Return
to Human Knowledge and Human Nature contents
11. THE PROBLEM OF THE
EXTERNAL WORLD
In this chapter I shall begin to consider the
significance of my proposed account of the core of empiricism for the problem
of scepticism.
Empiricist Epistemology
What would an empiricist epistemology now
look like, if my proposals concerning the essentials of empiricism were
accepted? One issue here, is whether a contemporary empiricist, who may accept
the existence of innate knowledge, and hence of one form of substantial a
priori knowledge, will have anything of interest to contribute to contemporary
debates in epistemology. At least a partial answer may be provided from amongst
the ideas defended in chapter 10. For the strictures against platonism, and
indeed against any claims to knowledge where we cannot give a naturalistic
account of the reliability of the belief-acquisition processes involved, remain
in force. The empiricist constraint on knowledge-claims may be expected to have
a distinctive impact on epistemology. Many claims to knowledge may be
disallowed because of it.
However,
the main issue I wish to consider is whether my contemporary empiricism is any
better placed to combat scepticism, thus allowing that we can have knowledge of
the physical world around us. Traditional empiricism has been closely allied to
scepticism, or at least to phenomenalism. While Locke thought that we could
have knowledge of a physical reality independent of our minds, his arguments
have generally been considered weak. Most subsequent empiricists, from Berkeley
and Hume through to Russell and Ayer, have doubted whether we could have
knowledge of a physical world which is independent of our minds. Instead, they
have embraced some form of phenomenalism, maintaining that our ordinary talk
about 'physical objects' may be reduced to descriptions of immediate
experience, or sense-data, and to descriptions of recurring patterns within
that experience. It is worth asking, then, whether the traditional empiricist
commitment to phenomenalism is in any way connected with opposition to nativism.
It may also be worth looking again at the pressures which have tended to push
empiricists towards scepticism.
However,
we need to be clear at the outset about the different levels involved in our
enquiry. For given a reliabilist conception of knowledge, of the sort defended
in chapter 5, we shall in fact have knowledge in any domain where we have
reliably caused true beliefs. So provided that there is in fact a physical
world, and that the processes which give rise to our beliefs about that world
(either perception, or natural selection, or some combination of the two) are
in fact generally reliable ones, then we do indeed have knowledge of the world.
But this will be of little help to us in combatting scepticism. For the
knowledge in question would be entirely first-order, whereas the knowledge that
we want is second-order. We want to know (second-order) whether we do indeed
have knowledge (first-order) of the physical world. If we cannot know this,
then we cannot answer the sceptic. Our task then, is to see whether a
contemporary empiricist may have sufficient reason to believe that we have
reliably caused true beliefs about the physical world.
Anti-Nativism and
Phenomenalism
Phenomenalism may plausibly be viewed as an
attempt to reconcile sceptical arguments with common-sense beliefs. Suppose we
accept that all knowledge must be grounded in the data of immediate experience.
Suppose also, that we agree that there can be no knowledge, on this basis, of
anything which is independent of experience. Then the phenomenalist should be
seen as attempting to accommodate this conclusion within common-sense, by
analysing the contents of ordinary beliefs into mere descriptions of patterns
within immediate experience itself. This project has proved remarkably difficult
to carry through to completion. Indeed, so far as I am aware, no remotely
plausible phenomenalist analysis of any proposition about physical objects has
ever been provided. But most empiricists have, in one way or another, been
committed to just such a programme of analysis.
The
reasoning behind the idea that all knowledge must be grounded in immediate
experience, which also seems to entail the possibility of a phenomenalist
analysis of our concepts, comes partly from the anti-nativism inherent in
traditional empiricism. As we saw briefly in chapter 1, if none of our
knowledge is innate, and none a priori, then all knowledge must be learned from
experience. And if no concepts are innate, and there are no innate
domain-specific mental structures embodying information about the world, then
the experience in question must be unconceptualised and unstructured. So from
this anti-nativist perspective, the basis of all knowledge would have to be a
set of experiences which are not structured in such a way as to contain
representations of anything outside of themselves - that is to say, in a set of
sense-data.
Anti-nativists
will of course allow that the mind is innately divided into different
faculties. They might also allow that the mind has an innate ability to
distinguish the differing attributes of our various sense-data, for example
recognising the difference between a sensation of red and a sensation of green,
and distinguishing a pain from a tickle.1 For these abilities do not
embody information about anything beyond our sense-data themselves. But for an
anti-nativist this would be the full extent of the foundation on which the
superstructure of our knowledge would have to rest.
