Published in the American Philosophical Quarterly 33/1 (July 1996): 247-261
MODAL BLOOPERS:
WHY BELIEVABLE IMPOSSIBILITIES ARE NECESSARY*
Roy A. Sorensen
Dartmouth College
Here is a debate I cannot lose. I
argue that it is possible to (at least unwittingly) believe the
impossible, say, that
there is a largest prime number. The impossibilist objects that
I am mistaken. Wrong move! By trying to correct me, the impossibilist
concedes that I believe a false proposition. The proposition in
question (i.e. that impossibilities can be believed), if false,
is necessarily false. Thus, the impossibilist would be conceding
that an impossibility can be believed.
If believing the impossible is impossible,
then so is believing one can believe the impossible, and believing
one can
believe one can believe the impossible, and so on. The "and
so on" elevates my debating point beyond the status of a
rhetorical zinger. Most transcendental arguments can be circumvented
with judicious rephrasing. Claims about essential preconditions
have a notorious tendency to beg the question. However, the soundness
of "I believe that it is possible to believe an impossible
proposition, therefore, it is possible to believe an impossible
proposition" can only be challenged in ways that also undermine
its intended adversaries. For example, a skeptic about modalities
might judge the premise as false on the grounds that my thesis,
<>(Ep)(Ex)(~<> p & Bxp), is meaningless. However,
the impossibilist can only accept skepticism about modality by
undermining his own thesis. For the negation of a modal proposition
is itself a modal proposition. Or to emphasize a related symmetry,
the negation of any meaningful sentence must itself be meaningful,
so the impossibilist must view his adversary's position as meaningful.
This prevents him from portraying me as a "believer"
who lacks an object of belief. If the impossibilist can get into
the belief relation with `No one can believe an impossibility',
then I can get into the
belief relation with its negation.
I. BELIEF IN BELIEVABLE IMPOSSIBILITIES IS INFALLIBLE
I take myself to be in a debate that philosophers normally only dream about. To win, I need only believe my thesis. (Indeed, later it will be demonstrated that I need only appear to believe it.) The reason is that belief in `Some impossibilities are believable' guarantees its own truth. Believing so makes it so. (Self-intimation is the converse where being so makes it believed so.) This logical entailment ensures that the argument does not rely on dialectical hocus pocus. Those who affirm Bp > <>p tend to subscribe to substantive modal logics such as the highly popular S5. Therefore, no question is begged by recasting my argument for the infallibility thesis with the help of an asocial, though heart-warmingly quick and dirty proof:
1. Ba[<>(Ep)(Ex)(~<>p & Bxp)] Premise.
2. ~<>(Ep)(Ex)(~<>p & Bxp) Assume for the sake of a reductio ad absurdum.
3. ~<><>(Ep)(Ex)(~<>p & Bxp) 2, Whatever is impossible is not possibly possible.
4. ~<>[<>(Ep)(Ex)(~<>p & Bxp)] & Ba[<>(Ep)(Ex)(~<>p & Bxp)] 3, 1 Conjunction.
Intermission: Notice that line 4 gives us a concrete example of believing the impossible. To make the point a matter of principle, we need only name the bracketed proposition and apply existential generalization. So let m name [<>(Ep)(Ex)(~<>p & Bxp)]. And on with the show:
5. ~<>m & Bam 4, Synonymous redescription of the proposition with a name.
6. (Ex)(~<>m & Bxm) 5, Existential generalization.
7. (Ep)(Ex)(~<>p & Bxp) 6, Existential generalization.
8. <>(Ep)(Ex)(~<>p & Bxp) 7, Whatever is actual is possible.
9. ~~<>(Ep)(Ex)(~<>p & Bxp) By reductio ad absurdum from 2 through 8.
If we venture from the modal system S4 to S5, we can go on to invoke the principle that whatever is possible is necessarily possible. This yields the modest but fertile necessary truth:
[]<>(Ep)(Ex)(~<>p & Bxp).
If no one actually happened to believe
<>(Ep)(Ex)(~<>p & Bxp), then my thesis would still
be rendered true by a possible believer. `Some impossibilities
can be believed' may look contingent because the natural demonstration
is prompted by my actual reply in a real-life debate. From the
contingent fact that I believe <>(Ep)(Ex)(~<>p &
Bxp) the possibility of believing in impossibilities follows by
the principle that whatever is actual is possible. Since whatever
is possible is necessarily possible, my contingent belief constitutes
decisive evidence for a
necessary truth. Or to put the point cruelly, impossibilism is
itself impossible. Belief in the thesis that impossibilities cannot
be believed is a subtle instance of what it precludes!
The modal status of the commitment
to believable impossibilities exposes a quasi-Godelian tension
between deductive closure and consistency. In addition to driving
another nail into the coffin of doxastic logic, []<>(Ep)(Ex)(~<>p
& Bxp) ruptures theories that link belief and necessity such
as Brian Ellis' epistemic
semantics and Robert Stalnaker's possible worlds analysis of propositions.
This will be backed up at the end of the
paper.
Another welcome feature of my thesis
is that its infallibility is dialectically robust; my adversaries
cannot wriggle
free by rejecting discretionary background assumptions. Some might
think that the strength of this result is inversely proportional
to its interest. If it is certain that impossibilities can be
believed, then the impossibilists already have their trousers
half-way down. Am I just pulling their trousers all the way down?
Only in the fashion praised through
the ages. The skeptic, the egoist, and the hard determinist are
partly refuted
by obvious counterexamples -- the sort volunteered by sophomores
and refined by common sense philosophers such as Thomas Reid and
G. E. Moore. However, our teachers showed how these (usually authentic)
counterexamples beg the question against subtle characters such
as the skeptic, the egoist, and so on. Since
question begging is relative to an audience, the counterexamples
may rightly shape the opinion of other people. Partial refutations
are instructive. However, universal refutations have special value.
These use premises that even one's adversary must grant.
Better yet would be an explanation
of how one's adversary mis-reasoned. But all I have is my negative
point. I
have not found any master fallacy behind impossibilism. Indeed,
their arguments strike me as careful deductions from attractive
premises. The impossibilists certainly do not have their pants
half down in the sense of being slovenly dialecticians. Their
premises are tenets from our most promising models of belief.
