sceptical scenarios are not error-possibilities
TRANSCRIPT
ORI GIN AL ARTICLE
Sceptical Scenarios Are Not Error-Possibilities
Tim Kraft
Received: 16 November 2009 / Accepted: 5 December 2012 / Published online: 25 December 2012
� Springer Science+Business Media Dordrecht 2012
Abstract On a common view of scenario-based sceptical arguments sceptical
scenarios are error-possibilities, i.e. their point is to introduce the possibility of
having only false beliefs (about the external world). However, global error is
impossible for purely logical/conceptual reasons: Even if one’s beliefs are consis-
tent, the negations of one’s beliefs need not be consistent as well. My paper deals
with the question of what the consequences of this result are. Two attempts (made
by Muller and Genova) at repairing scenario-based sceptical arguments within the
framework of understanding sceptical scenarios as error-possibilities are found
wanting. Instead, what should be given up is the assumption that sceptical scenarios
are error-possibilities. What is thought-provoking about the scenario of the brain in
a vat is not that none of its empirical beliefs are true, but that all of its empirical
beliefs fall short of knowledge at the same time. Hence, sceptical scenarios are not
error-possibilities, but ignorance possibilities. If this is so, both the closure argument
and the underdetermination argument commit a subtle mistake and should be
replaced by slightly different arguments. The principle of excluded ignorance-
possibilities turns out to be an epistemological principle that is faithful to scepti-
cism’s tenets without misinterpreting sceptical scenarios as error-possibilities.
Cartesian sceptical arguments are sceptical arguments that take a so-called sceptical
hypothesis or sceptical scenario as their starting-point: I might e.g. be a brain in a
vat connected to a supercomputer in an otherwise empty universe. On a common
understanding of scenario-based sceptical arguments the point of presenting such a
scenario is to introduce the possibility of global error. The possibility of being a
brain in a vat is a threatening possibility for the reason that if I were a brain in a vat,
all or almost all of my empirical beliefs would be false. Unfortunately, this cannot
T. Kraft (&)
Institut fur Philosophie, Universitat Regensburg, 93040 Regensburg, Germany
e-mail: [email protected]; [email protected]
123
Erkenn (2013) 78:59–72
DOI 10.1007/s10670-012-9423-2
be the point of presenting a sceptical scenario because global error is impossible for
purely logical reasons. To give an example, my beliefs that there are trees and that
there no trees on which meat grows cannot both be false (cf. Muller 2003; Gemes
2009, 2010). In this paper I discuss what consequences, if any, the impossibility of
global error has for the interpretation of scenario-based sceptical arguments. Against
two attempts to accommodate the impossibility of global error within the framework
of understanding sceptical scenarios as error-possibilities (Muller 2003; Genova
2010) I argue that this impossibility indicates that something is wrong with the very
idea that sceptical scenarios are all about error. Instead, sceptical scenarios should
be understood as illustrating the possibility of global ignorance, i.e. the possibility
of all empirical beliefs falling short of knowledge at the same time. In a nutshell,
sceptical scenarios are not error-possibilities, but ignorance-possibilities.
I first explain in more detail why global error is impossible (Sect. 1). Second, I
argue that Muller’s and Genova’s strategy for dealing with such examples is
unconvincing (Sect. 2). Third, I argue on independent grounds that sceptical
scenarios are ignorance-possibilities (Sect. 3). Fourth, I present a diagnosis of why it
is commonly assumed that sceptical scenarios are error-possibilities: Both the
closure argument and the underdetermination argument rest on the assumption that
sceptical scenarios are error-possibilities (Sect. 4). On the constructive side, I offer a
suggestion about how scenario-based sceptical arguments are to be understood
instead, namely as relying on the principle of excluded ignorance-possibilities (Sect.
5), and explain how this proposal helps to make sense of the distinction between
scepticism and fallibilism (Sect. 6).
1 Not All of Our Empirical Beliefs Can be Jointly False
Can all of our beliefs—or at least all of our empirical beliefs1—be false? Muller
(2003) and Gemes (2009) argue for a negative answer by giving examples for
beliefs that cannot be jointly false. This is Muller’s example2:
(M1) There are no unicorns.
(M2) There are animals. (2003: 47)
Given the conceptual truth that all unicorns are animals the negation of (M1)
implies (M2). Hence, only one of the two beliefs can be false. Both beliefs are
obviously empirical beliefs. Thus, not all of one’s empirical beliefs can be false.
Gemes presents a different example:
(G1) I have a hand.
(G2) It is not the case that I have a hand with a wart on it. (2009: 219)
1 I assume throughout that scenario-based sceptical arguments are best understood as being targeted at
our empirical beliefs and not at our beliefs about the external world. Although the classes overlap, they
are not coextensive.2 Muller credits me for this example and the objection behind it (Muller 2003: 47).
60 T. Kraft
123
Given the logical truth that hands with a wart on it are hands the negation of (G2)
implies (G1).3 Hence, only one of the two beliefs can be false. Both beliefs are
obviously empirical beliefs. Thus, not all of one’s empirical beliefs can be false.
With their examples Muller and Gemes are successful in establishing what I shall
call the Muller-Gemes theorem: Not all of our empirical beliefs can be false.4 This
theorem can be established by reflecting solely on the logical and conceptual
interrelations between our beliefs. No consideration of sceptical scenarios is needed.