However,
as soon as we accept that our perceptual faculties may be innately structured,
the picture changes dramatically. For in that case the basis of empirical
knowledge would not be sense-data, but perceptions of objects in space outside
us. I argued in chapter 6 that our visual faculty, in particular, is innately
structured so that in-coming information is automatically presented to us in
the form of representations of an array of physical objects in
three-dimensional space. It will be the beliefs arising out of these
representations which form the foundations of our empirical knowledge. It will
also be these representations which provide the materials for many of our
concepts, perhaps in the manner outlined in chapter 7. Notice that these
concepts will constitute modes of classifying physical objects, rather than
being abstractions out of, and modes of classifying, sense-data, as
phenomenalists would maintain. Indeed, viewed from this nativist perspective,
sense-data themselves become entirely marginalised, playing no significant role
within the structure of our belief-systems, except in so far as those beliefs
happen to be concerned explicitly with our subjective sensations.
It
is very likely true that our experience is innately determined as representing
a world of physical objects in space around us. Once this is accepted by contemporary
empiricists, even as a mere possibility, then they immediately lose any motive
for attempting a phenomenalist reduction of concepts. For there is then no
reason to think that all concepts must be constructed out of sense-data.
Indeed, if this form of nativism is true, then we have found an explanation of
why such reductions have proved so difficult to complete. But would this mean
that we should also have laid to rest the threat of scepticism? Unfortunately
not. For it is one thing to say that our perceptions innately represent to us a
world of physical objects outside us. It is quite another thing to say that
they truly represent such a world. It may be innately determined that
some of our beliefs, concerning perceived physical objects, constitute the
foundation of our cognitive system, while those beliefs themselves are false.
To say that it is beliefs about physical objects, rather than beliefs about
sense-data, which are foundational, is not to say that the whole system of our
beliefs might not be built upon sand.
To
see that this is so, notice that we can have apparent perceptions of objects in
space outside us which are illusory. I can seem to see a man lurking in the
bushes, while all that is really there is a play of shadows. Indeed, in cases
of hallucination, and perhaps also in dreams, we can be presented with apparent
perceptions of a three-dimensional reality which is wholly unreal. Moreover,
one only needs to recall Descartes' hypothesis of the omnipotent demon to
realise that it is at least conceptually possible that all of our experience
might be just as it is, and all of our foundational beliefs about physical
reality be just as they are, while there is in fact no physical world for those
experiences or beliefs to concern. While nativism might help us to avoid a
commitment to phenomenalism, it does not, on the face of it, help us to avoid
scepticism.
Varieties of Foundation
It might in any case be questioned whether
the sort of empiricist nativism sketched above is really coherent, or at any
rate consistent with foundationalism. For how can there be foundations to our
beliefs which are not completely certain? It might be claimed that to say that
a belief is foundational, is to say that it does not require support from any
other belief. And does this not entail that the belief in question must be
certain? For if a foundational belief P were uncertain, this would
surely mean that there must be some other possible belief Q which would lead us
to overturn P. But then, it may be claimed, P does require support after
all, namely from the belief that not Q; which contradicts the hypothesis that P
is foundational.
However,
this argument involves a confusion of levels. For what is foundational to a
cognitive system (a system of first-order beliefs) is not necessarily what is
foundational in epistemology, where our concern is to know what we know. We can
insist that our perceptual beliefs about the physical world constitute the
innately-determined foundation of our first-order system of beliefs, since
these beliefs are primitive, and are held independently (in general) of our
other beliefs. But at the same time we can allow that these beliefs, in being
uncertain, do not constitute an appropriate foundation for epistemology. If our
concern is to show that our perceptual beliefs about the world constitute
knowledge, then we clearly cannot simply take the truth of those beliefs for
granted.
It
is worth stressing again that the task of epistemology remains essentially the
same, whether we endorse justificationalism or reliabilism as a theory of
knowledge - namely, to justify our first-order claims to knowledge. In the case
of justificationalism this is so because knowledge itself requires
justification. But for a reliabilist it will also be true, given that our aim
is to know what we know (to achieve second-order knowledge). For the only
generally reliable method of forming second-order beliefs concerning which of
our first-order beliefs are both true and reliably caused, is to find
convincing reasons for believing that we have knowledge. (Another point is that
the very most that we could hope to obtain in any case, from attempts to combat
scepticism by providing reliabilist definitions of the key concepts, would be
to shift scepticism from one place to another. Even if scepticism about
knowledge, and about justified belief, could be blocked - by virtue of a
reliabilist account of these concepts - there would remain scepticism about internally
justified belief. So there would be no real gain. It is just as challenging
to be told that we have available to us no convincing reasons for belief in the
physical world, as to be told that we do not have knowledge of such a world.)