Their logic is state of the art. I will survey the impossibilist's
basic arguments and offer some criticisms. But I cannot specify
where each argument goes wrong. Some have me stumped. Nevertheless,
I know that all of the impossibilist's arguments
are unsound. For I have an infallible belief in the negation of
each argument's conclusion.
II. THE DISAPPEARANCE OF HIGHER ORDER APPEARANCES
Impossibilists grant that people often
act as if they believe impossibilities. A shopper will heatedly
assert that
4.39 + 9.84 = 13.23. A chess novice who only has a king and a
knight will attempt to checkmate his adversary's lone king. Experimenters
report that some subjects judge town A as northeast of town B
but B as not southwest of town A (Moar and Bower 1983). All philosophers
have felt the pain of contradiction. Brother, I've had my share!
Happily, reductio ad absurdum can't hurt me now. If an impossibilist
were to extract an absurd consequence
from my belief in believed impossibilities, then my belief would
be self-exemplifying. If I'm wrong, I'm right, therefore, I'm
right.
Impossibilists try to keep up appearances
by distinguishing between what people really believe and what
they believe they believe. For instance, the resemblance between
a contradiction and a consistent statement may lead us to mistake
the believer of the consistent statement as a believer in the
contradiction. But given impossibilism, believing someone believes
impossibilities should be a special case of believing an impossibility.
Nor can they save
appearances by scaling to the higher order hypothesis that we
only look as if we seem to believe impossibilities. For higher
order appearances would also be precluded by the impossibilist's
shackle of impossibility results:
(p)(~<>p > ~<>Bp > ~<>BBp > ~<>BBBp > ... ~<>Bnp).
What goes for thought, goes for talk; the chain of impossibilities precludes (even insincere) assertion of impossibilities. One can only assert what one can give the appearance of believing. Nor can the appearance of appearances be reclaimed by distinguishing between senses of `believe'. Any plausible reading of `believe' must permit disagreement over the believability of impossibilities.
Of course, one can stipulate a sense
of `believe' that entails logical omniscience. It's a free country.
But this
liberty invites the charge of misleading advertising. For example,
if the impossibilist has a technical sense of `believe' in which
it means `consistently believe', then his audience will protest
that they assumed he was using `believe' in an ordinary sense.
Another stratagem is to abandon impossibilism
about belief in favor of another propositional attitude that resembles
belief (Marcus 1990). However, an attitude can resemble belief
only if it allows for opposed attitudes to the same object. One
can only countenance this dissonance by tolerating disharmony
about the impossibility issue. The range of what is assumable
or presumable is at least as wide as the range of what is believable.
This difficulty will also haunt attempts to construct an interesting
sort of impossibilism with respect to a technical term such as
"acceptance".
One of the early impossibilists, George
Berkeley (1710, 273), paraphrases apparent belief in terms of
what we imagine we believe. However, like most impossibilists,
Berkeley also denies that we can imagine impossibilities. After
all, Berkeley subscribes to David Hume's (1739, 32) principle
equating conceivability with possibility. Hume's principle generates
its own nesting of impossibilities: if it is impossible to imagine
an impossibility, then it is also impossible to imagine imagining
an impossibility, and so on. Since imagination is the minimally
encumbered propositional attitude, the progression of unimaginables
is maximally constraining. Therefore, we
cannot appeal to propositional attitudes to explicate the impossibilist's
concession that we at least seem to believe
impossibilities. But if this explication fails, all fail. Hence
the impossibilist cannot afford to admit that we seem to believe
impossibilities. Even the illusion of disagreement is lethal!
Obviously, however, there is at least the appearance of disagreement.
Therefore, we should infer that it is possible to believe impossibilities
and should reject Hume's principle that conceivability implies
possibility. This also scotches the logical positivist's claim
that logical impossibilities can be sharply distinguished from
empirical impossibilities by means of the unthinkability of contradictions
(Schlick 1932/33: 42).
In one respect, the impossibilist
is a lonelier debater than the solipsist. Sure, the solipsist
has the same problem of
self-defeating disputation; if the solipsist admits that someone
else disagrees, the solipsist loses. However, the belief that
only I exist is compatible with it seeming as if other people
exist. So the solipsist can at least invent an imaginary friend
and have pretend debates about solipsism. The impossibilist is
also worse off than an eliminative materialist such as Paul Churchland
(1979). Those who believe there are no beliefs can sustain debate
by re-describing the controversy in a vocabulary free of "folk
psychology". Or at least these eliminativists could
issue the promissory note that future science will furnish this
redescription. The same promissory note can be waved at the problem
of higher order appearances. Just as one cannot prove to the eliminativist
that there are beliefs by appealing to the infallibiilty of belief
in `There are beliefs', one cannot prove to the eliminativist
that some impossibilities can be believed by appealing to the
infallibility of belief in `Some impossibilities can be believed'.
However, the impossibilist believes in belief; he restricts his
skepticism to belief in impossibilities. He
cannot emulate the eliminativist's indiscriminate rejection of
belief in the impossible. Therefore, the anti-impossibilist argument
is effective against its intended adversary.
III. CAN IMPOSSIBILISM BE RESTRAINED?
The impossibilist might try to salvage his thesis by restricting it to a special class of beliefs. Richard Purtill distinguishes between a strong and a weak sense of `believe'. "To believe in this weak sense is to do no more than to sincerely give it as one's opinion that p" (1970, 19). This sense is lenient enough to permit belief in impossibilities. The strong sense is intended to forbid belief in the impossible. An individual a "believes p in the strong sense if and only if a understands p, and a has some reason (which seems to him to be a good reason) for thinking p to be true." (20)
The essay you are now reading is jolly
good evidence that I understand my thesis, <>(Ep)(Ex)(~<>p
& Bxp), and have "some reason" for thinking it true.
So if Purtill applies his distinction to me in accordance with
his stated criteria, he should count me as a believer in the strong
sense. But then I will have achieved just what the distinction
was intended to thwart. My second objection is to the whole idea
of treating `belief' as having disparate senses. If `belief' is
as ambiguous as `bank', reflective equilibrium is impossible.
To consider the case at hand, just how do my strong beliefs interact
with my weak beliefs? A modest degree of epistemological holism
will require that I draw inferences from mixtures of strong and
weak beliefs. Yet applied logic demands an underlying unity in
the attitude adopted towards the premises. The Bayesian meets
this requirement by treating degrees of belief as commensurable
units. But Purtill has made no gesture towards finding a common
element between weak
and strong belief. If he were to opt for a reductionist strategy,
he would need to take strong belief as the fundamental attitude.