The crucial point is purely logical: If a set of propositions is consistent, the set
consisting of the negations of the propositions in the original set need not be
consistent as well.
What Muller and Gemes disagree about is what the philosophical import of this
theorem is. Gemes’ conclusion is the most straightforward one: According to the
title of his paper (‘‘A refutation of global scepticism’’) he claims to have refuted a
position called ‘‘global scepticism’’. Note, however, that Gemes uses ‘‘global
scepticism’’ in a somewhat idiosyncratic way. He defines global scepticism as the
claim that all of one’s empirical beliefs can be false. This is—as he is happy to
admit (Gemes 2010: 64 f.)—a far cry from the epistemological claim that I do not
know anything empirically. Proving that at least one of my empirical beliefs is true
does not imply that I know at least one proposition empirically. Thus, Gemes’
refutation of global scepticism is not to be equated with a refutation of
epistemological scepticism.5 It does not follow, however, that the refutation of
global scepticism is completely irrelevant for an assessment of epistemological
scepticism. Sceptical scenarios are relied upon in sceptical arguments not as some
sort of decoration or embellishment, but are supposed to play an important
argumentative role. Since the reason why they are invoked is to establish the
possibility of global error, the Muller-Gemes should leave us wondering how to
understand scenario-based sceptical arguments. Do they already get off on the
wrong foot?
3 Gemes (2009: 218) and Genova (2010: 59) both claim that the negation of (G2) implies (G1) logically.
However, it is questionable that ‘‘hands with a wart on it are hands’’ is a logical truth. Since ‘‘waistcoats
with two sleeves are waistcoats’’ and ‘‘a vegan salad with a yoghurt dressing is a vegan salad’’ do not
seem to be true, ‘‘hands with a wart on it are hands’’ might as well be a conceptual truth. Since the
example could easily be altered to avoid confusions of this sort, I will ignore this point.4 Unlike Muller, Gemes and Genova explicitly accept a stronger version of the theorem, namely that at
least one of my empirical beliefs must be true (Gemes 2010: 63; Genova 2010: 60). The theorem, as
stated in the main text, is only equivalent to the claim that at least one of my empirical beliefs is not false.
If there are no truth-value gaps, my formulation of the theorem is equivalent to the formulation by Gemes
and Genova. However, if there are truth-value gaps, a belief is not false iff it is either true or lacks a truth-
value. Accordingly, if there are truth-value gaps, the examples by Muller and Gemes only establish the
weaker theorem. To prevent a possible misunderstanding having to do with scope: The theorem states that
my actual beliefs are such that it is impossible that all of them are false. The claim is not that it is
impossible that my beliefs are such that all of them are false (Muller 2003: 3 f.; Gemes 2010: 70). If
someone only has a very limited number of beliefs, it may be that they are all false.5 Gemes does not address the issue of what the consequences for scenario-based sceptical arguments are.
Since this paper and his paper do not address the same question, this paper is not meant as a reply to his
paper. Thanks to two anonymous referees for urging me to clarify this.
Sceptical Scenarios Are Not Error-Possibilities 61
123
2 How Not to Rescue Scenario-Based Sceptical Arguments
Muller and Genova do not think so. Both argue that scenario-based sceptical
arguments survive the Muller-Gemes theorem with no substantial loss. To show
this, they propose to partition the set of beliefs under attack from sceptical
challenges into two subsets, a set of beliefs that can be jointly false and a set of
inferred beliefs. I shall call the beliefs that admit of joint falsity ‘‘basic beliefs’’ and
the remaining beliefs ‘‘non-basic beliefs’’.
This is Muller’s version of the partitioning:
To do justice to this fact [=the Muller–Gemes theorem] in our investigation of
massive error we would have to view the belief system of the speaker as the
maximal conjunction of mutually independent beliefs (and the remainder of
his beliefs as implications of this conjunction). (Muller 2003: 47, my
translation)
To apply this proposal to his example, there are no unicorns, there are cats, thereare dogs etc. are basic beliefs. Together with the conceptual truths cats are animals,
dogs are animals etc. these beliefs imply there are animals. With such a partitioning
at hand, sceptical arguments can proceed in two steps. In a first step, it is argued via
the claim that all basic beliefs can (by definition) be jointly false that none of them
is knowledge. In a second step, it is argued that all non-basic beliefs are not
knowledge, because beliefs that are (or, as Muller’s wording suggests, can be)
inferred from something believed but not known cannot be knowledge. Unfortu-
nately, Muller does not explain any further which of our beliefs are the basic ones.
He merely postulates that there is a suitable set of mutually independent beliefs.
In contrast, Genova is not very explicit about the overarching strategy, but more
informative about which beliefs are the basic ones. What is called ‘‘basic beliefs’’
here is called ‘‘experience-based beliefs’’ by Genova.6 Experience-based beliefs are
not appearance reports, but are directed at objects outside our own minds. What
distinguishes an experience-based belief from other empirical beliefs is that an
experience-based belief is ‘‘engendered by its experienced content’’ (2010: 61), i.e.