While many contemporary writers have thought that a shift to a reliabilist
conception of knowledge has great significance for epistemology, I disagree,
except as regards the possibility of innate knowledge. In particular, we still
face the problem of scepticism.
However, to concede that this is so,
need not imply a retreat towards phenomenalism. For what we may surely be
certain of is that we at least seem to be presented with an array of objects in
three-dimensional space. (Of course to be certain of this, we have to be
granted an adequate grasp of the concepts which we use to describe that
appearance. But then a similar concession must in any case be granted to
phenomenalism. It appears that any enquiry must take for granted that we
possess the concepts used in the expression of that enquiry.) There is then
nothing to force us to concede that the basis of epistemology must be
descriptions of sense-data. Rather, we may take as our starting-point that we
at least seem to perceive a physical world, and that many of our beliefs at
least purport to concern the distribution of objects within that world. Our
task is to discover whether we can provide sufficient reason for thinking that
most of these beliefs are true, perception itself being a reliable means of
acquiring beliefs.
What
form may such reasons take? Is our task to deduce the existence of the
physical world from the existence and nature of our perceptual beliefs? If so,
then the task is a hopeless one. For we have already conceded that it is
conceptually possible that our beliefs and experiences should be just as they
are, while there is in fact no physical reality corresponding to them. This
means that there can be no valid argument from the former (our perceptual
experience) to the latter (the world). For a valid argument is precisely one
for which it is impossible that the premisses should be true while the
conclusion is false. (And even if there could be such an argument, two problems
would remain: first, to give a natural explanation of how reason can possess
such power, which is the empiricist challenge; and secondly, to show why the
demon, or the fact that I am dreaming, may not be making me go wrong in my very
use of reason.) So if reasons for belief in the physical world had to consist
of deductive arguments, then we should be faced with a sceptical conclusion.
Taking for granted only that there appears to us to be a physical world, we
could not know ourselves to know that there is such a world.
However,
why should it be maintained that reasons for belief have to take the form of
deductively valid arguments? Why should reasons not include non-deductive
arguments, such as inference to the best explanation of a given phenomenon? It
is easy to understand why a traditional empiricist should place restrictions on
allowable reasons for belief. For such an empiricist is committed to the thesis
that all knowledge must be constructed on a two-fold basis only: the data of
immediate experience, together with analytic truths, including those embodied
in deductive arguments. (These latter they understand to be concerned merely
with relations between the ideas which we derive from experience.) Hence
deductive argument, and only deductive argument, may be taken for granted at
the outset as a method for providing reasons for belief.
With
nativism allowed to be a possibility, the situation takes on a different
aspect. For it may then be that non-deductive principles of belief-formation,
particularly inference to the best explanation, are an innately given part of
our cognitive apparatus. Indeed, we saw in chapter 7 that they very likely are.
In which case these principles will count as foundational, alongside our
perceptions and perceptual beliefs. But this is not sufficient to show, as yet,
that we may make free use of inference to the best explanation in epistemology.
One reason for this is that the question whether or not such a principle of
inference is innate is an empirical one, needing to be settled by the sorts of
considerations adduced in chapter 7. So we cannot know that the principle is
foundational until we have provided some evidence of it. But such evidence will
very likely presuppose that we can have knowledge of the physical world, which
is precisely what is in question.
Moreover,
the same point can be made here that we insisted on in connection with our
perceptual beliefs. To say that a principle of inference is innately given is
to say that it belongs to the foundations of our first-order system of beliefs.
And if that principle is in fact generally reliable, and the beliefs formed
through its use are in fact true, then it will contribute to our first-order
knowledge. But it does not follow that we may take that principle for granted
when we come to enquire what we may know ourselves to know. Since it is by no
means self-evident that inference to the best explanation is a reliable method
of acquiring beliefs, there is a question-mark over its status when we come to
do epistemology, just as there is a question-mark over the status of our
first-order perceptual beliefs. We have no option but to attempt to justify our
reliance on such a principle, even if we know it to be innate.