Anything less renders the deduction of further strong beliefs
intractable. But then the thinker will need to be able identify
which beliefs are strong beliefs. If these selectors must themselves
be strong beliefs, there will be a vicious infinite regress because
each strong belief must be backed with "some reason"
and this reason must be a strong belief.
The interaction problem does not arise
if the impossibilist merely distinguishes between species of belief.
(Then `belief' is general, not ambiguous.) For example, Curtis
Brown (1990: 283) maintains that one cannot directly believe an
impossibility. A direct belief holds in virtue of intrinsic facts
about the believer, not facts about his environment and society.
So one test for directness is to imagine whether an exact replica
of the believer must share the belief. Indirect beliefs can be
analyzed as those that arise from the interaction of direct beliefs
and
extrinsic facts. Since the believer has limited knowledge of external
conditions, he has an imperfect understanding
of his indirect beliefs. For example, ignorance of the fact that
the man called "Samuel Clemens" is identical to the
man called "Mark Twain" might lead a student to (consistently)
believe that the man called Twain is a better writer than the
man called Clemens. Under actual conditions, a referential convergence
generates the indirect belief in the impossibility that Mark Twain
is a better writer than Samuel Clemens. Since direct beliefs are
not vulnerable to these misfortunes, the object of direct beliefs
must always be possible. A parallel line is pursued by
Albert Casullo (1979). He defends Hume's principle that conceivability
implies possibility by distinguishing
between basic and derived conceivability.
Brown (1992) supplements the motivation
for the direct/indirect distinction by applying it to the belief
puzzles raised by Hilary Putnam, Tyler Burge, and Saul Kripke.
The guiding precedent is perception. Analysis of illusion forces
us to abandon naive realism in favor of representative realism.
Brown's search for "immediate objects of belief" is
reminiscent of the perception theorist's rummagings for sense
data (and the action theorist's quest for "basic actions").
As J. L. Austin's Sense and Sensibilia documented, the direct/indirect
distinction is a slippery
fish. So the historical record cannot be heartening to Professor
Brown.
But in addition to these general worries,
the direct/indirect distinction is particularly ill adapted to
the specimen
under scrutiny. My belief that impossibilities can be believed
enjoys a Cartesian aloofness from the extrinsic factors cited
by Brown. My Twin Earth doppelganger also believes that some impossibilities
can be believed. The object of my belief is a necessary truth
that only uses the concepts of belief and possibility.
Brown could reply that I only indirectly
believe that impossibilities can be believed. After all, Tyler
Burge
(1986) has shown that factors outside the speaker's psychology
affect the meanings of words such as `arthritis',
`sofa', and `contract'. Perhaps similar examples could be constructed
for `possible', `belief', and `there is'.
The first problem with this strategy is that Burgean cases always involve an omissive or commissive error about the term. It is implausible that every possible believer in `It is possible to believe the impossible' has a misconception about one of the terms in the sentence. All I need to prove my thesis is the possible existence of a single conceptually well-adapted believer in believable impossibilities.
The second problem with the strategy of expanding the domain of indirect beliefs is that no direct beliefs would remain. Even Descartes' cogito uses `I', `think', and `exists'. Notice that the Cartesian immediacy of my basic idea can be underscored by switching to a first person claim; `I can believe an impossibility' is just as infallibly believed as the logically weaker `Someone can believe an impossibility'. My thesis has as much independence from external conditions as the cogito.
IV. THE SOPHISTICATED PREFACE PARADOX
Our Cartesian meditations draw out
a resemblance between my thesis and the preface paradox. Prefaces
frequently contain apologies for the errors that are bound to
exist in the book. Although the author believes each of his assertions,
he does not believe the conjunction of these assertions. In addition
to refuting (Bp & Bq) > B(p & q), the preface paradox
shows that rational inconsistent belief is possible. For it is
impossible for a believer in
`At least one of my beliefs is false' to have all of his beliefs
turn out true.
This last step in the reasoning becomes problematic if we let the statement become self-referential. That is, instead of letting `At least one of my beliefs is false' cover just the beliefs in the text, let it also include itself. Now, if the beliefs in the text are all true, then `At least one of my beliefs is false' is true only if it is false and false only if true. There is kinship here with contingent variations of the liar paradox. Consider list A:
(A1) At least one of the statements on list A is false.
(A2) Bill Clinton is left-handed.
Since Clinton is indeed a lefty, the
other sentence on list A is true only if it is false. And if false,
it is true! The
paradox is contingent in the sense that (A1) would have been unproblematically
true if Clinton had been
right-handed.
Notice that my proof of the infallibility
of believing `It is possible that someone believes an impossibility'
includes
that statement in its own domain of discourse. After all, the
proof is a reduction ad absurdum which points out that false belief
in my thesis entails its truth. But what if all other possible
beliefs are beliefs in genuine possibilities? Then `It is possible
that someone believes an impossibility' would be true only if
it itself is an impossibility and hence false. And if false, it
would be true. Hence, if all other beliefs must be beliefs in
the possible, my thesis becomes as paradoxical as the contingent
liar paradox.
Fortunately, this enigma rests on a big `IF'. `All other possible beliefs are beliefs in genuine possibilities' is a necessary falsehood. So my thesis is no more paradoxical than the situation arising from list B:
(B1) At least one of the statements on list B is false.
(B2) Some triangles have six sides.
However, even those who believe that
impossibilities can be believed may object that I am begging the
question
against the impossibilist. After all, the impossibilist's thesis
entails that all other possible beliefs must be beliefs in
genuine possibilities. So even if my thesis is semantically unparadoxical,
I haven't demonstrated its probity to the
relevant audience.
This charge of question-begging is
questionable because it rests on reasoning that injures my adversary
more than myself. If the impossibilist thinks that the negation
of his thesis is as paradoxical as the liar then he should think
that his own thesis has the same flaw. For the "negation"
of a meaningless utterance is itself meaningless and so cannot
be affirmed in preference to its negation. Furthermore, if the
impossibilist accuses me of believing a paradoxical sentence,
then he contradicts his thesis that believing p implies the possibility
of p. His position would
also be inductively implausible. If I can manage to believe paradoxical
sentences, what stops me from believing a
necessarily false proposition? Impossible propositions at least
have a stable truth value. So the charge of question-begging can
only be sustained by representing the impossibilist as a kamikaze
dialectician whose only hope is to make his self-destruction jointly
damaging.
