certainly you
can answer some of the same questions about its visual appearance that you could answer
with them open, but many of them you could not. For example, where are the brightest
highlights? Where are the darkest shadows? Where is the Intel logo in relation to the
Windows logo? Where is it in relation to the P key? Now imagine that you get better
and better at answering such questions. And bear in mind: This is not simply a matter of
certain words popping into your head. You have the same demonstrative visual abilities,
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

17

PSYCHE: http://psyche.cs.monash.edu.au/

and so the same demonstrative access, that you have with your eyes open. You can see
complicated visual patterns at a glance that you would be hard pressed to describe in
words, but they are ones you could match with confidence to congruent patterns flashed
on a screen. You can trace the outlines of such patterns, you know how to move your
head to get a better view, and you know when you have moved it too far to see the whole
keyboard. You can obscure any feature of the keyboard by holding your finger at an
appropriate location between it and your eye. And instead of describing the relation
between the Intel logo and the P key, you can use your finger to trace a connecting path.
Moreover, when performed, these are full-blooded actions. They are given in experience
as doable before being performed, they are performed with confidence, and they are
given as successful when the performance is complete. Can you really imagine all this
and not imagine that your experience is phenomenologically equivalent to seeing?
My second point starts from the side of visual experience and is offered to help
defuse the idea that the proposed treatment leaves something out. The point is a critique
of the sense-datum view, but on phenomenological, not theoretical, grounds. In fact, I
shall argue for what I have called elsewhere an inflationary/deflationary phenomenology
of experience (White 2004b; White 2004a pp. 281-282). By inflationary and
deflationary I mean inflationary and deflationary relative to a strict sense-datum theory.
And by this I mean one according to which, for visual perception, the subjective visual
field is modeled on the retinal image. On such a theory, the only sense-datum properties
will be properties that have obvious retinal counterpartsshape, color, and relative size.
An inflationary/deflationary phenomenology, then, is one according to which what we are
given directly in experience (and I shall be concerned primarily with visual experience) is
both richer (in some respects) and more impoverished (in others) than such a sense-datum
theory would allow.
Consider deflation. Here there is a wide range of examples suggesting that our
most direct access is not to sense-datum properties because, the evidence suggests, we
have no such access at all. Most people, for example, cannot approximate the apparent
shape of their car windshield (as viewed from the drivers seat) using four straight lines.
Of course, in the case of such relatively simple shapes, there is a tendency to think that
the problem must be a matter of memory. But even when allowed to look, most people
cannot, while standing, estimate with any accuracy the ratio of the apparent length of
their hand held four inches from their eyes to the apparent width of their foot. Moreover,
there is a clear argument that visual sense-data could not have the character that the
theory attributes to them. For the visual field as it is given to us seems to have the
following properties. On the one hand, it seems to encompass a visual angle of roughly a
hundred fifty degrees, and, on the other, shapes of objects seem undistorted, even at the
peripheries. In addition, objects seem to maintain their shapes, even as we turn our heads.
But images cannot have these three properties at the same time. This is apparent if one
watches the shapes of objects outside the middle of the frame in any film in which a
wide-angled lens camera pans across the scene. And this is not an empirical point about
current lens technology. Rather, it is a point about the constraints that can and cannot be
satisfied simultaneously in mapping three-dimensional space onto a planeand one that
was well known to Renaissance theorists (White 2004b p. 211).

PSYCHE 2007: VOLUME 13 ISSUE 1

18

PSYCHE: http://psyche.cs.monash.edu.au/

On the side of inflation, notice, as Berkeley did, that even depth could not be a sensedatum property, strictly speaking. Richer examples include the fact that a staged fight will
look different when one thinks it is real and when one realizes it is not. (Imagine that you
see the same set of physical movements twice, once under the first impression and once
under the second.) Similarly, a park, viewed from the same angle, will look different
when you are completely new to the city in which it is located and when the city and the
parks orientation in it have become completely familiar. The most compelling examples
for many, however, are those involving the expressive features of other people. We can
be, and normally are, given anothers distress directly, and not on the basis of an
inference from the purely geometric features of the persons face or bodily posture (much
less their sense-datum counterparts). This example works as well as an example of
deflation, since if we are given the person as distressed, it is difficult to see how we could
be given the pure geometry of the relevant body parts. There would, it seems, be
incompatible determinations of the same part of the visual fieldsomething the sensedatum theorist cannot accommodate. In fact, if this is right, inflation entails some form of
deflation. If we have inflation, then what we are given directly are richer properties than
the strict sense-datum theory allows. But if this is the case, then we are not given some of
the sense-datum properties that the theory postulates, and we have (some degree of)
deflation. Moreover, we should not forget the important research traditions stemming
from the work of Michotte and J. J. Gibson (Michotte 1963; Gibson 1986). Empirical
studies deriving from the former involve the direct perception of the rich properties
constitutive of causation. Those following on the latter concern the direct perception of
humanly relevant functional properties of the environment (physical-geometric features
may be given directly as hiding places or escape routes, for example), as well as
intentional and mentalistic properties.
The third point is intended to clarify the theoretical payoff of the agential
approach, while anticipating somewhat the discussion of agency to follow. Recall the
general structure of the argument. I argued that language must be grounded subject to
Freges constraint and that this requires that some perceptual experience be both direct
and perspectival. This raises the question how any experience could be both. I therefore
introduced the agential approach to perception, which will make it possible to appeal to
the notion of know-how in beginning to answer this question. And shortly I shall argue
what I have already claimed in passingthat the agential analysis does not leave out
what experience is like because agency has the rich phenomenology I have just been
describing. Moreover, it will be shown that this is not merely a disputable intuition about
agency, but a requirement of our making sense of action. But then the following will have
been shown: We need perception to ground language subject to Freges constraint, we
need action to give us perception that is both direct and perspectival, and we need a rich
phenomenology (and, in particular, a rich phenomenology of action) to give us genuine
agency. And the conclusion that we need a rich phenomenology of action in order to have
a meaningful language was exactly the conclusion we wanted.
What, then, is it about action that entails that it must have a rich phenomenololgy
of the kind I have been gesturing toward? I shall give three arguments, none of which
appeals simply to our intuitions about what the phenomenology is like (of which there are
many). The first stems from the possibility of what I have called the passive subject
(White 2004a, p. 284; White 2004b pp. 201-202). Imagine a person who claims not to
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