In
the final chapter of this book I shall attempt such a justification, making
crucial use of the concession of nativism. But first let us consider whether
the principle of inferring to the best explanation could in any case help us to
overcome scepticism about the physical world. For if it would not be of any
help, then there is little point in discussing whether or not that principle
can be justified. Let us therefore proceed for the remainder of this chapter on
the assumption that inference to the best explanation is justified. This
assumption will be vindicated in the chapter which follows.
Inferring to the Best
Explanation
What we need to decide, in effect, is whether
the hypothesis of our reliable perception of a physical world provides the best
available explanation of our perceptual experience, and of the existence of our
first-order perceptual beliefs. There are only four candidate hypotheses that
we need to consider.
The
first is that there is indeed a physical world, and that it is a reliable cause
of our perceptual beliefs. According to this hypothesis, the explanation for
the fact that there seems to be a room full of people in front of me, and of my
consequent belief that there is, is that there really is a room full of people
which causes my experience. Let us call this 'the realist hypothesis'.
The
second hypothesis also maintains that there is physical world, but does not
suppose that this world is a reliable cause of our beliefs. According to this
hypothesis, there may indeed be such things as rooms and people, but the
explanation for the fact that I seem to see a room full of people may be some
quite unrelated physical situation, such as that some scientists are
stimulating the visual centre of my brain, which is in fact floating in a vat
of chemicals in their laboratory. Let us call this 'the brain in the vat
hypothesis'.
The
third hypothesis maintains that there is no physical world, and no cause of our
experiences outside of our own psychology. According to this hypothesis, our
experience is one gigantic hallucination. Then the explanation for the fact
that I seem to see a room full of people will be some unknown feature of my
psychology, which results in me having this experience now, rather than any
other. This might be called 'the dream hypothesis'.
The
final hypothesis is again that there is no physical world, but that our
experience is caused by some non-physical fact independent of our minds. For
example, the explanation for the fact that I seem to see a room full of people
may be that there is an all-powerful demon who decides to cause me to have just
such an experience. This might as well be called 'the demon hypothesis'.
Which
of these available hypotheses provides the best explanation of my experience,
and of my perceptual beliefs? Plainly, I think, the realist hypothesis. This
can supply a detailed and highly sophisticated explanation of the course of my
experience, at the same time warranting subjunctives about what I would
experience were I to be differently situated, and counterfactuals about what I
would have experienced had I been differently situated in the past. This
hypothesis can not only explain the bare fact that I seem to see a room full of
people, but can also explain the details of what I seem to see in terms of the
details of my physical orientation and the distribution of people in the room.
It can also predict that I would seem to see a largely empty corridor if I were
to leave the room, and that I would have been seeming to see such a corridor if
I had stopped just before I came into the room. So the hypothesis has all the
hall-marks of a good explanatory theory: it provides a simple, but powerfully
predictive, explanation of the phenomena.
All
of the other three hypotheses, in contrast, are lamentably weak. The dream
hypothesis may be dismissed straight away, precisely because it makes no
attempt to explain the details of my experience. The brain in the vat
hypothesis, likewise, is incapable of providing detailed explanations. For if
my experience of a room full of people is dependent upon the whim of an
experimenter, then I cannot explain in detail why I have the experiences that I
do. Nor can I predict what I shall experience next, nor say what I might have
experienced in other circumstances. The demon hypothesis too, while being
ontologically simple (it only requires us to postulate a single thing outside
of our minds), is, as it stands, lacking in predictive power. For who knows
what the demon might cause me to experience next?
The
demon and brain in a vat hypotheses also explain less, because they leave the
demon's decisions (or those of the experimenter) themselves unexplained.
Explanation will come to an end with such claims as that the demon chooses that
it should appear to me that there is a room full of people. In contrast, the
hypothesis of reliable perception can not only explain why I am having such an
experience ('There really are people present'), but can also explain what does
the explaining ('They have come to listen to me lecture').
The
demon and brain in the vat hypotheses also have less coherence, in that they
employ separate principles where the hypothesis of reliable perception can make
use of a single set of laws. For example, apparent motion can be explained on
the realist hypothesis, either by movement of the object perceived, or by
movement of the perceiver; the same set of physical principles being involved
in either case. But on the demon hypothesis, we should have to suppose that
there are three distinct principles which the demon chooses to employ. On the
one hand he sometimes chooses that it should appear to me that things are in
motion. On the other hand, in a situation where he would have decided not
to cause an appearance of motion, if I think to myself 'I shall move', then the
demon does cause such an appearance. There are also rare cases in which,
where he would otherwise have caused an appearance of motion, if I think to
myself 'I shall move', then he decides not to cause such an appearance
(for example, cases where I walk parallel to a moving object in an otherwise
darkened room, so that the object appears stationary).