Mercifully, we can side-step a convoluted
controversy about circularity by mounting an alternative challenge
to
the paradox-monger's supposition that all other beliefs are beliefs
in the possible. In the course of my proof that
some people can believe the impossible, I first derived the lemma
that I can believe the impossible (line 4). This shows the inconsistency
of the supposition that all of my beliefs are beliefs in genuine
possibilities except for `It is possible for someone to believe
an impossibility'. The impossibilist will admit that if I believe
that I can believe an impossibility, others can believe that they
can believe impossibilities. Moreover, they can believe I can
believe impossibilities, I can believe they can believe that,
and so on. In short, the impossibilist should admit that if one
impossible belief is possible, then indefinitely many impossible
beliefs are possible.
V. TWO KINDS OF IMPOSSIBILITY?
Can the impossibilist avert dialectical
overkill by distinguishing between kinds of impossibility? He
might say that it is only metaphysically impossible to believe
a contradiction. `Something is red all over and blue all over'
seems impossible even though it is not a contradiction (because
`red' and `blue' are primitive terms). Many impossibilists follow
Saul Kripke (1972, 1980: 3) in thinking that the indiscernibility
of identicals is as "self
evident as the law of contradiction" and hence maintain that
false identity statements are necessarily false. Indeed, Kripke
thinks you cannot imagine that pain is c-fiber stimulation. He
says you only think you can imagine you were born from different
parents and that you only think you can imagine his lectern is
actually made of ice.
However, distinctions between kinds
of impossibilities will only be relevant if one reading opens
a possibility for
the believer that is not available for his object of belief. Contradictions
are stronger than metaphysical impossibilities in the sense that
whatever is contradictory is metaphysically impossible but not
vice versa. But this derivational difference does not reveal an
opportunity to believe an impossibility. If belief in an impossibility
is only metaphysically impossible, then there is still no possible
world in which someone believes an impossibility.
On the other hand, if belief in an
impossibility is only physically impossible, then there are other
possible worlds
with different physical laws that contain believers in impossibilities.
Although this would not secure an actual
opponent to impossibilism, the argument would open the possibility
that an actual person could seem opposed. Pseudo-adversaries could
also be recruited with appeals to biological or technical impossibility.
But none of these weakenings harmonize with the arguments deployed
by actual impossibilists.
VI. FOUR IMPOSSIBILIST ARGUMENTS
Antecedents of the main arguments
against relations with impossibilities can be found in Plato's
dialogues. In the
Theaetetus and Sophist it is argued that we cannot believe what
is false because what is false is not the case and what is not
the case does not exist. This is the absent relatum argument:
To believe is to believe something, so where there can be no object
of belief, there can be no belief. Subsequent philosophers flushed
out plausible relata for consistent false beliefs. However, they
still have the same problem with belief in the impossible. Causal
theories of representation, for example, take the object of belief
to be what causes it when the belief is formed under optimal conditions.
But there are no optimal conditions under which an impossibility
is true. Others take sets of possible worlds as the objects of
belief. However, there is no possible world in which an impossibility
holds. Sentences purporting to attribute belief in the impossible
appear to be meaningless danglers.
Plato's discussion of false belief
also touches upon the appeal to understanding. The human predilection
for viewing words as names led Plato to view all errors as mis-identifications.
For example, in a fog Socrates mistakes Theodorus as Theaetetus.
However, Plato realized that this mental switch model is implausible
because
no one is irrational enough to judge that Theodorus is Theaetetus.
As soon as the alleged object of belief becomes intelligible,
it becomes unbelievable.
As philosophers gradually freed themselves from the assumption that all words are names, they were able to focus on more tractable cases of false beliefs. Most beliefs use the `is' of predication rather than the `is' of identity. Generalizations really use variables rather than names of abstract entities. Names were increasingly viewed as minority members of language. Bertrand Russell's theory of descriptions showed how ordinary names might be analyzed as definite descriptions. W. V. Quine modified this theory and recommended that scientific language avoid all names in favor of definite descriptions. However, Kripke's attack on the description theory of names has dis-interred the Platonic problem of mis-identification.
Prominent philosophers of language
in the twentieth century have equated understanding with knowledge
of truth conditions (Heidelberger 1980). This link has strengthened
the reasoning against the possibility of believing contradictions.
Their opening premise is that one can only believe what one understands
and regards as true. Understanding requires knowledge of truth
conditions. A contradiction has no conditions under which it can
be true, so understanding a contradiction precludes belief. (Indeed,
understanding a contradiction automatically
produces disbelief.) Therefore, belief in a contradiction is itself
a contradiction (Foley 1986, 350). Although each
of the three premises have been challenged, the argument has considerable
influence.
To get the third argument against
the possibility of believing the impossible, the appeal to charity,
switch attention
from the interpretee's need for understanding to the interpreter's
need for understanding (Davidson 1967, 605). The point of assigning
beliefs and desires is to make sense of the agent's actions by
setting up applications of belief-desire psychology. If we ascribe
conflicting beliefs, then no predictive progress has been made.
(So even instrumentalists are hostile to belief in contradiction
(Dennett 1987, chapter 4).) Charity is forced upon the interpreter
because his project requires selection of the most rational portrayal
of the agent's attitudes. Therefore,
the appearance of incoherence is just a projection of the interpreter's
confusion. This confusion reflects a lack of
ingenuity on the part of the listener rather than the presence
of contradictory beliefs in the speaker. The interpreter is free
to admit that he is stumped. He is also free to doubt whether
the preconditions for interpretation obtain. (Perhaps some forms
of insanity amount to resistance to belief attribution.) However,
the
interpreter is never free to infer that the speaker is irrational.
More positively, "In our need to make him make sense, we
will try for a theory that finds him consistent, a believer of
truths, and a lover of the good (all by our own lights, it goes
without saying)" (Davidson 1970, 97).