its content is the content of an experience, not a proposition inferred from the
content of an experience.7 What this amounts to is explained best by looking at
6 A note on terminology: Genova borrows the term ‘‘experience-based belief’’ from Gemes. But unlike
Gemes, Genova does not use ‘‘experience-based belief’’ as a synonym for ‘‘empirical belief’’ or
‘‘aposteriori belief’’ (Gemes 2010: 65). Throughout this paper I use ‘‘empirical belief’’ in this wide sense
and ‘‘experience-based belief’’ in the more restricted sense proposed by Genova.7 I take it that it is not enough that the belief was de facto not inferred. (Here I disagree with Gemes’
reading of Genova’s notion of an experience-based belief, Gemes 2010: 65). According to Genova, as I
interpret him, a belief ceases to be an experience-based belief if its content is not the content of an
experience. All the cups in this room are white is not an experience-based belief—no matter how I formed
that belief. Even if I trained myself to form beliefs of the form ‘‘All Fs are Gs’’ by looking at a particular
F and without going through intermediate steps (e.g. this is a white cup, there’s no other cup in thisroom), my belief would not be an experience-based belief. There is a gap between the content of the
experience (this is a white cup) and the content of the belief (all cups in this room are white).
Independently of the details of the psychological process that led to the belief, the belief is not an
experience-based belief. (I do not mean to defend this conception of experience and its content. The point
of this footnote is purely exegetical.).
62 T. Kraft
123
Genova’s way of dealing with Gemes’ example. According to Genova, the example
fails because (G2) is not an experience-based belief. According to Genova,
(G3) My hand does not have a wart on it.
is an experience-based belief, while
(G2) It is not the case that I have a hand with a wart on it.
is not (2010: 61). The reason why the latter claim is not an experience-based belief
is that it is inferred from an experience-based belief like (G3). The content of an
experience cannot be represented by a sentence with external negation. The content
of an experience can (always?) be expressed by a statement of the form ‘‘a is F’’ or
‘‘a is non-F’’. Genova offers the following truth-conditions for internal negation:
‘‘a is non-F’’ is true iff a exists and it is not the case that a is F. (The latter negation
is the usual external or truth-functional negation.) Otherwise the sentence is false.
What motivates this conception of experience is (apparently) the thought that the
content of an experience is never existence-neutral. If a belief does not imply the
existence of some particular object—namely the object that is perceived to be such-
and-such—it cannot be an experience-based belief.
Effectively Genova offers the same way out as Muller: For the sceptical
argument to be successful it is enough that all our experience-based beliefs can be
false. Global scepticism restricted to an appropriately chosen subset of one’s beliefs
stands unrefuted: There is a subset of my beliefs that admits of joint falsity. Since all
other empirical beliefs are inferred (or at least inferable) from experience-based
beliefs, we do not know the former if we do not know the latter.
The Muller-Genova strategy is unsatisfying. Since Muller does not state a
criterion for being a basic belief, we are left wondering what that subset of our
beliefs is supposed to be. Since Genova merely rejects one example, it is merely
asserted, not argued for, that our experience-based beliefs admit of joint falsity. He
only invites our ingenuity to design a new example. It is no surprise that Gemes
(2010: 67 f.) comes up with new counterexamples.8 I restrict myself to presenting an
example of my own: I am not good at discriminating between different musical
instruments. Hearing a sound I come to believe:
1. This is not a trumpet.
and
2. This is an instrument.
As in all the other examples, (1) and (2) cannot both be false. But if anything is an
experience-based belief at all, (1) and (2) are. I believe (1) because it just does not
8 However, Gemes’ examples are not very convincing. The ultimate example should consist of beliefs for
which it is beyond doubt that the beliefs in question are experience-based beliefs. I do not think that
Gemes’ examples fulfil this condition. His last (and allegedly best) example consists of the pair there arebuffalo in Rourke’s valley and all the members of the buffalo herd in Rourke’s valley stampeded. Genova
should reply that the second belief is not an experience-based belief, because the content of the
experience is not that all buffalo stampeded, but rather that a buffalo herd stampeded (where the herd is
the object that is perceived to be such-and-such). The first belief is not an experience-based belief,
because the content of the experience can only be that these buffalo are in Rourke’s valley.
Sceptical Scenarios Are Not Error-Possibilities 63
123
sound like a trumpet. I do not believe (1) because it sounds like an oboe. Being such
a music ignoramus I have no idea how an oboe sounds like. I believe (2) because it
just sounds like an instrument. Again this belief is not inferred from another belief.9
Hence, the strategy of Muller and Genova fails. There is no guarantee that we can
isolate a core of basic and independent beliefs. (1) and (2) are basic beliefs in Genova’s
sense, but not independent. By contrast, they are not basic beliefs in Muller’s sense
(since they are not independent). But they are also not non-basic beliefs. For they are
not consequences of any other beliefs held by the subject. Hence, they are neither basic
nor non-basic beliefs and Muller’s partitioning of our beliefs is not exhaustive.
3 Sceptical Scenarios are Ignorance-Possibilities
The conclusion to draw from the discussion so far is that if scenario-based sceptical
arguments rest on the premise that global error is possible, they cannot succeed. Hence,
we need to reinvestigate the role of sceptical scenarios in scenario-based sceptical
arguments. What is the point of sceptical scenarios? What are they meant to illustrate?
I propose to answer this question by having a look at some ordinary empirical beliefs
and what happens to them when one is the victim of a sceptical scenario.
