Wright
on Truth and Superassertibility
Crispin Wright
argues persuasively that truth cannot be understood in terms of warranted
assertibility, on the basis of some very simple facts about negation. The argument, he claims, undermines not only
simply assertibility theories of truth, but more idealized ones according to
which truth is to be understood in terms of what is assertible in the long run,
or assertible within some ideal scientific theory. He nonetheless thinks that there can be an
assertibility-based understanding of truth that satisfies the minimalist
account of truth he defends. On the
minimalist account, there are two distinctive features about truth. The first is that any truth predicate must
satisfy the Disquotational Schema (DS) ““p” is true iff p” (and thereby satisfy
the related Equivalence Schema “It is true that p iff p”), and the second is
that to assert that p is to claim that p is true. Wright takes these claims to show that any
predicate satisfies the minimalist conception of truth so long as it is
“normatively coincident” with warranted assertibility, and yet potentially
different in extension from it. Wright
claims to have found in the concept of superassertibility an
assertibility-based truth predicate.
The argument
Wright develops against assertibility theories of truth is simple but
ingenious. We begin with the DS
(1) “p” is true
iff p
and introduce negations on either
side
(2) It is not the
case that “p” is true iff it is not the case that p.
(1) and (2) together yield
(3) “It is not
the case that p” is true iff it is not the case that “p” is true.
If we replace the truth predicate in
(3) with “is warrantedly assertible”, we encounter the problem of neutral
states of information. Neutral
states of information regarding a particular claim are states in which grounds
are lacking for asserting that claim or its denial. Such neutral states of information show that
the warranted assertibility analogue of (3)
(3a) “It is not
the case that p” is warrantedly assertible iff it is not the case that “p” is
warrantedly assertible,
is false. For if our state of information regarding p
is neutral, failure of entitlement to assert “p” does not generate entitlement
to assert “It is not the case that p”.
The possibility of neutral states of information thus argues forcibly
against warranted assertibility accounts of truth.[i]
But Wright does
not abandon hope for an assertibility-based predicate that can survive
this attack and satisfy the minimal constraints on a truth predicate. Besides the formal argument just given, he
notes two other difficulties that assertibility theories face, difficulties
that I will argue pose problems for his own theory. He agrees with the Reason, Truth, and
History animadversion of Putnam that assertibility theories cannot account
either for the stability of truth (warrant is something that comes and goes for
propositions, truth is not) or its absoluteness (warrant comes in degrees,
truth does not). But he wishes to avoid the
idealization move made by Putnam and Pierce of understanding truth in terms of
what inquiry would reveal at the (idealized) limit. Instead, he wants an assertibility-based
concept that is found instantiated by ordinary assertions. So he proposes the concept of
superassertibility as a truth predicate.
Superassertibility
is built out of warrant for assertion, or assertibility. He says, “We can ask whether an ordinary
carefully controlled investigation ... justifies the statement, and whether,
once justified, that statement continues to be so no matter how much further
information is accumulated” (p. 47). He
then defines superassertibility “more carefully”:
[A]nother
property constructible out of assertibility which is both absolute and, so it
is plausible to think, may not be lost, is the property of being justified by
some (in principle accessible) state of information and then remaining
justified no matter how that state of information might be enlarged upon or
improved. ... A statement is
superassertible, then, if and only if it is, or can be, warranted and some
warrant for it would survive arbitrarily close scrutiny of its pedigree and
arbitrarily extensive increments to or other forms of improvement of our
information. (pp. 47-8)
This formulation, as I’m sure Wright
is aware, could use (a lot) more precision, but I think the idea is clear
enough to avoid detaining us here.
Instead, I want first to look at his reasons for thinking that
superassertibility is a truth predicate.
We will see that the argument is sorely lacking. I will then turn to argue that it is not
simply that Wright lacks a good argument for his position. The truth of the matter is that Wright’s
position is simply untenable.
I.
Superassertibility and the Equivalence Schema
Wright’s argument is tortuous, but the
central moves in it are these. The
fundamental question Wright poses is whether superassertibility satisfies the
Equivalence Schema, which requires that
(Es)
It is superassertible that p if and only if p.
He first notes that (Es)
is adequate if and only if
(Fs)
It is true that it is superassertible that p iff it is true that p.
Wright worries that counterexamples
to (Es) and (Fs) are obvious and straightforward: “A
suitably chosen proposition (Goldbach’s Conjecture, say) may be undetectably
true, and hence not superassertible; and a suitably chosen superassertible
proposition (perhaps that we are brains-in-a vat) may be undetectably false”
(p. 51). His response to such
counterexamples appeals to his pluralism about truth predicates. Since minimalism about truth does not
guarantee that one and only one predicate could be a truth predicate, the
proper attitude for a minimalist about truth to take is that there might be a
multiplicity, a plurality, of truth predicates.