Appeals to understanding frequently
fuse with the defeasibility of belief. There is some temptation
to think that we have immediate and infallible access to our own
beliefs. However, defeasibilists note that we often describe ourselves
as having been mistaken about our beliefs once an absurdity is
exposed. Indeed, magnanimous practitioners of reductio ad absurdum
describe themselves as showing that people do not really mean
what they
said. Patronizing? Not if thorough understanding is a precondition
of belief. Just as knowledge must be a relation with actual states
of affairs, belief must be a relation with possible states of
affairs (Marcus 1983, 324). One can then go on to construe `conceive'
as a success verb, and dismiss the possibility of imagining an
impossibility in the same way we dismiss the possibility of perceiving
an impossible situation (Hart 1988: 28).
Defeasibilists admit that there are
other defeaters of belief ascriptions. However, they under-represent
the wide
generality of the phenomenon. For the epicenter, go back to Plato's
dialogues. In Protagoras 358 and Meno 77-8, Socrates contends
that we cannot desire what is bad. If a man is unaware that something
is evil, then he might say he wants it and act as if he wants
it. But once he recognizes it as evil, he realizes that he really
never wanted it. Socrates concludes that virtue is knowledge.
Defeasibility undergirds paternalism
in medical ethics. A physician once illustrated how much baldness
bothers
men with an anecdote. A young patient kept insisting that there
must be some cure for his baldness. The exasperated physician
finally declared that the only way to stop baldness was castration.
The young man thought a bit. And said "Okay, whatever it
takes". But the story actually shows how little weight we
assign baldness. For our reaction is that the young man hasn't
thought the matter through; he doesn't understand what is entailed
by castration. The paternalist insists that the patient should
get the treatment he really wants -- not the treatment he
thinks he wants. What the patient really wants is determined by
what he would want if he had rationally considered all the relevant
information.
The political implications of defeasibility were taken up by some nineteenth century idealists. They reasoned that real freedom lies in the satisfaction of our ideally rational and well-informed desires. Thus the state was duty bound to ignore the apparent wishes of its citizens in order to make them free. Totalitarian Marxists incorporated this theme and concluded that they were the real democrats.
Ruth Marcus (1983) bases her impossibilism
on the principle that a belief attribution is defeated when it
is discovered that no state of affairs could make the belief true.
Just as knowledge requires truth, belief requires possibility.
In addition to applying this defeasibilism to reductios and false
identity statements, Marcus (1981) extends it to the Pierre puzzle.
Pierre is a monolingual Frenchman who hears that "Londres
est jolie" and so
believes it. He moves to a place he knows as London and begins
to learn English by immersion. Since he does not realize that
Londres is London and happens to be confined to its shabby parts,
he sincerely says that London is not pretty. So it appears that
Pierre has rationally acquired contradictory beliefs. For when
we translate his sincere French utterance and conjoin it with
his sincere English utterance, we obtain `London is pretty and
London is not pretty'. Marcus' solution is that the contradiction
cancels the belief reports. Upon learning that London is
Londres, Pierre would disavow his assertions about London.
Pierre may well recant after learning
of the impossibility. But if defeasibilist intuitions sufficed,
then a vitiatingly
broader program of re-description would be in order. For people
retract reports of beliefs and desires upon the discovery of almost
any kind of unwelcome consequence. (They also frequently just
admit error.) Some repudiations may be interpreted as a reaction
to absent relata. A biology student may retract his claim to have
believed that the largest single organism is a mushroom in Michigan
after learning that "the mushroom" is composed of clones
and so is only a borderline case of a `single organism'. Perhaps
a statement is borderline when there is no state of affairs that
would satisfy it or its negation. This might explain why we are
tempted to view vagueness as a kind of meaninglessness.
However, we also retract when the object of belief has defects other than non-existence. If a lecturer tells a student `Although mental events have bodily effects, you are an epiphenomenalist', the student might believe it on authority. Once the student learns that epiphenomenalists deny that causes run from mind to body, he can recognize the resemblance between the teacher's remark and the Moorean variant "It is raining but you do not believe it." So the student denies that he ever really believed the lecturer's remark but not on the grounds that there is no state of affairs under which the remark could be true.
Many flaws provoke disavowals: triviality, empirical bizarreness, heresy, rudeness, etc. When President Gerald Ford assured Americans in the midst of the cold war that Eastern Europe would never be dominated by the Soviet Union, his aides swiftly discounted the statement as a verbal slip. A central duty of Ronald Reagan's spokesman was to furnish reporters with official interpretations of presidential remarks. Unless we embrace the vintage real selves of the nineteenth century Idealists, we will need to rein in the implications of disavowals.
VII. PROPOSITIONAL GUISES OF THE PROBLEM
Impossibilists are inconsistent about inconsistency. Plato depicts Socrates as embarrassing his interlocutors by exposing hidden inconsistencies and thereby compelling them to change their minds. But if Socrates' adversaries never really had the conflicting beliefs, what are they embarrassed about? How could Socrates be the gadfly of Athens?
In the Meno, Plato partly agrees that
the Socratic method is uninformative. The concession is prompted
by the paradox of inquiry: If Socrates really knows nothing, how
can he hope to acquire knowledge by asking others? Recognition
of the correct answer requires some background knowledge. Socrates
responds by conceding that in one sense, learning is impossible.
Not because we know nothing, but because we know everything! In
particular,
his doctrine of recollection credits us with atrophied omniscience.
In a pre-existent state, we dwelt amongst the
forms and so knew all. The trauma of birth causes us to forget.
The role of Socratic interrogation is to revive these heavenly
memories.
The recollection model (minus the
fairy tale about pre-existence) has some prospect of accounting
for conflicting beliefs about contingent propositions. If I latently
believe p but become persuaded that ~p, then I might seem to believe
p & ~p. The problem of finding an object for this apparent
single belief can be resolved by analyzing it as two beliefs.
However, this divide and conquer strategy fails for belief that
there is a largest prime. This belief is not a conjunction of
opposed propositions. Thus the doctrine of recollection has no
advantage in explaining how inquiry about necessities can be informative.
This kind of cognitive advance cannot be reduced to the elimination
of one member of a conflicting pair of beliefs. (Indivisible contradictions
are also troublesome for Peter Strawson's (1952, 3 and 21) contention
that contradictory beliefs cancel each other out.)