1. Some empirical beliefs are true ex hypothesi. Their truth follows directly from the
description of the sceptical scenario. In the scenario of the brain in a vat e.g. I have abrain, electrons have negative charge, the half-life of carbon-14 is 5730 years,
there are no unicorns remain as true as ever. Note that this is not an artefact of that
particular scenario. It is of the nature of a sceptical scenario that it provides an
explanation of why the victim believes what she does. Unless some story or other is
told about the sources of the beliefs, it is implausible that both the victim of the
sceptical scenario and someone living in a world that is more or less the way she
believes it to be can have the same beliefs. Whatever the details of that story, some
beliefs turn out to be true in both situations. Therefore, beliefs that remain true in the
sceptical scenario can simply be read off from the description of the scenario.10
9 A defender of negative free logic might object that (1) and (2) are both false if ‘‘this’’ fails to refer. (For
the sake of the argument I assume that even in the context of my example ‘‘this’’ might fail to refer.) I am
confident that a new example can be crafted against such an opponent. At this point I only want to point
out that sceptical arguments should not rest on highly controversial assumptions. If scenario-based
sceptical arguments are cogent only if one accepts Genova’s conception of experience and negative free
logic, sceptical arguments are avoided easily, namely by rejecting one of the controversial assumptions.
Moreover, sceptical scenarios are not scenarios in which ‘‘this’’ fails to refer. If I am a brain in a vat, my
aural perception exists and it is caused by a supercomputer; a fortiori what I hear is not a trumpet. Hence,
even if global error were possible, sceptical scenarios would not illustrate this possibility.10 Since descriptions of the sceptical scenario can be more or less detailed, it depends highly on the
actual description which beliefs remain true. In the main text I assume e.g. that the description includes a
statement like ‘‘the laws of physics are unaltered’’. What cannot be read off from the description of the
scenario is what happens to beliefs like Goldbach’s conjecture has been computer-checked for allnumbers up to at least 1018 or there is no exaflops computer yet. Since the technical details of the
supercomputer are usually not mentioned in the description, the truth-value of these beliefs is left open in
much the same way in which the truth-value of Sherlock Holmes’ blood type is A is left open by Conan
Doyle’s stories.
64 T. Kraft
123
2. Some empirical beliefs are vacuously true. This includes universally quantified
propositions and material conditionals with a false antecedent. Examples are
cats are furry, lemons are sour or if I go to a party tonight, I won’t be in bedbefore midnight. Since there are no cats, lemons or parties in the sceptical
scenario, all these beliefs turn out to be true.
3. Some empirical beliefs are neither true nor false. This includes many beliefs
whose content is a singular proposition. If I was a brain in a vat, Anne is adoctor, Sarkozy is the present president of France, this is a tree and so on would
be neither true nor false. In the possible world in which I am a brain in a vat
‘‘Anne’’, ‘‘Sarkozy’’, ‘‘France’’ are empty names and ‘‘this’’ a non-referring
demonstrative. On at least one plausible semantics of empty singular terms
these beliefs are neither true nor false.11
4. Some empirical beliefs are necessarily true. Many apriori truths are necessary.
Although they can be justified non-empirically, they need not be justified non-
empirically. Many examples can be used to illustrate this. Here is one in which
even a trivial logical tautology is believed because of an experience: Looking at
the mist I mutter ‘‘A dog is coming towards me… or it isn’t’’. Some seconds
after acquiring the belief I notice that it is a tautology. For a short period of time
I believed a tautology on the basis of an experience.
To sum up, in a sceptical scenario some empirical beliefs are true because of the
nature of the scenario, some are vacuously true, some are neither true nor false and
some are necessarily true. These beliefs do not form a small, negligible subset of
one’s empirical beliefs. Moreover, they clearly fall within the scope of sceptical
arguments. Thus, mere reflection on examples of empirical beliefs shows that there
is something wrong with the idea that sceptical scenarios are meant to illustrate the
possibility of global error. If sceptical scenarios were meant to accomplish this, they
would obviously be inadequate for their task. Charity forbids understanding
sceptical scenarios as dealing with the possibility of global error.
But what do they illustrate if it is not global error? Although the victim of a
sceptical scenario has a lot of true beliefs, she nevertheless does not know all these
matters. She does not have knowledge, because her beliefs are somehow or other
unfounded. Although the details are complex, it is not difficult to point out why a
brain in a vat does not know all these matters.12 What I have a hand, I have a brain,
electrons have negative charge, Anne is a doctor, there are no unicorns, cats arefurry etc. have in common is the absence of an adequate causal link to the relevant
object(s), i.e. my hand, my brain, electrons, Anne, cats. There are different ways in
11 If you prefer an account of empty singular terms that does not have this consequence, note that it
suffices for my argument that a sceptic is not committed to a view of empty singular terms according to
which sentences with empty singular terms have a truth-value. The force of scenario-based sceptical
arguments does not depend on the semantics of empty singular terms. (see also fn. 9.).12 It is often said that sceptical arguments establish an epistemic ‘gap’ or ‘divide’ between my beliefs and
the external world. In this paragraph I try to explain in a non-metaphorical way what this ‘gap’ or ‘divide’
amounts to. Usually the ‘gap’ is explained—if it is explained at all—in terms of falsity, i.e. the gap is
supposed to open up because of the possibility that it seems to me that P although it is false that
P. Obviously, this explanation is only adequate if sceptical scenarios are error-possibilities.
Sceptical Scenarios Are Not Error-Possibilities 65
123
which an adequate causal link can be absent: Sometimes the link is absent because
the relevant object does not exist (e.g. my hand, Anne) and sometimes there is a
causal link, but the link is not adequate for knowledge (e.g. electrons, my brain).