19

PSYCHE: http://psyche.cs.monash.edu.au/

get the notion of agency. (He is, at the outset, capable of speaking, and so not
completely passive. But we can imagine him rapidly getting worse and eventually falling
silent.) In the initial phase, we can imagine that the passive subject is completely
competent and comfortable in using the language of objective science to explain and
predict events, and that it is only the specifically agential vocabulary (including perhaps
evaluative notions used in the specification of the goals of action) that raises a problem.
We can, however, suppose that he was once a normal subject and thus that he has, in one
sense, a perfectly good grasp of the use of agential terms. The passive subject knows
what generalizations it is appropriate to accept (e.g., One is responsible in the first
instance only for what one does) and even what is appropriate in particular instances. (If
someone is the causal source of spilled milk and wasnt pushed, tripped, etc., then it is
appropriate for that person to own up and say, I did it.) We could say, then, that the
passive subject still grasps the inferential role, but only the inferential role, of the agential
vocabulary. Alternatively, we could say that his knowledge is like an ideally wellinformed anthropologists knowledge of the demonological language of some group
when the anthropologist has no belief in demons. The passive subject finds, we might
say, nothing in the world that could justify or motivate the specifically agential
vocabulary.
What is it that we have and that the passive subject lacks? It will be obvious that
the example is so constructed that the answer that strongly suggests itself is something
perceptual. For what else could it be? First, we can suppose that the passive subjects
motor systems are all intact, indeed that this has been tested. We can imagine, for
example, that while the passive subject still had some capacity for spontaneous action, he
was made to think that someone was about to put poison in his intravenous bottle. And
we can suppose he jumped up without thinking to knock the poison away and then
immediately collapsed back into passivity. Second, recall that we are assuming that the
passive subject agrees with us about all the facts that can be characterized without a
specifically agential vocabulary and that he still retains the inferential role of the agential
expressions. And finally we can assume that there is no lack of motivation, that the
passive subject wants desperately to be active again. What is it, then, that the passive
subject is missing?
It seems that all the passive subject could lack, given that he retains the inferential
roles of all the terms in the specifically agential vocabulary, is something on the side of
perception and know-how. It must, it seems, be something at least closely related to what
Gibson describes as the perception of affordancesthe perception of things as affording
certain actions. As we have seen, according to Gibson things like chairs are not normally
given simply as solid objects with certain geometric configurations. Rather chairs are
given as things one can sit on (or stand on to change a light bulb, etc.), and, more
generally, things are given in their agentially relevant aspects. The passive subject, then,
lacks the rich phenomenology in virtue of which things are given as doable, situations are
given as opportunities, and so forth. And the point of the example is to suggest by
contrast what we have and thus the richness of the perceptual experience necessary for
agency.
Some may find the example of the passive subject difficult to imagine. But
consider the past. For us the past is dead with respect to actionwe have no idea what it
PSYCHE 2007: VOLUME 13 ISSUE 1