The
only recourse for the demon hypothesis (as for the brain in the vat
hypothesis), is to suppose that for some reason the demon (or the experimenter)
chooses to cause me to have experiences exactly as if there were the sort of
physical world which I intuitively believe there to be. The only way for the
demon hypothesis to get the explanatory and predictive power of realism, is for
it to be cast in the form 'The demon decides to give me experiences for which
the best explanation would otherwise (before mention of the demon) be ...',
followed by every claim of realism. Then I can not only explain my experience
of a room full of people, but explain what does the explaining - it is because
the demon wishes me to believe that there are people who want to listen to me
lecture. And I can predict that I should have experience of a largely empty
corridor if I were to leave the room - it is because the demon wishes me to
believe that there are corridors which remain in existence when unobserved, and
which are generally empty during lecture hours. And so on.
But
this now renders the demon hypothesis more complex than realism, since it contains
all of the latter's complexity, together with the additional supposition of the
demon and his decisions. Although the demon hypothesis is ontologically
simpler, since it only postulates a single thing outside of myself, it is
structurally more complex, since every statement within realism will have to
appear within the demon hypothesis, preceeded by a claim about the demon and
his decisions. (This must be so if the latter is to share the former's
predictive power.) And it is important to see that in assessing explanations it
is structural, rather than ontological, simplicity that matters. This is
because, in giving an explanation of some phenomenon, it is always facts rather
than objects which do the explaining. For example, the mere existence of rooms
and people cannot explain my experience of a room full of people; it is the
fact that there are people in a room within my line of vision which explains.
So when we are told that better explanations do not multiply entities beyond
necessity (which sounds like a direct exhortation to ontological simplicity),
this is only because theories which postulate more kinds of thing will also, in
general, have to postulate more kinds of fact - but not always, as the
comparison of realism with the demon hypothesis makes clear.
Another
point is that on the present proposal the demon hypothesis becomes entirely parasitic
upon realism. The only way to actually work the hypothesis, is to forget about
the demon, and to get on with the job of constructing the best available
version of realism. This means that the demon hypothesis, besides being more
complex, also lacks another virtue of a good explanatory theory - that of
fruitfulness. One of the marks of a good theory, it is generally agreed, is
that it should be fruitful, leading to new predictions which could not
previously have been made, and suggesting new lines of research. Yet in this
respect the demon hypothesis is entirely inert. It predicts nothing which is
not also predicted by realism, and not only does it fail to suggest new avenues
of research, but it is impossible to do research except by forgetting
the demon and constructing the best theories that one can within the orbit of
realism.
It
is clear, then, that the hypothesis of reliable perception provides
overwhelmingly the best available explanation of the course of our experience,
and of the existence of our perceptual beliefs. (I shall return in the next
chapter to the sceptical suggestion that the demon may also be interferring
with my powers of best explanation, leading me to reject as worse an hypothesis
which may actually be better.) It is therefore a matter of some importance to
decide whether we can legitimately rely upon inference to the best explanation
in attempting to combat scepticism. If we can, then it would appear that the
sceptic can indeed be answered. But before turning to that task, I propose to
consider one further objection.
Must Knowledge be Foolproof?
The objection is as follows. Even if it is
granted that we may somehow know inference to the best explanation to be a
generally reliable method of acquiring beliefs, it is certainly not a foolproof
method. On no account of the matter will an inference to the best explanation
of some phenomenon provide any sort of guarantee of the truth of the
conclusion. Even if it is generally reliable, it is also undeniably sometimes
fallible. So, even given that the hypothesis of reliable perception provides
the best explanation of the course of our experience, it remains possible that
it should be false. Indeed, it remains possible that the demon hypothesis,
although worse (in terms of simplicity, fruitfulness, and so on), is in fact
the correct one. It may then be objected that our belief that we have knowledge
of the physical world cannot, on this basis, constitute knowledge. Since it is
not completely certain, we cannot really know it.