Some qualifications are needed to apply this claim to there are no unicorns and catsare furry. There must be an adequate causal link to the garden (and/or some of its
parts) if I am to know that there are no unicorns in the garden. There must be an
adequate causal link to the earth (and/or some of its parts) if I am know that there
are no unicorns on earth. Analogously, there must be an adequate causal link to the
external world (and/or some of its parts) if I am to know that there are no unicorns.
Another qualification is needed to apply my claim to cats are furry. There are two
ways of knowing this. I might know this because I am adequately linked to some
cats so that I am in a position to know that they are furry. I might also know this
because I am adequately linked to the external world so that I am in a position to
know that there are no cats. A brain in a vat is barred from both ways.
Hence, the most charitable interpretation of sceptical scenarios understands them
as illustrating the possibility of global (empirical) ignorance, the possibility of not
knowing anything from experience. A sceptical scenario is not defective just
because it does not illustrate the possibility of global error. To be successful it has to
illustrate the possibility of global (empirical) ignorance, i.e. the possibility that all of
my (empirical) beliefs fall short of knowledge.13
4 Why it is Attractive to Misunderstand Sceptical Scenariosas Error-Possibilities
So far I have argued that sceptical scenarios are misunderstood if one thinks of them
as error-possibilities. Interestingly, however, Gemes claims that what I claim to be a
misunderstanding is a ‘‘widely accepted view’’ (2009: 219). It is indeed striking how
often sceptical scenarios are described in terms of error and false belief. Hence, my
claim would be more plausible if I could offer a diagnosis of why it is tempting to
think that sceptical scenarios are error-possibilities although they are not. I do not
think that this misunderstanding is widespread because nobody so far has thought
thoroughly enough about sceptical scenarios. Instead, the misunderstanding is due
to the argumentative role sceptical scenarios are supposed to play. Consider the
argument from ignorance (DeRose 1995: 1). This argument relies on two premises.
In a first step, which gives the argument its name, it is argued that I do not know that
I am not e.g. a brain in a vat. In a second step it is argued that if I do not know that I
am not a brain in a vat, none of my empirical beliefs constitutes knowledge. These
premises entail that I do not know anything from experience. Especially the second
premise is in need of a defence. I want to argue that the standard defences of the
second premise lead to the misunderstanding of sceptical scenarios.
13 According to some accounts, ignorance is lack of information, i.e. S is ignorant that P iff it is true that
P but S does not believe that P. On my account, ignorance is lack of knowledge, i.e. S is ignorant that P iff
it is not the case that S knows that P. A useful side-effect of this stipulation is that on my account all error-
possibilities are ipso facto ignorance-possibilities (but not vice versa).
66 T. Kraft
123
The most popular defence of the second premise utilises the closure of
knowledge (under known implication): If one knows that P and also knows that
P implies Q, one is in a position to know that Q as well.14 Hence, (assuming that one
knows that being a brain in a vat implies not having a hand) one can only know that
one has a hand if one knows that one is not a brain in a vat. It is attractive to rely on
closure here, because closure is, at least prima facie, a very plausible principle.
Unfortunately, relying on closure has one crucial disadvantage: The closure
argument works only for beliefs that are indeed false when one is a brain in a vat. As
we have seen, it is far from obvious that a large set of beliefs fulfils this condition.
Although it is fulfilled by the ever so popular example I have a hand, it is not even
fulfilled by e.g. I have less than seven hands. I contend that the reason why it is
attractive to misunderstand sceptical scenarios as error-possibilities is that the
sceptical argument does not seem to get off the ground if they are not.
Replacing the closure principle by the underdetermination principle (Brueckner
1994; Cohen 1998) does not change the situation much. According to the
underdetermination principle, I do not know that I have hands if I do not know that
I am not a brain in a vat, because I can know that P only if my evidence for P favours
that P over all incompatible possibilities.15 Whatever the advantages of this argument
compared to the closure-argument, it suffers from the same problem. Since I cannot
both be a handless brain in a vat and have a hand, the question whether my evidence
favours I have a hand over I am a handless brain in a vat puts pressure on my belief
that I have a hand. But since I can both be a brain in a vat and have a brain, the question
whether my evidence favours I have a brain or I am a brain in a vat does not put
pressure on my belief that I have a brain. This can be argued for by an analogy:
Suppose I have evidence that I am sick. It is an interesting question whether my
evidence also favours the stronger claim that I have got the swine flu. But, obviously, I
can know that I am sick regardless of whether my evidence for the belief that I am sick
does or does not favour the belief that I have got the swine flu. Again I contend that the
reason why it is so attractive to misunderstand sceptical scenarios as error-possibilities
is that the argument does not seem to get off the ground if they are not.
5 How Scenario-Based Sceptical Arguments are to be Understood Instead
Either scenario-based sceptical arguments fail at a very early stage or another
epistemological principle than closure or underdetermination should be cited as a
defence of the second premise of the argument from ignorance. I want to propose
such an epistemological principle, the principle of excluded ignorance-possibilities.
According to this principle, S knows that P only if S can rule out all possibilities in
which S does not know that P. However, this principle is itself in need of a defence.
It is far from obvious that knowledge requires eliminating ignorance-possibilities.