20

PSYCHE: http://psyche.cs.monash.edu.au/

would mean if someone were to suggest that we act toward the past. But why? Certainly
it seems there are no conceptual difficulties in the idea of backward causal chains. To
suppose that instances of backward causation exist simply means that where the direction
of time has been established by the preponderance of causal processes, there are local or
isolated exceptions that run in the opposite direction. What, then, would it be like to act
toward the past? Imagine that you can control the cursor on our computer screen with
your mind as a basic action (without touching the mouse, etc.). (This is not currently
science fiction.) Suppose the cursor is now in the middle of the screen, moving slowly to
the right on a horizontal path. Your are asked to make it the case that the cursor reached
its midpoint position on the screen from a point above the midpoint. You tune in to a
point in the past when the cursor was at the left edge of the screen, midway between the
top and bottom and moving slowly and horizontally to the right. And you guide the
cursor up, toward the top of the screen, and then down to the midpoint that it occupied
when the instruction was given.
What, though, would it be to be given the past in the right way? One would
have to be given some part of the past directly (not via present representations) if we
want it to be a basic action in virtue of which we bring about something in the past. But
vivid memories with which one is fully engaged, and in which one is given past events
and not present images of those events, are experiences in which one is given the past
directly. Ones being given the past in the right way, then, could be thought of as being
like thiswithout, however, ones being given the past events necessarily as things one
originally experienced oneself.
There is another way in which we might conceive of acting toward the past. We
might begin with a common representation in films of the experiences of psychics.
Typically the subject has a vision of how events went in the pasta vision of events that
he or she did not witness. Imagine having such visions. At first one doesnt know how to
interpret them. With experience, however, it comes about that the events are given as in
the past. How does this happen and what is it like? We can imagine it by analogy with the
theory-and-practice-laden experience of the trained scientist for whom the equipment is
no longer given as metal and wires, but as a particle accelerator. And, given that we have
the idea of a direct perception of some part of the past, there is an even simpler way in
which one might act toward the past. If the instruction is bring it about that the cursor,
now in the middle of the screen, is at a point above the middle at some time in the past,
one might focus on an appropriate time and put it therei.e., move it to that place at that
time. This would be no more mysterious than the fact that in moving the cursor as a basic
action aimed at the immediate future one can make it appear where one wants itor that
one can do so with ones arm if one is told to raise it.
With regard to the passive subject, then, we can say this. We can well understand
the phenomenology of the experience of the passive subject because we can understand it
by analogy with our own experience of the past. Unlike the passive subject, however, we
experience ourselves as agents over a broad range of cases. And because we do, we can
understand by analogy what it would be like to experience some aspects of the past as a
domain open to our purposive intervention.
If, as these arguments suggest, the example of the passive subject is intelligible,
then we have an argument that agency requires a very rich phenomenology of experience
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

21

PSYCHE: http://psyche.cs.monash.edu.au/

(rich relative to the strict sense-datum theory). And the argument supports what, in any
case, we know on experiential grounds. We see things and the world as affording us a
great many possible actions. And we do so by seeing things as tools or as means (the
rocks in the stream are a bridge), seeing things as opportunities, (the distance between
two cars is not a space of so many meters, but an opportunity to change lanes), and, as I
shall argue later, seeing things as desirable or valuable. This is on the inflationary side of
the inflationary/deflationary phenomenology. On the deflationary side, we may cease to
see things at all. The automobile we are driving is, like our own bodies, very often not
experienced; rather we experience the world through it. The space between cars that is an
opportunity to change lanes is an opportunity for us, given that we are driving a mid-size
sedan. But it might not be an opportunity (nor given as one) if we were driving a tractortrailer truck. And to repeatthese affordances must be given to us directly in experience.
Were objects given in perception in an agentially neutral way and then interpreted as
providing opportunities, say, as the result of a linguistic description, we would have no
explanation of what the passive subject lacks. Furthermore, we would have no
explanation of the meanings of the expressions in the alleged linguistic interpretations.
For such expressions would not be definable in non-agential terms. If they were, we
could tell the passive subject what he is missing, and we could simply be told, without
appeal to any experiential analogies, what we are missing regarding the possibility of
acting toward the past.
The second argument that agency requires a rich phenomenology of experience
stems from the apparent inadequacy of current attempts to characterize action.
Davidsons original suggestion that a piece of behavior was an action just in case it was
caused by an appropriate belief-desire pair is undermined by his well-known climber
case. In the example, one climber is suspended by a rope held by another, who is in
danger of losing his grip on the rock face. The latter suddenly thinks, and thus believes,
that if he drops his companion he can ensure his own survival. And certainly he has the
desire to survive himself. As the example is told, however, the recognition that he could
even think such a thing so shocks and unnerves him that he accidentally lets go of the
rope. Thus an appropriate belief and desire cause the behavior, but, by hypothesis, the
climbers dropping his companion is an accident.
The temptation is to require that that the appropriate belief and desire cause the
behavior in the right way, where this is to be specified in neurophysiological terms. But
this seems clearly wrong, giving, as it does, the neurophysiologist the last word about our
actions. For in so doing, it threatens the notion of action itself by undermining the
authority that as agents we normally suppose we have. Suppose we have what looks to
the subject, and to fully informed (non-expert) spectators, like a paradigm case where
action is concerned. The subject is acting not only on the basis of genuinely held desires,
but deep value commitmentscommitments that he or she would be prepared to defend
in detail. Furthermore, there are no hidden sources of interference of the sort that nonexperts could recognize as undermining agency. There are no mad scientists manipulating
the subjects brain, no rampant self-deception, and so forth. Is it really plausible to
suppose that all the non-expert judgments could be overturned on the basis of the details
of the causal chain that connects the belief-desire pair with the behavior? (The discovery
that the connection was merely accidental, would be a different story.) We normally
know which are our actions without knowing anything about the relevant
PSYCHE 2007: VOLUME 13 ISSUE 1