The contemporary paradox of analysis
focuses on this problematic form of investigation. If a conceptual
analysis states exactly what the original statement says, then
the analysis is trivial. If it says something different from than
the original, then the analysis is mistaken. Hence, all analyses
are either trivial or false. The case for `Sameness of meaning
= triviality' meshes with the case against believing contradictions:
Since disbelieving p is just believing ~p, the incredibility of
contradictions precludes disbelief in tautologies. Could one be
an agnostic about a
tautology? If one understands a proposition, one understands the
conditions under which it is true and false. But in
the case of a tautology, this entails recognition that the proposition
is true and hence, knowledge that it is true.
Knowledge implies belief. Consequently, to understand a tautology
is to believe it. In addition to precluding agnosticism about
tautologies, the connection between understanding an analytic
truth and believing it "demonstrates" that conceptual
analysis is never informative.
The self-intimating nature of tautologies
would also preclude unwitting belief in them, that is, belief
unaccompanied by the recognition of their tautologous nature.
For example, some philosophers believe that `All causes precede
their effects' but are not sure whether the proposition is a tautology.
That would be impossible if belief requires understanding of the
nature of the proposition. Moreover, belief in a tautology would
always have
proper grounds. Fallacious proofs of mathematical theorems would
be impossible because anyone who understood the tautology would
automatically have cogent proof of it.
Contingencies would be partially self-intimating; although their truth-values could still be hidden, their modal status would always make itself known. For understanding a contingent proposition means understanding some conditions under which it is true and some conditions under which it is false. Moreover, contingencies would always look like contingencies. No one could believe they believe of a contingency that is non-contingent.
Propositions bear necessary relations
with other propositions, so understanding a proposition would
mean understanding what entails it and what it entails. But the
entailers and entailees must themselves be understood. This holistic
slippery slope is reminiscent of the idealist's arguments that
to know anything one must know everything. But it is even more
overbearing. At least the idealists could concede that we appear
to know some
things without knowing everything. The purveyor of logical omniscience
cannot admit that we seem to understand
some propositions without understanding their logical relations
with all others.
Some philosophers accept this dismal
conclusion and cite the failure of reductive programs as vindication.
But
most philosophers view the paradox of analysis as unsound or misleading.
For example, my impression had been that the paradox goes through
if one decides to set the high standard for `belief' implicit
in the principle that belief requires understanding. The background
idea is that `understand' is an absolute term like `flat'. Just
as a flat surface cannot be flatter than another flat surface,
someone who understands a proposition cannot understand it better
than another person who understands it. This follows from an absolutist
definition of `understand': A understands p iff A can discriminate
p from all other propositions. This definition quantifies over
propositions, thus a person who can discriminate p from foils
drawn from one domain of propositions might not be able to discriminate
p from the alternatives drawn from another set. When our quantifiers
are "wide open", there are so
many look-alikes that one understands very few propositions. As
with `flat', I concluded that although one can consistently relativize
`understand' to high standards, it is impractical to do so. Thus
I agreed that the paradox of analysis is correct for the limit
case in which we quantify over all propositions but not for the
ordinary cases. The philosopher's preoccupation with the paradox
could then be explained as an artifact of idealization. Usually,
we simplify a problem by considering extreme cases. But when it
comes to understanding the informativeness of
analysis, the strategy eliminates the resources needed for the
solution.
Although I still think that the paradox
of analysis is partly due to context shifting shenanigans, I now
think that there are semantic restrictions on how high one can
set the standards. To put the matter another way, the paradox
of analysis rests on two premises that are open to a context independent
refutation. Recall that the paradox proceeds by eliminating the
two ways in which an analysis might enlighten a thinker. The first
path is by
revealing that a belief is actually contradictory. The anti-analyst
blocks this possibility by appealing to the
principle that belief requires understanding. However, my belief
that "Belief does not require understanding" is infallible.
I could only be wrong if belief requires understanding. But then
I would understand "Belief does not require understanding"
and so know that there are no conditions under which it could
be true. This knowledge that the proposition is false would prevent
me from believing it -- contradicting the original assumption
that I
believed it. The proponent of "Belief requires understanding"
faces the same dialectical trap as his impossibilist
ancestors. If he concedes that I believe "Belief does not
require understanding", then he loses. Nor can he admit that
I believe that I believe my thesis, because my belief that I believe
"Belief does not require understanding" would itself
be a belief in a contradiction. Like the impossibilist, he must
deny even the appearance of disagreement. He must insist that
the infallibility claim, "If someone believes that belief
does not require
understanding, then he must be right", is only vacuously
true. The proponent of "Belief requires understanding"
must say the antecedent of the conditional expresses a logical
impossibility -- and he must say the same for all the higher order
infallibility theses. And he must say it for all contexts.
The paradox of analysis also attacks the second way analysis can be informative -- revealing tautologies to those who had not previously believed them. The anti-analyst attacks the possibility of being agnostic about tautologies by arguing that understanding a tautology entails believing it. However, "Understanding a tautology entails believing it" would then itself be a tautology. So if I understand the principle, then I must believe it. However, I understand the principle and do not believe it. Therefore, some tautologies can be understood without being believed.
This proof that one can understand
a tautology without believing it might be challenged on the grounds
that I only appear to understand "Understanding a tautology
entails believing it". The idea would be that I mistakenly
believe that I understand the principle. But this is impossible
given the anti-analyst's other principle that belief requires
understanding and his allegiance to principle that understanding
is divisional (one can understand a complex proposition only if
one understands its component propositions). For then my belief
that I understand the
proposition that p would be infallible. For `I believe that I
understand p' would entail `I understand I understand
p'. But I can only understand `I understand p' by understanding
all of the components of the complex proposition -- which includes
the p component.
VIII. DEDUCTIVE GUISES OF THE PROBLEM
The puzzles of deduction distend in similar fashion. If we cannot believe impossibilities, we cannot mistake valid arguments as invalid. To believe that an argument is invalid is to believe it is possible for the premises to be true while the conclusion is false. But if the argument is valid after all, this would be belief in an impossibility. The possibility of error in the opposite direction can be "refuted" by adapting the argument against agnosticism about tautologies. For uncertainty about the validity of an argument requires uncertainty about the status of the corresponding tautology. This challenge to the informativeness of deduction is best known in the form of John Stuart Mill's (1843, 120) thesis that all valid arguments are circular. As with tautologies, the informativeness of a deduction seems to evaporate under the light of reason. The more one can be said to understand, the more trivial the deduction appears.