14 Or closure’s close cousin, the principle of excluded error-possibilities. According to this principle, one
knows that P only if one can eliminate all possibilities in which P is false.15 Brueckner and Cohen state the underdetermination principle explicitly in terms of incompatiblehypotheses (Brueckner 1994: 830; Cohen 1998: 144).
Sceptical Scenarios Are Not Error-Possibilities 67
123
The distinction between non-epistemic and epistemic defeaters is useful here.16 A
belief can be defeated either by something that speaks in favour of the falsity of
what is believed or by something that is neutral with respect to the truth of what is
believed. The first kind of defeaters are non-epistemic defeaters, the second kind are
epistemic defeaters. Among epistemic defeaters are defeaters that speak against a
proposition the subject inferred the belief from, but also defeaters that speak against
the reliability or trustworthiness of the belief-forming process. Another useful
distinction is the distinction between non-modal and modal defeaters. The libraryhas new opening hours is a non-modal defeater of my belief that the library is still
open, whereas the library might have new opening hours is a modal defeater of that
belief. Since sceptical scenarios are obviously modal defeaters, I restrict my
attention to modal defeaters in what follows.
To give an example, suppose we are walking down the street. You claim that the
person with the umbrella that went into the store was Anne. I can defeat your belief
by pointing out that you might be too shortsighted to tell who went into the store
(epistemic defeater, possibility that is incompatible with your reliability) or that
someone of similar height might own a similar umbrella (epistemic defeater,
possibility that is incompatible with your reason for believing that it was Anne) or
that the person might be Anne’s sister (non-epistemic defeater, possibility that is
incompatible with the proposition believed). Hence, eliminating error-possibilities,
i.e. non-epistemic modal defeaters, like the possibility that it was Anne’s sister is
not sufficient for knowledge. At least some ignorance-possibilities, i.e. epistemic
modal defeaters, must be ruled out as well.
With the distinction between epistemic and non-epistemic defeaters at hand I can
help the sceptic to a defence of the principle of excluded ignorance-possibilities.
Although an example alone cannot establish the principle of excluded ignorance-
possibilities, what an example can show is that the existence of epistemic modal
defeaters is not an invention of the sceptic. The existence of these defeaters has to
be admitted independently of sceptical worries.
An argument that is closer to sceptical worries can be given as well. Lewis
defends the principle of excluded error-possibilities by pointing out that a
concessive knowledge attribution like ‘‘S knows that P, but she has not ruled out
the possibility that P is false’’ ‘‘just sounds contradictory’’ (Lewis 1996: 419). A
similar argument can be given for the principle of excluded ignorance-possibilities.
‘‘S knows that P, but she has not ruled out the possibility that P but she does not
know P’’ just sounds contradictory, too. Consider the following concessive
knowledge attributions:
1. Anne knows that it is 9 a.m., but she cannot rule out looking at a stopped watch
at exactly 9 a.m.
2. Ben knows that there are computers, but he cannot rule out that a computer
tricks him into believing that there are computers.
16 The distinction between two kinds of defeaters is implicit in Stroud’s discussion of the goldfinch
example vs. the Cleveland example (1984: 24–26). It is explicitly drawn using various names by e.g.
Pollock and Cruz (1999: 196, rebutting/undercutting), Williams (2001: 149, non-epistemic/epistemic) and
Casullo (2003: 44–46, overriding/undermining).
68 T. Kraft
123
3. Carla knows that she is holding a speech, but she cannot rule out holding a
speech while dreaming that she is holding a speech.
To my ears these knowledge attributions are as contradictory as Lewis’ concessive
knowledge attributions. Hence, I propose on the sceptic’s behalf to adopt a new
principle as the backbone of the argument from ignorance, namely the principle of
excluded ignorance-possibilities: S knows that P only if there is no uneliminated
ignorance-possibility. According to scepticism, sceptical scenarios are uneliminated
ignorance-possibilities (or epistemic modal defeaters). They are special defeaters
only insofar as they defeat all of one’s empirical beliefs at once.
6 How to Distinguish Scepticism from Fallibilism
So far I have taken for granted that scenario-based sceptical arguments utilise
exactly one scenario. Given this assumption scenario-based sceptical arguments
cannot take the form of either the closure or the underdetermination argument since
both arguments cannot show how one sceptical scenario can defeat all knowledge
claims at once. My proposal is to retain the assumption—one sceptical scenario
defeats all empirical knowledge claims—but to reject the claim that sceptical
scenarios are error-possibilities. Another, prima facie attractive option is to reject
the assumption and retain the understanding of sceptical scenarios as error-
possibilities: Although there is not one sceptical scenario defeating all empirical
beliefs, all beliefs are defeated by one or other sceptical scenario.17 My belief that I
have less than seven hands is brought down by one sceptical scenario (e.g. being a
dreaming seven-handed creature) and my belief that I have two hands is brought
down by another scenario (e.g. the standard brain in a vat scenario). The advantage
of that reading of scenario-based sceptical arguments is that the popular closure and
underdetermination arguments can be rescued. These could be understood as
templates: For every empirical belief there is an adequate scenario with which to run
the argument. I shall call this ‘multiple scenario scepticism’.