22

PSYCHE: http://psyche.cs.monash.edu.au/

neurophysiology, and, in fact, the Davidson climber example would ordinarily be


understood as just such a case.
This objection to an appeal to the neurophysiological details of the causal chains
connecting belief-desire pairs to behavioral outputs suggests another strategy for
characterizing action. Rather than making the characterization of action depend on
neurophysiological facts, we might make it depend on facts that are accessible to the
agent. In particular, we might require that the causal chains terminating in behavior go
through some appropriate piece of practical reasoning in order for the behavior to count
as action. But this seems open to a fundamental objection. Consider weak-willed action.
At a party, Smith sees the waiter approaching with a Martini tray. He recalls that he will
have a long day of important work before him in the morning and that he has had four
Martinis already. He reasons correctly that, all things considered, it would be best not to
take another drink. Nevertheless, when the tray arrives, he reaches out and takes his fifth
Martini.
We ask Smith what happened. Did he forget the salient points in his piece of
practical reasoning and reason again on the basis of different premisesfor example that
he had nothing to do the next day and that there was no reason to refrain from another
drink? Or did he entertain two logically incompatible beliefs at the time of the action
that, all things considered, it was best to abstain and that, all things considered, it was
best to have the Martini? Smith is indignant. He is not he assures us, despite his four
Martinis, lacking in either lucidity or any kind of reasoning ability. He had, he claims, all
the relevant beliefsabout the importance of his work the next day, about the debilitating
effects of a fifth Martini, etc.consciously and unwaveringly in mind as he reached for
the glass. Nor, he claims, was he guilty of any irrationality of the sort involved in
consciously holding two logically incompatible beliefs at the same time about what was
best all things considered. The action was, he claims, merely an instance of practical
irrationalitya weak-willed action, pure and simple.
Weakness of the will thus understood seems entirely possibleindeed, to some
of us, all too obviously actual. But now the question arises: What makes Smiths taking
the Martini an action? And the difficulty here is acute because, far from being supported
by an appropriate piece of practical reasoning, this action flies in the face of the relevant
practical argument. And again, mere causation by an appropriate desire, even ones
strongest relevant desire, cannot by itself make the movement of the arm an action. Such
causation is compatible with the arm moving, like Dr. Strangeloves, as though with a
will of its own. And certainly such alienated movement is incompatible with action.
Indeed, we can imagine an example of an alienated subject as an extension of the
passive subject example. Suppose that while action remains incomprehensible to the
passive subject, his beliefs and desires come to cause appropriate bodily movements. This
is compatible, as we have seen, with his being completely alienated from the
movementsnot because he disvalues their consequences, but because they are given as
though they were happenings on a par with knee jerks and involuntary twitches. Even if
such movements happened (accidentally as far as the passive subject is concerned) to
satisfy his strongest desires, he could experience such events while still claiming
sincerely not to find action intelligible. In such a case he would refuse to acknowledge
such events as actions, and we, it seems, would have to concur. Furthermore, even adding
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

23

PSYCHE: http://psyche.cs.monash.edu.au/

privileged knowledge, including privileged knowledge in advance, of the movements in


question would not be enough. This is clear, first of all, because such privileged
knowledge is still something the passive subject could have while still finding the concept
of action unintelligible. In addition, however, the passive subject would have special,
privileged knowledge of what bodily movements would occur only under their
descriptions as bodily movements, whereas the privileged perspective of the normal agent
would involve action descriptions. In fact, the normal agent might be almost as ignorant
of the movements described as movements or behavior as the passive subject is about
them described as actions.
What is missing, both for the alienated subject and for the weak-willed subject, as
was the case for the passive subject, is something perceptual. The argument for this claim
that was given in the case of the passive subject applies to the alienated subject, who, like
the original passive subject, fails to find the concept of action intelligible. Certainly the
perception of affordances or the direct perception of opportunities goes a long way
toward eliminating the most blatant forms of alienated behavior. And it helps to ground
the agential vocabulary for the passive and alienated subjects who understand its
inferential role, but, nonetheless, fail to grasp its meaning. But what the example of
weakness of the will suggests is that this is not enough. For what seems to be missing is
any sense of the teleological structure of action. This includes the fact that we aim at an
outcome as a goal, rather than being pushed toward it from behind, and that the agents
take on the object at which he or she aims rationalizes, rationally motivates, and makes
intelligible the agents aiming at it. In the case of weakness of the will, there is no
possibility of such a take being given in linguistic terms. For the agents considered
belief, linguistically expressed, is that the qualities of the Martini do not make aiming at it
rational, rationally motivated, or in the relevant sense intelligible. We have to suppose
that such a take that does rationally motivate and make intelligible the subjects action
will be given not in linguistic, but in perceptual terms. This again is an argument that
agency requires a rich phenomenology of perception.
That this requirement that the teleological structure necessary to rationalize
(rationally motivate, make intelligible) action is not completely captured by the
perception of affordances leads to the third argument. Suppose we ask how the agents
desires are given to him or her. We can see immediately that an objective representation,
say as a state of the brain, will not do. Represented in this way, there is no necessary
connection between and agents having (and being aware of having) desires and being
rationally motivated to satisfy them. Suppose, for example, that you were told by a
neurophysiologist that your strongest desire (not yet, but soon to be, manifest to you) was
for great wealth. Then, given that you were planning an academic career, a perfectly
rational response would be to try to get rid of it, rather than making any plans to increase
the likelihood of its satisfaction. And, it seems, no characterization that focuses on the
desire will reveal any internal connection between the desire and rational motivation.
Hence no such characterization could show how the presence of a desire could rationalize
a subjects action, and so no such characterization could reveal its teleological character.
What is necessary, it seems, is that we ask how the object of the desire is given to the
subject when it is given through the desirehow it is given when it is given as the
desire presents it. This, then, is what it seems we have to ask in order to answer the
question of how an object must be presented to the subject in order to be intelligible as
PSYCHE 2007: VOLUME 13 ISSUE 1