Let's sharpen the deductive dilemma
with a slippery slope argument. For openers, nothing can be learned
from an
invalid deduction. And even some sound deductions are epistemically
impotent. To argue `Caesar's wife must be
above suspicion, therefore Caesar's wife must be above suspicion'
is to just run in place. Ditto for any argument of the form `P
therefore P'. Conjoining P with another proposition brings no
improvement: `P and Q, therefore, P' is just as question-begging.
Separate assertion of the premises does not matter (`P, Q, therefore,
P'), so an argument with truth functionally equivalent premises
only looks more persuasive: Q > P, Q, therefore, P. The
apparent persuasiveness can be increased by further tactics of
obfuscation such as rewording. But rearrangements of the same
information don't increase knowledge. Repackaging a circular argument
leaves it circular. However, a combination of these camouflage
techniques can take us from `P therefore P' to any valid argument
form. Therefore, all deductive reasoning is impotent. Or so runs
this self-defeatingly interesting
deduction.
Notice now that these implications can be expanded by ascent to higher order appearances: People cannot commit fallacies and cannot even appear to. People cannot be informed by deduction and cannot even appear to. This extra layer of commitment topples extremists into an abyss of absurdity.
IX. CLOSURE CONFLICTS WITH UNIVERSAL CONSISTENCY
In 1968, Leonard Linsky pointed out
that Jaakko Hintikka's (1962) model-theoretic account of belief
condemns as indefensible the belief that someone else is inconsistent.
Roughly, a proposition is doxastically indefensible if one cannot
consistently believe it even though it might be true. For example,
Moore's sentence `It is raining but I do not believe it' is logically
consistent but believing it commits the speaker to conflicting
beliefs. Thus Hintikka's system generates a weak analogue of impossibilism.
Instead of dismissing `Someone believes an impossibility' as a
contradiction, it condemns belief in that proposition as indefensible.
This implication might be viewed as awkward side-effect of using model theory. A model set cannot contain an inconsistency because it represents a (partial) description of a possible set of affairs. So a believer in `Someone else believes an inconsistency' cannot embed the object of belief in a model set. This inability of the belief to come out true is not the same thing as logical inconsistency. There is no way to make belief in `No one has a belief' come out true but that does not show that `Someone has a belief' is a necessary truth.
Not all view the result as a technical embarrassment. Some make a virtue out of necessity and defend the result as a bold discovery: although `Someone else believes an inconsistency' does not sound odd, it is a subtle Moorean sentence. Bolder still are the impossibilists who think that Hintikka's result is an understatement; instead of being merely indefensible, `Someone else believes an inconsistency' is a contradiction (Purtill 1970).
However, the infallibility of belief
in `Some impossibilities are believable' suggests that it is Linsky's
result that is
understated. Linsky's complaint is that Hintikka's system cannot
permit belief that someone else is inconsistent. The real problem
is that a model theoretic approach cannot require belief in the
believability of impossibilities. Hintikka's believers are supposed
to have deductively closed beliefs. Hence, every necessary truth
should appear within every belief system. But given the necessary
truth that `It is possible to believe an impossibility', each
believer must then think that a belief system can be inconsistent.
However, a model set is inherently consistent. Thus all of Hintikka's
ideal believers reject Hintikka's theory of belief.
The ingratitude generalizes. Any theory
that correlates belief and necessity will create Frankensteins
who repudiate the theorems of their creators. In Brian Ellis'
epistemic semantics, necessity is defined in terms of belief.
In particular, a proposition is possible if, and only if it is
believed in at least one rational belief system. A proposition
is necessary only when believed in every rational belief system.
So the necessary truth of `Some
impossibilities can be believed' demands its presence in every
system. Consequently, each system is committed to
the possibility of a rational belief system containing an impossibility.
But this is possible only if there is at least
one rational belief system containing an impossibility. Such a
rational belief system is inconsistent. Thus the attempt to reflect
all necessities by means ideal belief comes into conflict with
the requirement that all ideal belief systems be consistent.
The same conflict between closure
and universal consistency arises with the converse attempt to
analyze belief in terms of necessity. For example, Robert Stalnaker
(1984) takes the object of belief to be a subset of possible worlds.
When I believe my cat or my dog tipped a vase, I believe I am
in a possible world in which my cat or my dog tipped a vase. After
learning that the dog was outdoors when the vase broke, I narrow
down the range of possible worlds to those in which the cat is
the culprit. The possible worlds account neatly models contingent
beliefs by providing a clear contrast between the worlds I may
occupy and those I exclude. However, this contrast is not to be
found for necessary truths. Necessary truths hold in all possible
worlds, so they all have the same
content, namely, the set of all possible worlds. Since necessary
truths appear in every possible world, everybody believes all
of them. Stalnaker deftly defends deductive closure and the second
consequence of consistency. However, the details of this defense
do not protect his account from the objection used against epistemic
semantics. As a necessary truth, `Some impossibilities can be
believed' must appear in every possible world and hence every
belief state. But for an impossibility to be believed is to appear
in a belief state, that is, a possible world. But no possible
world contains an impossibility. Hence closure conflicts with
universal consistency.
Stalnaker does have a resource for
accommodating the appearance of inconsistent beliefs. He allows
a single agent to be in more than one belief state. That is, in
one complete belief state p holds and in another complete belief
state ~p holds. Each belief state is "a separate center of
rationality". Deduction is "the integration of the separate
belief states of a single agent" (Stalnaker 1984, 87). However,
this maneuver does not apply when the object of belief is a single
necessary truth. For `Some impossibilities are believable' must
be reflected within each and every belief state. There is no variation
in belief to integrate, so belief in the believability of impossibilities
does not
stimulate the search for deductive relief of internal dissonance.
Neighborly conflict between closure
and consistency should be distinguished from the self-referential
sort associated with Kurt Godel's incompleteness theorem. Roughly,
Godel showed that one can express `This statement cannot be proved
in this system' in a language strong enough to do arithmetic.
This creates an internal
conflict between consistency and completeness. The conflict precipitated
by `Some impossibilities are believable' is
between systems. For a belief system need not attribute belief
in an impossibility to itself. It can say that some other belief
systems contain impossibilities. The attribution pattern for inconsistency
could resemble one that some anthropologists allege for cannibalism;
all groups abominate some groups as cannibals but no group admits
to being cannibals. This self-serving finger-pointing carries
a price. An anti-cannibal group cannot say that its forbearance
is guaranteed by human nature. Likewise, a belief system that
abominates inconsistency cannot claim
that its own consistency is guaranteed by the nature of belief
systems.