If so, however, there is a danger of evaluating a potential candidate
for a truth predicate in a question-begging way, by implicitly using some other
truth predicate in purported counterexamples to the instance of the Equivalence
Schema for the potential candidate.
Such, Wright thinks, occurs in the counterexamples above, which assume a
strongly realist truth predicate in order to cast doubt on the adequacy of
superassertibility as a truth predicate.
This response may
blunt the force of the counterexamples, though I will express my doubts
later. Even if the response is useful,
however, it gives us little help in determining how to evaluate whether any
potential truth predicate satisfies the Equivalence Schema. For that, we need something further. Here is Wright’s proposal. He claims that to avoid question-begging
counterexamples to (Fs) (and hence to (Es)), one’s only
recourse is to replace both occurrences of ‘true’ in it with the candidate
truth predicate ‘superassertible’. Thus
the proposal is that the question-begging difficulty can only be avoided by
showing the inadequacy of
(Gs)
It is superassertible that it is superassertible that p iff it is
superassertible that p.
For the generalization of (Gs)
(G) ΠΠp
iff Πp,
let us say that any operator
satisfying G is an idempotent operator.
Wright thus proposes that a positive answer to the suitability of
superassertibility as a truth predicate (for a particular discourse) can be
returned if, and only if, superassertibility is an idempotent operator.
In the next
section I will argue that this approach is thoroughly wrongheaded, but here I
wish to consider Wright’s argument that superassertibility is such an
operator. He claims that a sufficient
condition for superassertibility to be an idempotent operator is the
following: “to have warrant for P is to
have warrant for the claim that P is superassertible, and conversely” (p.
54). Though it takes an argument to show
that this condition is sufficient, I will grant the point here. I will also grant the converse claim in order
to focus on the more difficult, left-to-right, direction. His argument for this claim, the claim that
to have warrant for P is to have warrant for the claim that P is
superassertible, depends on three crucial data about justification or warrant:
(I) “What I am
warranted in believing I remain warranted in believing sine die unless I
acquire defeating collateral information.” (p. 55)
(II) “In
warrantedly believing any statement, P, a subject is thereby warranted in
believing that a sound investigation, to whatever extent one is possible, would
bear her out.” (p. 55).
(III) “The firm promise
of justification for what one has no reason to doubt is already justification.”
(p. 56)
For my purposes, the details of the
argument from these data to the claim that warrant for a claim is automatically
warrant that the claim is superassertible are not as important as the data
claims themselves. But some summary is
important, for any attack on the data claims may leave us wondering whether a
minor revision may still save Wright’s argument. So here is how the argument goes. Suppose that one has a warrant for p. By (I), any later time at which I investigate
the issue will be a time at which my warrant remains, as long as defeating
information was not acquired. And I
would be warranted at that future time even if my investigation at that time
were different, for by (II), I am warranted in believing that any further sound
investigation would bear me out. Such a
different investigation would, by my present lights, be merely an enlargement
of my present justification, since by (I), my present warrant will still be in
place. So I presently have warrant for
thinking that this enlargement would warrant what I believe. And by (III) the promise of warrant from this
enlarged state of information is present warrant. So, if one is warranted in believing p, one
has warrant for thinking that any enlargement or improvement of my state of
information will continue to warrant p, so long as no new information was
acquired before entering this improved state of information. This last qualifier is, Wright claims, all
that stands in the way of showing that warrant implies warrant for
superassertibility, and Wright thinks it can be eliminated as follows:
[I]t suffices to
reflect, presumably, that any I* prior to entering which I do acquire
p-relevant information additional to any possessed in I, must be the terminus
of a finite chain, áI, I2, ... , I*ñ, each
element of which satisfies the qualification with respect to its immediate
predecessor. (p. 56)
The idea, I take it, is that there
can’t be an infinite sequence of information states going from I* back to I
each of which is less of an enlargement on I than is I*, and yet each of which
involves a prior improvement of I. Given
this, we can conclude that warrant for p just is warrant for p’s being
superassertible, and hence that superassertibility is an idempotent operator.
The fundamental
problem with this argument is that all three of the crucial data statements on
which it relies are false, and no minor revision of the errors can salvage the
argument. Claim (I), that I remain
justified in believing what I am presently justified in believing in the
absence of acquired defeating information, is incompatible with important
cognitive limitations on memory from which we all suffer. Wait 10 years and see how much warrant you’d
have for thinking that you read this paper at this exact time, date, and place;
unless you’re miraculously different than I, you’ll have little warrant for
it. But it won’t be because you’ve
acquired defeating information; for that to be true, there would need to be
some further proposition you come to believe which, in conjunction with your
present evidence, no longer justifies your believe that you are now reading
this paper. No, the explanation is much
simpler: you’ll simply forget, and forgetting is a much different phenomenon
than learning something new that forces a change of rational opinion in you.