24

PSYCHE: http://psyche.cs.monash.edu.au/

the object of an action aimed at securing it. (And such action is, as we have seen, to be
contrasted with mere behavioreven behavior caused by an appropriate belief and
desirethat has the same effect)
The only answer, it seems, that captures the teleological character of action as it is
experienced by the subject is that the object or goal of the action is given as in some
sense valuable or desirable. What this means in most cases is that the object is given in
some specific way that entails its value or desirability. Such an object will be given, e.g.,
as beautiful, fascinating, delicious, captivating, virtuous, compelling, right, morally right,
cool, hot, perfect, alluring, exemplary, intoxicating etc. It will not, then, be given as
merely desired, or as desired by oneself. The experience of the object must have, to use
John Mackies term, to-be-pursuedness built into it (Mackie 1977, p. 40). The reason
is, as we have seen, that desires by themselves dont automatically or necessarily make
action aimed at their objects intelligible. Consider again the person who is told that his
strongest desire, which is as yet not manifest to him, is for great wealth. Suppose he is
aware that the desire can be removed safely, effectively, at minimal cost, and without
unwanted side effects. If, nonetheless, he immediately begins applying to business
schools, he cannot explain this action by saying that he has it on the best authority that
great wealth is what he most strongly desires. Given the alternatives, we want to know
what makes wealth desirableand we wont accept an explanation of the action until its
object is portrayed as having some desirability feature.
We are now in a position to see the upshot of this treatment of action for the
analysis of weakness of the will and the implications of that analysis for both parts of the
original transcendental argument: the argument for a phenomenology of experience and
the argument for a distinctive phenomenology of agency. In the case of weakness of the
will, the desirability feature is completely missing from the subjects linguistically
expressed thinking and reasoning. Therefore, the weak-willed action can only genuinely
be an action on the basis of some desirability feature given directly in the agents
perceptual experience. In the case of Smith, the Martini is given in perception as
manifesting some feature relevantly similar to those listeda fact that would normally be
elicited in a subtle and detailed description of the Smiths subjective experience.
The intelligibility of agency, then, presupposes a rich phenomenology of agency.
That is, it presupposes one that includes not simply the direct perception of affordances,
opportunities, action possibilities and the like, but also the direct perception of a wide
range of desirability features. And this perceptual treatment of action should defuse the
worry that what looked like an agential treatment of perception would leave out the facts
about what it is like to have the perceptual experience. Neither analysis is more basic than
the other, and, in fact, basic action and direct perception go hand-in-hand. Roughly, basic
actions require that their objects be given in a direct and rich perceptual experience, and
direct perceptions are those that are intimately tied to appropriate packages of basic
action possibilities. Furthermore, as we have seen, meaningful language presupposes the
possibility of grounding subject to Freges constraint. Since grounding involves a
connection between words and the world unmediated by further language, such
grounding requires perception. The perception required, however, is perception whose
content is specified neither in linguistic terms, nor in terms of sense-data. For the first

WHITE: THE TRANSCENDENTAL SIGNIFICANCE

25

PSYCHE: http://psyche.cs.monash.edu.au/

would generate an infinite regress and preclude grounding, and the second, as we have
seen, leads to epistemological skepticism and its meaning-skeptical counterpart.
We need, then, access to the world that is direct in the sense that it is unmediated
by any mode of presentation that can be fully characterized independently of the object
presented. In the absence of such access, the Humean argument for skepticism goes
through. This access, however, must, nonetheless, be perspectival. And the agential
treatment of perception seems to have just these features. Different modes of presentation
involve different action possibilities, so the differences are clearly intentional and
perspectival differences. But recall that the phenomenology of agency and perception is
both inflationary and deflationary relative to the strict sense-datum theory. Often, when
we act directly on an object, there is no individuating characterization independent of the
object itself. Thus had the object not existed, there would have been no complete
thoughts or intentional contents among or associated with the relevant statesi.e., the
intentional states in which the object is given directly.
Consider again, for example, that we may see someones distress in the persons
face without having access to any features of the face that would be sufficient to
distinguish distress from an enormous range of other possible states and emotions. Nor
need we have access to any features that would explain (at the personal level) why we
immediately respond to distress rather than to any of those other possibilities. If the
person in question isnt really distressed, there is no way of capturing, either for an
observer, or for the subject himself or herself, the content of the experience. I was
having a perceptual experience in which it seemed that so-and-so was distressed wont
do, because we can give no account of what the experience of the persons seeming to be
distressed was like.
Here is another example. Imagine that in the Matrix Neo has an experience that he
would describe as seeing a woman in red. Suppose that he says in the course of it, That
woman is wearing red. Then imagine that he comes out of the Matrix and realizes that
there was no such woman. Can he fully characterize the experience or the content of the
experience he had inside the matrix from his new perspective outside? It seems not.
Certainly he could say It seemed to me that I was seeing a woman in red. But a woman
in red clearly doesnt capture the singularity and particularity of the experience, which
is, we may assume, irreducibly demonstrative (See Austin 1990). Also, one could have
the demonstrative experience that Neo has in the Matrix, without believing the content It
seems to me that I am seeing a woman in red. The reason is that one could have the
perceptual experience without having the conceptual sophistication required to express a
content that is neutral as between a real seeing and an apparent seeing. In fact, it seems
that one could have the perceptual experience without having the conceptual
sophistication to express a content about an apparent seeingindeed, one might not need
anything like a full-blown concept of perception at all.
As a final example of the directness and transparency of most perceptual
experience, imagine that you are operating a robotic device in an alien environment
(under water, inside a volcano, on another planet, etc.). Images on your screen are not
photographic, but are computer generated on the basis of information from light, heat,
chemical analyses, and so forth, and look nothing like images of natural features with
which you are familiar. You learn to manoeuvre the robot in order to complete the
PSYCHE 2007: VOLUME 13 ISSUE 1