This
objection represents a perennial temptation in philosophy - namely, to insist
that you cannot really know anything unless you can be completely
certain of it. So even if we have reason for believing inference to the best
explanation to be generally reliable, and reason for believing that we have
knowledge of the physical world, we cannot properly be said to know
these things, since we have no absolute guarantee of their truth. On this
account, nothing can properly constitute knowledge, except where it is
inconceivable that it should be false given the evidence. (Or perhaps: except
where it is inconceivable that it should turn out to be false given the
evidence - which is not quite the same thing.)2
It
is worth noting that we never in fact insist upon absolute certainty in our
practical lives, no matter how much may be at stake. For example, imagine
yourself as a juror in a murder case, in a State which retains the
death-penalty. And imagine that the evidence against the defendant is, as we
say, overwhelming. Nevertheless, as a scrupulous person, you ask yourself
whether you really know that the defendant is guilty. Now suppose that one of
the other jurors has proposed an alternative hypothesis - namely, that a group
of super-intelligent Martians may have landed on Earth undetected, and planted
all the evidence against the defendant for purposes known only to themselves.
This hypothesis is certainly a possible one, in the same way that the
hypothesis of the omnipotent demon is. It therefore shows that your belief that
the defendant is guilty is not absolutely certain. But plainly this alternative
proposal would not even cause you to hesitate in handing down a guilty verdict,
despite the fact that someone's life hangs in the balance.
What
emerges is that it is far from obvious why we should even possess the concept
of knowledge, if knowledge really did require complete certainty. For this is
not a concept that we have any use for in our practical lives. I suppose it
might be replied that such a concept would have a use, as a kind of ideal, to
which our ordinary standards of justification may approximate but never attain.
But in fact I think that the tendency which philosophers have, to insist that
knowledge requires certainty, is an illusion, produced by a feature of the
concept of knowlege which it shares with many of our other concepts - namely,
that it is purpose relative. Let me explain.
Notice
that the standards which we ordinarily insist on for a belief to count as
knowledge may vary according to the seriousness with which we take the
situation in which the question of knowledge arises. For example, suppose that
I ask you, as a casual inquiry, whether you know that Mary will be at the
Philosophy Department party. You reply that you do know this, because she told
you that she would be there. In context, I think we should ordinarily be
inclined to allow that you do know that Mary will be at the party. But now
imagine an exactly similar exchange taking place in the context of a murder
trial. I ask you whether you know that Mary is the murderer, and you reply that
you do know this, because she told you so. Even granting your sincerity, it is
plain that what you say would be insufficient to convict her. Here, since a
life or a life-sentence hangs in the balance, we would insist on greater
justification than this before allowing that you have knowledge. This sort of
variability is easily explained if 'knows' means 'is a true belief which is
reliably enough grounded for the purposes in hand'. Moreover, since the
contexts in which we employ a concept of knowledge may contain a variety of
different purposes and interests of varying importance, it is easy to see that
our need for such a concept is fulfilled by the possession of just such a
purpose-relative concept.3
It
is arguable that a great many of our concepts are purpose-relative in precisely
this sort of way. For example, consider the term 'flat'. Suppose that you are a
farmer, who owns a number of fields which are at present under grass. You are
first approached by someone from your local Territorial Army, who is looking
for a flat field on which to practice manoevres. You point to a particular
field and say 'That one is flat'. On another occasion you are approached by
your local bowls club, who are looking for a flat field on which to play while
their bowling green is under repair. You point to the very same field and say
'That one is not flat'. On both occasions, surely, you might say something
true. The explanation for this is that 'flat' means 'is flatter than enough
things of the kind in question for the purposes in hand'. Where the purposes
differ, so does the correct application of the concept.
Now
the important point to notice is that, where there are no purposes in
hand, the standards for applying such concepts are apt to become unlimitedly
high. This may be because one then naturally sets standards which would be apt
for any conceivable purpose. For example, suppose that I ask, in the
context of a theoretical discussion of scepticism, where no particular
practical project is in hand, 'Is this table really flat?' There would be a
very natural tendency to reply in the negative. You would immediately think,
for example, that the surface of the table is bound to contain slight
irregularities, even if they are undetectable to sight or touch. Thus, with no
particular purpose in hand, you are inclined to insist that nothing can really
be flat unless it is absolutely flat.
Similarly, then, in connection with the concept of knowledge. When we ask, outside of the context of any practical purpose or project, 'Do we really know that such-and-such?', there is a very natural tendency to deny it. This explains the impulse to reject our claim to have knowledge of the physical world, even given that the hypothesis of reliable perception provides overwhelmingly the best explanation of the course of our experience. But for all that, our belief that we have knowledge of physical objects may be well enough justified to count as knowledge for any purpose which we might actually have. I therefore conclude that, if we can somehow justify our reliance upon inference to the best explanation, scepticism about the physical world will have been decisively refuted.