However, multiple scenario scepticism does not result in a convincing
interpretation of the role of sceptical scenarios in sceptical arguments. It is a
common theme in recent discussions of scepticism that scepticism is to be
distinguished from fallibilism. According to fallibilism, any (empirical) belief might
be false.18 Scepticism is a serious thread only if it makes a stronger claim than that
17 A third option is to think of the sceptical argument as a paradigm case argument. The idea here is that a
sceptical argument begins with casting doubt on a paradigmatic or even best claim to knowledge. If I do
not even know that I have two hands—so the argument goes—I cannot know anything else (cf. Cavell
1979: chs. 6 and 8; Williams 1996: 135). However, in order to evaluate the merits of this argument we
need answers to at least two questions: First, we need to know why this or that belief is taken to be a
paradigm case of knowledge. Secondly, we need to know how the failure of a single knowledge claim is
supposed to generalise to all knowledge claims of some kind. I conjecture that if these questions are
answered carefully, the paradigm case argument will turn out to be an enthymematic version of one of the
other arguments considered in the main text.18 In the main text I deal only with fallibilism about belief and not with fallibilism about justified belief or
knowledge.
Sceptical Scenarios Are Not Error-Possibilities 69
123
(cf. Williams 2001: 59–61). I will present two ways in which scepticism should be
stronger than fallibilism. Multiple scenario scepticism falls short of both of the
stronger claims and is, hence, only fallibilism in disguise.19
There is—or so I shall argue—a difference between sceptical and ordinary, non-
sceptical error-possibilities. Fallibilism can be equated with the claim that there is an
uneliminated ordinary error-possibility for every belief, whereas scepticism amounts
to the stronger claim that there is also an uneliminated sceptical error-possibility for
every belief. Suppose I claim to know that my bicycle is parked outside the
philosophy department. One interlocutor asks me if I can rule out that it was stolen
2 min ago. A second interlocutor asks me if I can rule out being a brain in a vat. A
sensible answer to the first interlocutor is to concede that I cannot rule out that
possibility and to insist at the same time that it is nevertheless quite improbable that
my bicycle was stolen because of the high quality of the lock I use. Even if I cannot
rule out the error-possibility, I can nevertheless offer some, albeit only non-
conclusive reasons that speak against it. The parallel answer is not a sensible answer
to the second interlocutor. I have no evidence at all pertaining to the sceptical
hypothesis. This is so because if I were a brain in vat I would take myself to have the
same evidence I take myself to have now. Everything I could offer against the
obtaining of the sceptical scenario would also be offered by a brain in vat. Reliance
on sceptical scenarios is an essential feature of scepticism, because otherwise one
could evade scepticism simply by lowering the standards for knowledge, i.e. one
could concede, on the one hand, that indubitable knowledge is impossible and insist,
on the other hand, that justified beliefs or probable opinions are still to be had.
Hence, what multiple scenario scepticism requires is that there is not just an
ordinary error-possibility, but also a sceptical error-possibility for each belief.
However, crafting an appropriate sceptical error-possibility is not as easy as it may
seem. At least simple adjustments of one of the standard sceptical scenarios will not
always do the trick. Consider as an example my belief that there are no snakes in my
office. It is false only if some room is in fact my office (and contains a snake). However,
planting the envatted brain in a room will not make the room the envatted brain’s
office—no matter what the room contains or how it looks like. One might hope to
circumvent this problem by supposing that I was envatted only recently, say last week.
Of course, then some room could be the envatted brain’s office. However, my total
evidence speaks against this possibility. Given all the evidence I acquired until last
week it is rather unlikely that there is a computer powerful enough to feed a human
brain with appropriate stimuli and it is even more unlikely that in addition to that a
snake ended up in my office. Although my evidence is not conclusive, it is not
completely neutral either. To sum up, on the one hand, multiple scenario scepticism
cannot rely on standard sceptical scenarios because some beliefs are not false in them.
On the other hand, appropriate modifications turn the scenarios into ordinary error-
19 An additional difficulty for multiple scenario scepticism: Since empirically justified beliefs can be
necessarily true, not all empirical beliefs can be the target of multiple scenario scepticism. An
interpretation of sceptical arguments should name a property that makes a belief susceptible to sceptical
attack. On my reading that is the property of being an empirical belief, on the alternative reading it is the
property of being a contingent empirical belief. This seems to be an unmotivated, ad hoc restriction. It is
an advantage of my reading of scenario-based sceptical arguments that no such restriction is needed.
70 T. Kraft
123
possibilities, i.e. into possibilities against whose obtaining I can offer some evidence.
Without proof to the contrary it is merely wishful thinking to suppose that there is a
sceptical error-possibility for every empirical belief a person might happen to have.
The second way to draw the difference between fallibilism and scepticism is to
conceive of it as a scope distinction.20 According to fallibilism, for all (empirical)
beliefs there is an error-possibility that I cannot rule out. According to scepticism,
there is an error-possibility for all (empirical) beliefs that I cannot rule out. If we
accept this distinction, what I have called ‘multiple scenario scepticism’ is
obviously not a form of scepticism. What is not so obvious is that this distinction
does indeed mark the border between fallibilism and scepticism. According to
multiple scenario scepticism, I do not know that P because I cannot rule out being in
scenario S and I do not know that P* because I cannot rule out being in scenario S*.
For some P and P* the respective scenarios S and S* are incompatible. Since I
cannot be both in S and S*, there is something that I can rule out, namely that all of
my beliefs are in a bad situation at the same time. This may provide only cold
comfort, but is still a reassuring thought: Bad things can happen to any of my
beliefs, but not to all beliefs at the same time. However, if sceptical scenarios are
ignorance-possibilities, this reassuring thought is blocked. On this reading of
sceptical scenarios, being e.g. a brain in a vat is bad for all empirical beliefs, not
only for those empirical beliefs that happen to be false in that scenario.