26

PSYCHE: http://psyche.cs.monash.edu.au/

assigned tasks, while avoiding the most serious risks of damage to the machine itself, and
initially you do so by memorizing associations between two-dimensional images on the
screen and real features of the robots environment. Quickly, however, you begin to
recognize the real environmental features directly, without conscious awareness of the 2D images or any inferences based upon them. If you are then asked to describe what you
are doing, you will say that you are traversing a certain area, avoiding certain obstacles,
monitoring the levels of certain dangerous chemicals, and so forth. And if it turns out that
there is no such environment, you will be at a loss to provide an adequate characterization
of your experience. (Compare: How long was the parking space the last time you parked?
Just long enough to accommodate the car. Now imagine that what you took to be your
driving experience was one long nightmareperhaps not that difficult in Boston.) It
seems, then, that in the absence of the objects of such direct perceptions, there is no way
to capture the contents of such experiences, where contents are (as usual) individuated in
terms of cognitive significance.
Such contents may be called de re. And understanding them in terms of an
inflationary/deflationary phenomenology, we can see how contents could be both de re or
direct and perspectival. There may be no way of characterizing the contents of our
experiences when we operate in the alien environment that is independent of the
existence of the environment itself. But that is not to say that that content is adequately
captured by a singular proposition, where such a proposition is represented, if we focus
on the subject place alone, as an ordered pair of an object and a property that we ascribe
to it. The reason is that in operating in the alien environment, one may see the same
obstacle from two perspectives (once directly and once in a mirror, say) and fail to
recognize it as the same. (This is the demonstrative version of Freges problem again.)
Moreover, what we might call our perceptual-agential perspectives and capacities
understood in terms of the inflationary/deflationary phenomenology have exactly the
combination of features in question. They are direct or de re, and they are perspectival.
Hence they are the appropriate intentional entities to ground our capacity for meaningful
language, subject, as always, to Freges constraint. Such a capacity for meaningful
language presupposes them, but such perspectives and capacities per se presuppose no
such linguistic capacity. (Of course, since on this view much or most of our actual
perception will be theory laden, many of our actual perceptual-agential capacities will
have just such a linguistic presupposition.) Let us refer to this as the primitiveness of de
re thought.
We are now in a position to see how the transcendental argument is concluded. If
the meaningfulness of our language presupposes such de re thoughts or attitudes, then
there is no possibility of our stepping back from all such attitudes simultaneouslyof our
putting them all into doubt simultaneously and entertaining the possibility that they are
the illusory products of a coherent dream or an evil demon. For there is no place to step
back to, no way to describe our experience in noncommittal, external-world-neutral
terms. If we are to have a meaningful language, we must be talking about such things as
tables and chairs, even to characterize that subjective experience necessary to make our
words something more than elements in a formal calculus. And the generalized Quinean
criterion of ontological commitment that says, roughly, that we are committed to the
existence of those things about which we must talk in order to say the things we are
committed to saying entails that we cannot be neutral about all such external objects
WHITE: THE TRANSCENDENTAL SIGNIFICANCE

27

PSYCHE: http://psyche.cs.monash.edu.au/

simultaneously. Thus we have our transcendental argument against Humean skepticism


about the external world. And the rich phenomenology of agencymore generally the
inflationary/deflationary phenomenogyplays an ineliminable role.