Belief in cannibalism could be a cultural universal without there being any cannibals. Perhaps each group accuses other groups to increase local solidarity. However, those who link belief and necessity cannot hope that the believability of impossibilities is a universal myth. For they take the presence of a belief in every system to be sufficient for its truth, indeed, it must count as a necessary truth.
X. THE LIMITS OF ILLUSION
I have purposely reviewed the impossibilists'
arguments in broad historical strokes. The modern refinements
surpass Plato's first efforts. Nevertheless, they inherit a rare
disorder. Philosophers can typically help themselves to an appearance/reality
distinction. A Francis Herbert Bradley or a George Berkeley or
a Zeno can admit that their thesis appears false and then triumphantly
argue that it is, against all odds, really true. Indeed, much
of the value of speculative reasoning lies in the prospect of
dispelling an illusion. The preconditions for illusion are so
modest that they are commonly ignored. However, impossibilism
is an unprecedented case in which this minimal
condition of speculative debate is violated. Consequently, students
of belief are in a unique position to lay down an
adequacy condition with instructive connections to puzzles about
reasoning, representation and analysis: All discriminative accounts
of belief must permit belief in impossibilities.
* I salute some troopers on the bloopers: Jonathan Adler, Curtis Brown, John Carroll, Daniel Goldstick, and Carsten Hansen. Audiences at the CUNY Philosophy of Logic Seminar and the University of Rochester corrected meta-bloopers. All meta-meta-bloopers are mine.
REFERENCES
Altrichter, Ferenc (1985) "Belief and Possibility" Journal of Philosophy 82: 363-82.
Ambrose, Alice and Lazerowitz, Morris
(1984) "Assuming the Logically Impossible" Metaphilosophy
15/2:
91-99.
Berkeley, George (1710) Principles
of Human Knowledge (New York: Bobbs Merrill, 1970) ed. C. M.
Turbayne.
Brown, Curtis (1990) "How to Believe the Impossible" Philosophical Studies 58: 275-85.
____________ (1991) "Believing the Impossible" Synthese 89: 353-64.
___________ (1992) "Direct and Indirect Belief" Philosophy and Phenomenological Research 52, 2: 289-316.
Burge, Tyler (1986) "Intellectual
Norms and Foundations of Mind" The Journal of Philosophy
vol. LXXXIII
no. 12, 697-719.
Casullo, Albert (1979) "Reid and Mill on Hume's Maxim of Conceivability" Analysis 39, 4: 212-19.
Churchland, Paul (1979) Scientific Realism and Plasticity of Mind, New York: Cambridge University Press.
Davidson, Donald (1967) "Meaning
and Truth" Basic Topics in the Philosophy of Language ed.
Robert M.
Harnish, Englewood Cliffs, New Jersey: Prentice Hall, 1994: 598-614.
_______________ (1970) "Mental
Events" in L. Foster and J. Swanson eds. Experience and Theory
London:
Duckworth.
Dennett, Daniel (1987) The Intentional Stance Cambridge, MIT Press.
Ellis, Brian (1979) Rational Belief Systems Totawa, N. J.: Rowman and Littlefield.
Elugardo, Reinaldo (1986) "Marcus's
Puzzle about Belief-Attribution" Southern Journal of Philosophy
24:
201-18.
Foley, Richard (1986) "Is it
Possible to have Contradictory Beliefs?" Midwest Studies
in Philosophy 10,
Minneapolis, Minnesota: University of Minnesota Press: 327-55.
Goldstick, D. (1989) "When Inconsistency is Logically Impossible" Logique et Analyse 125-6: 139-142.
Hart, W. D. (1988) The Engines of the Soul Cambridge University Press.
Heidelberger, Herbert (1980) "Understanding
and Truth Conditions" Midwest Studies in Philosophy 5,
Minneapolis, Minnesota: University of Minnesota Press: 401-10.
Hintikka, Jaakko (1962) Knowledge and Belief Ithaca: Cornell University Press.
Hume, David (1739) A Treatise of Human Nature, ed. L. A. Selby-Bigge, Oxford: Clarendon Press, 1978.
Kripke, Saul (1972) Naming and Necessity Cambridge, Mass.: Harvard University Press, 1980.
Linsky, Leonard (1968) "On Interpreting Doxastic Logic" Journal of Philosophy 65: 500-2.
Marcus, Ruth Barcan (1981) "A
Proposed Solution to a Puzzle about Belief" Midwest Studies
in Philosophy 6:
501-10.
_________________ (1983) "Rationality and Believing the Impossible" The Journal of Philosophy 80: 321-38.
________________ (1990) "Some
Revisionary Proposals about Belief and Believing" Philosophy
and
Phenomenological Research vol. 1 Supplement 133-153.
Margalit, Avishai (1983) "What
the Tortoise Told Tertullian" How Many Questions? Essays
in Honor of Sidney
Morgenbesser ed. Leigh S. Cauman, Isaac Levi, Charles Parsons,
Robert Schwartz Indianapolis, Indiana:
Hackett: 195-207.
Moar, I. and Bower, G. H. 1983 "Inconsistency in Spatial Knowledge", Memory and Cognition 11: 107-113.
Mill, John Stuart (1843) A System of Logic London: Longmans, Green, and Co. 1919.
Purtill, Richard L. (1970) "Believing the Impossible" Ajatus 32, 18-24.
Schlick, Moritz (1932/33) "Positivism
and Realism" reprinted in The Philosophy of Science ed. Richard
Boyd,
Philip Gasper, J. D. Trout (Cambridge, Massachussetts: The MIT
Press, 1991): 37-55.
Stalnaker, Robert (1984) Inquiry Cambridge, Mass.: The MIT Press.
_________________ (1991) "The Problem of Logical Omniscience, I" Synthese 89: 425-440.
Strawson, Peter F. (1952) Introduction to Logical Theory New York: John Wiley & Sons.
Williams, John (1982) "Believing the Self-Contradictory" American Philosophical Quarterly 19: 279-285.
Zemach, E. M. (1986) "Can you avoid both inconsistency and conceit?" Metaphilosophy 17: 259-265.