I have argued that multiple scenario scepticism is only fallibilism in disguise. It is
an advantage of my reading of sceptical scenarios that the distinction between
scepticism and fallibilism can be upheld. If we replace talk of error-possibilities
with talk of ignorance-possibilities, this results in a new reading of both fallibilism
and scepticism: According to fallibilism, there is an uneliminated ignorance-
possibility for each (empirical) belief.21 According to scepticism, there are not only
ordinary but also sceptical ignorance-possibilities and there is exactly one
ignorance-possibility defeating all empirical knowledge claims. Hence, the distinc-
tion between scepticism and fallibilism can and should be understood in terms of
ignorance-possibilities, not in terms of error-possibilities.
7 Conclusion
To sum up, sceptical scenarios are not error-possibilities and, therefore, scenario-
based sceptical arguments must rely on an epistemological principle considerably
different from the closure or the underdetermination principles. My proposal is to
replace these principles by the principle of excluded ignorance-possibilities. But the
main thesis of this paper does not depend on this proposal: To pave the way for a
proper understanding of the role of sceptical scenarios in sceptical arguments we
20 This distinction is also drawn by Davidson (1983: 140). A related distinction is the distinction between
specific and general doubts. That nothing is exempted from doubt can mean that one can doubt anything
or that one can doubt everything at once (cf. Wittgenstein 1969: §115, §232).21 Hetherington (1999) and Reed (2002) offer similar readings of fallibilism. Their main argument,
however, is a different one: The standard account of fallibilism cannot explain how necessarily true
beliefs can be fallible. A note of caution: Both are interested primarily in fallibilism about knowledge.
Sceptical Scenarios Are Not Error-Possibilities 71
123
should forget once and for all the idea that sceptical scenarios are error-possibilities;
they are ignorance-possibilities.22
Acknowledgments I am grateful to Christian Beyer, Jan Gertken, Tobias Klauk, Clayton Littlejohn,
Dolf Rami and Hans Rott for helpful criticism of this paper and to Olaf Muller for years of discussions
about scepticism.
References
Beebe, J. (2010). Constraints on sceptical hypotheses. Philosophical Quarterly, 60, 449–470.
Brueckner, A. (1994). The structure of the skeptical argument. Philosophy and PhenomenologicalResearch, 54, 827–835.
Casullo, A. (2003). A priori justification. Oxford: Oxford University Press.
Cavell, S. (1979). The claim of reason. Oxford: Clarendon.
Cohen, S. (1998). Two kinds of skeptical argument. Philosophy and Phenomenological Research, 58,
143–159.
Davidson, D. (1983). A coherence theory of truth and knowledge. (Reprinted in Subjective,Intersubjective, Objective, pp. 137–157, Oxford: Clarendon, 2001).
DeRose, K. (1995). Solving the skeptical problem. Philosophical Review, 104, 1–52.
Gemes, K. (2009). A refutation of global scepticism. Analysis, 69, 218–219.
Gemes, K. (2010). A vindication of a refutation of global scepticism, a refutation of global perceptual
scepticism and a refutation of global existential scepticism. Analysis, 70, 63–70.
Genova, A. C. (2010). Has Gemes refuted global scepticism? Analysis, 70, 59–63.
Hetherington, S. (1999). Knowing failably. Journal of Philosophy, 96, 565–587.
Lewis, D. (1996). Elusive knowledge. (Reprinted in Papers in metaphysics and epistemology,
pp. 418–445, Cambridge: Cambridge University Press, 1999).
Muller, O. L. (2003). Wirklichkeit ohne Illusionen. vol. 1. Paderborn: Mentis.
Pollock, J., & Cruz, J. (1999). Contemporary theories of knowledge (2nd ed.). Lanham: Rowman &
Littlefield.
Reed, B. (2002). How to think about fallibilism. Philosophical Studies, 107, 143–157.
Schaffer, J. (2010). The debasing demon. Analysis, 70, 228–237.
Stroud, B. (1984). The significance of philosophical scepticism. Oxford: Clarendon Press.
Williams, M. (1996). Unnatural doubts. Princeton: Princeton University Press.
Williams, M. (2001). Problems of knowledge. Oxford: Oxford University Press.
Wittgenstein, L. (1969). On certainty. (edited by G.E.M. Anscombe and G.H. von Wright, translated by
G.E.M. Anscombe and D. Paul, Oxford: Blackwell).
22 Beebe (2010) and Schaffer (2010) argue for a similar conclusion. Schaffer (2010) introduces the
debasing demon who ‘‘preys not on the truth requirement but rather on the basing requirement’’ (2010:
231). He introduces this demon in order to present an argument for universal scepticism, i.e. scepticism
that is not limited to empirical beliefs. Beebe (2010) argues inter alia that ‘‘it is widely but incorrectly
believed that sceptical hypotheses must describe scenarios in which subject’s beliefs are false’’ (2010:
449). Beebe’s argument centres on the possibility of sceptical doubts about apriori beliefs and necessarily
true beliefs as well. Both do not claim, as I do, that the thesis that sceptical scenarios are error-
possibilities already fails with respect to empirical scepticism.
72 T. Kraft
123