6. Conclusion.
The standard objection to transcendental argumentsthat even if they work they show at
most that we must use certain concepts, or that if we use some we must use othersis
widely assumed to be conclusive. Transcendental arguments, it is supposed, though they
might tell us something important about ourselves, could never tell us anything about the
external world (Stroud 1999). But whether or not this objection is effective against
transcendental arguments such as those of Kant, Strawson, or Davidson (Strawson 1959;
Davidson 2001), it is not one that can be made against the argument I have given. For the
objection itself presupposes a division between, on the one hand, a subjective domain that
is external-world neutral (a domain that includes our uses of words and concepts), and,
on the other, the external world. And this is precisely the division that the transcendental
argumentin particular the conclusion that the primitiveness of de re thought is
presupposed by the meaningfulness of our languageshows to be incoherent. Moreover,
this demonstration of the incoherence of the domain-counter-domain division that the
skeptic needs does exactly what transcendental arguments have always purported to do. It
shows that something presupposed by the skeptics argument undermines the possibility
of his or her employing a meaningful language.
But how does this show that we arent, for example, brains in vats or subjects in
the Matrix? The answer is that it doesnt. It forces us instead to reconceptualize the nature
of such skeptical hypotheses. Or, better, it forces us to recognize that the only meaningful
hypotheses in the offing are the real ones. That is, the meaningful hypotheses are the ones
for which we could have real empirical evidence and the ones that, for lack of such
evidence, we could be justified in giving very little credence. Can we really make sense,
though, of the idea of empirical evidence for, say, a virtual reality hypothesis? In Vanilla
Sky the character played by Tom Cruise is approached by a stranger, who displays
uncanny knowledge and unexplained powersknowledge to which no one besides the
Cruise character could have access and powers that would be supernatural on the
hypothesis that his experience was of an ordinary external world. This is surely some
evidence against the ordinary world hypothesis, and we can easily imagine further
evidence (forthcoming in the film) that would make the virtual reality hypothesis more
attractive than a radical revision of the laws of physics.
It follows from the transcendental argument that Humean skepticism can be ruled
out on philosophical or a priori grounds as presupposing a false philosophical theory of
perception. The Humean argument for skepticism about the external world does not, as
we have seen, presuppose a sense-datum theory. But it does presuppose that we have
access to a domain, or the possibility of a stance, that is external-world neutral and within
which we can understand the skeptics argument. And if our having a meaningful
language means we must in fact have direct perception of the worldhave perceptual
states of a kind that if their (external-world) objects dont exist there is no complete
intentional contentthen no such domain exists. On the account derived from the
transcendental argument, our experience involves commitments that have taken us all the
PSYCHE 2007: VOLUME 13 ISSUE 1

28

PSYCHE: http://psyche.cs.monash.edu.au/

way out to the world already. And the fact that some such experience is presupposed by
the meaningfulness of our language implies that there is no stepping back as far as the
skeptics argument requires.
There is a sense, then, in which the skeptics hypotheses are incoherent. But there
is no mystery as to why it seems that the skeptic is saying something perfectly
intelligible. First, many philosophical theses that are demonstrably false on purely
conceptual grounds are, nonetheless, intelligible in any ordinary sense. And, indeed,
many such theses have a perennial appeal. (Example: all actions are aimed at the
satisfaction, pleasure, or happiness of the agent.)
Second, virtually every skeptical scenario understood as making a philosophical
or Humean point has a corresponding real scenario expressed in exactly the same words.
Imagine that you are about to enter the Library when one of your students runs up, says,
Professor, you may be under the influence of an hallucinogenic drug, and rushes off
without explanation. We can easily hear this remark in two ways. (1) The student is
simply pursuing, outside of class, the thread of an earlier philosophical discussion with
somewhat greater than average enthusiasm. (2) He has just come from a meeting with the
Dean in which his roommate has been accused of spiking the coffee in the Faculty Club.
The possibility of such real-world interpretations of standard skeptical scenarios suggests
again that we could, as I have argued, be brains in vats or virtual reality subjects (though
those who would consider seriously whether their beliefs leave them out of touch with the
world would be better advised to read Chomsky on the media than Hume on perception).
If so, the fact that we seem to understand the skeptic poses no problem for the view I am
maintaining.
Finally, that anti-skeptical strategies based on claims about meaning count as
senseless things we seem to understand is tied closely to another fact about themthat
they have a long history of having proved too much. In failing to make the distinction
between philosophical or transcendental skepticism and real skepticism, even such
transcendental arguments as Davidsons that most of our beliefs must be true (Davidson
2001) have seemed to rule out real possibilities. And they have seemed to do so merely
through reflection on the internal character of our experience. Thus transcendental
arguments have seemed to divide between those that are too ambitious to be believed and
those thatsince they tell us nothing substantive and non-psychologicalare (if the truth
be told) too modest to be very interesting. The distinction between philosophical and real
skepticism, however, reveals a third possibility: that transcendental arguments might
defeat skeptical hypotheses understood as philosophical, while remaining silent on their
phonological and orthographic counterparts understood as real. Such arguments need not
be thought to do the impossibleto give us substantive, non-psychological conclusions
on the basis of the internal character of our experience alone. Nor, however, are they so
modest as to tell us only that if in fact we think some things, we must think others. Such
arguments rule out as illegitimate what we might otherwise be tempted to say: Of course
we have commitments regarding the external worldbut thats just us! (I would not
want to be accused of sloganeering, but a useful mnemonic aid might be transcendental
arguments against transcendental skepticism, empirical arguments against real
skepticism.)

WHITE: THE TRANSCENDENTAL SIGNIFICANCE

29

PSYCHE: http://psyche.cs.monash.edu.au/

This understanding of transcendental arguments is, as we have seen, grounded in the


thesis of the primacy of de re thought, which is, in turn, grounded in the
inflationary/deflationary phenomenology. And this understanding is relevant to the
standard objection to coherentismthat a perfectly coherent set of beliefs and
experiences might amount to nothing more than a coherent dream (alternatively, that such
beliefs and experiences could float free of any connection with reality, or that nothing
could justify the assumption that coherence and truth are connected). Another way to put
the point is to say that coherentist theories involve an infinite regress. Something is
justified by its coherence with something else only if the latter is justified. But if all we
can point to in the case of any belief is that it coheres with some others, then it seems that
we could never regard any belief as justified at all.
Again, however, these arguments presuppose exactly the picture that I have
argued againstthat of a domain of belief and experience that involves no external-world
commitments. If we suppose, as I have argued we should, that as a matter of conceptual
necessity we cannot ever avoid such commitments, the situation changes. We are not, as
the objection suggests, trying to justify some beliefs in terms of their coherence with
beliefs that themselves are equally open to question. Rather, we are trying to justify those
beliefs whose truth and justification are open to doubt by reference to their coherence
with beliefs that, in context in question, are not up for grabsbeliefs that represent, at
least in that context, our commitments.
There is, then, in this answer to the standard objection to coherentism, a
foundationalist element. But it does not yield the absolute foundationalism that sensedatum theories have always been thought to support. It is rather what we might call
context-relative foundationalism. There are no empirical propositions that are immune to
either doubt or scrutiny. But such scrutiny, involving as it must the use of a meaningful
language, can only take place against a background of beliefs that are fixedand beliefs
that, in virtue of the capacities we must have for direct perception and basic action, take
us all the way to the world. Such beliefs thereby give us what we must have as well:
external-world commitments. And since such direct perception and basic action
presuppose a rich and distinctive phenomenology of action and perception, I conclude
that such a phenomenology has the transcendental significance I have claimed.4

References
Austin, D. (1990). Whats the Meaning of This?: A Puzzle about Demonstrative Belief.
Cornell University Press.
Davidson, D. (1973). Radical Interpretation. Dialectica 27, 313-28.
Davidson, D. (1983). A Coherence Theory of Truth and Knowledge. In D. Henrich (ed.),
Kant oder Hegel. Klett-Cotta, Stuttgart. Reprinted in D. Davidson, Subjective,
Intersubjective, Objective. Clarendon Press, 2001.
Evans, G. (1982). The Varieties of Reference. Oxford University Press.
Gibson, J. (1986). The Ecological Approach to Visual Perception. Laurence Erlbaum.

PSYCHE 2007: VOLUME 13 ISSUE 1

30

PSYCHE: http://psyche.cs.monash.edu.au/

Hume, D. (1968). A Treatise of Human Nature. Clarendon Press.


Hume, D. (2001). An Enquiry concerning Human Understanding. Open Court, Chicago.
Lewis, D. (1974). Radical Interpretation. Synthese 23, 331-44.
Mackie, J. (1977). Ethics: Inventing Right and Wrong. Penguin.
Michotte, A. (1963). The Perception of Causality. Basic Books.
Schiffer, S. (1978). The Basis of Reference. Erkenntnis 13, 171-206.
Searle, J. (1983). The Intentionality of Perception. In J. Searle, Intentionality: An Essay in
the Philosophy of Mind. Cambridge University Press.
Strawson, P. (1959). Individuals: An Essay in Descriptive Metaphysics. Doubleday.
Stroud, B. (1984). The Significance of Philosophical Skepticism. Clarendon Press.
Stroud, B. (1999). The Goal of Transcendental Arguments. In R. Stern (ed.),
Transcendental Arguments: Problems and Prospects. Clarendon Press.
White, S. (2004a). Skepticism, Deflation, and the Rediscovery of the Self. The Monist 87,
275-298.
White, S. (2004b). Subjectivity and the Agential Perspective. In M. De Caro and D.
Macarthur (eds.), Naturalism in Question. Harvard University Press.
White, S. (2007). Property Dualism, Phenomenal Concepts, and the Semantic Premise. In
T. Alter and S. Walter (eds.), Phenomenal Concepts and Phenomenal Knowledge. Oxford
University Press.
Williams, M. (1996). Unnatural Doubts: Epistemological Realism and the Basis of
Scepticism. Princeton University Press.

Notes
1. Those whom Michael Williams calls new Humeans are committed to our being able
to do so. Williams includes P. F. Strawson, Barry Stroud, Thomas Nagel, and, possibly,
Stanley Cavell in this category (Williams 1996, Ch. 1).
2. For suitably characterized sets of propositions corresponding to the domain and
counter-domain. Since I shall be arguing on other grounds that this example cannot
generate Humean skepticism, the complications in applying the independence condition
neednt concern us here.
3. For his doubts regarding the first assumption, see Stroud 1999, pp. 160-163. For his
reservations on the second score, see Stroud 1984, Chs. IV and V.
4. I am grateful to Massimo DellUtri for stimulating discussions of these issues and to
Ned Block and Susanna Siegel for comments and suggestions on an earlier draft.

WHITE: THE TRANSCENDENTAL SIGNIFICANCE

31

You might also like