Wright may
attempt to avoid this problem by idealizing the epistemic agents to which he is
applying the ordinary concept of warrant.[ii] There are three ways I can think of to do
this. One is simply to restrict all of
his claims about warrant to claims about the kind of warrant that idealized
agents have. Such a move is a
non-starter, for it won’t allow Wright to infer that warrant for p is warrant
for the superassertibility of p.
Instead, it will only allow him to claim that for idealized agents, warrant
for p is warrant for the superassertibility of p.
A more
interesting maneuver would be to argue that the true nature of warrant is being
masked by the cognitive limitations or defects of ordinary human beings. So, once we idealize the epistemic agents a
bit, the true nature of warrant will shine through, showing that warrant will
always remain unless undermined by defeating information.
Such idealization
is problematic for Wright, since he makes much of wishing to reject the
idealization involved in Putnam’s and Pierce’s conceptions truth. By requiring similar idealization, Wright
will be guilty of the subterfuge found in false advertising.
There is a more
that can be said, however, about the claim that the true nature of warranted
belief is being masked by our cognitive limitations. It is certainly true that the warranted
beliefs of idealized agents have certain properties different from those of the
warranted beliefs of the more cognitively challenged. The question we must ask, though, is why
would one favor some of these properties as being central to the nature of
warranted belief? The property of
warrant which beliefs have is a relation between, among other things, the
belief in question and the person holding the belief. It is not as if warrant is a simple property
of a belief, with a nature that can be corrupted when the belief comes to
reside in limited cognitive beings. The
attempt here strikes me as an attempt to build on a chemical analogy, that
warrant in its pure form can have properties that it loses when appearing in
certain compounds. But the analogy won’t
work; warranted belief is not a simple element but rather a relation that
already has persons among the relata.
The final defense
of idealization appeals by analogy to the appropriateness of idealizations in
scientific theories. Certainly, it might
be claimed, some idealization in scientific theorizing is allowable, so why not
in philosophical theories? The answer
depends, I think, on the source of value for the idealization. Scientific theories involve allowable
idealization for several reasons: enhancing of theoretical virtues such as
simplicity or predictive utility, getting closer to the truth, and idealizing
when the anomalies between theory and reality can be explained in other
terms. But none of these features of
allowable scientific theorizing will be of much help in defending idealizations
regarding normative philosophical theories.
For example, some versions of Bayesian epistemology self-consciously
idealize epistemic agents to the point of logical omniscience. Such theories might give us insight into the
nature of rationality regarding some beliefs (say, empirical beliefs):
that is the “getting closer to the truth” defense of idealizations. But it won’t do for such theorists to also
assert that the theory also gives us insight into the nature of rational belief
regarding logical truths; that is to assert that the very feature idealized in
fact accords with reality. The same
point holds regarding Wright’s first datum regarding warrant. To defend the point by appealing to the
appropriateness of idealizations in science amounts to nothing more than
admitting that the datum holds only for the idealization and then hoping we’ll
be silly enough to let the claim go through that the datum holds also in
reality.
Claim (I) is
therefore false. Claim (II), that in
warrantedly believing something you are warranted in believing that a sound
investigation will bear you out, is equally faulty. Wright notes the holistic character of
warrant, and ventures the claim that virtually anything can constitute warrant
for anything else, given the right background information (p. 45), but seems to
have forgotten the point here. You might
have strong inductive evidence that too much investigation is
counterproductive; that further investigation, even sound investigation, nearly
always gets you in (alethic) trouble.
You might justifiably believe that inquiry in general should satisfy the
advice often given to test-takers to trust their first instincts because
additional, or different, methods of generating an answer are more
unreliable. If so, then your present
warrant will also be warrant that a sound investigation won’t bear you
out.
Perhaps Wright
would try to salvage claim (II) by maintaining that more is packed into the
notion of soundness here than I am granting.
My assumption above was that careful and deliberate investigations by
competent inquirers are sound investigations.
But perhaps Wright would claim that the concept is more restrictive, so
that the idea of claim (II) is that your present warrant also warrants you in
thinking that any investigation that gets you to the truth (i.e., a
sound investigation) will be one that gets you to the same conclusion that is
presently warranted for you. Such a
restriction on what a sound investigation undermines the argument in which
claim (II) is used, however. For in that
argument, the warrant which a sound investigation produces needs to be
presently projected to be an enlargement of a present
justification. But if a sound
investigation is one that gets one to the truth, there will be no guarantee
that such an investigation (different from one’s actual investigation) is such
an enlargement. In fact, the very
counterexample I constructed to claim (II) in the last paragraph undermines the
enlargement claim on this understanding of soundness. For a person might be correct that no
expanded investigation would get one to the truth, and so that any
investigation that does get to the truth would fail to be an enlargement of
present justification. So no version of
claim (II) can play the role needed in Wright’s argument for the idempotence of
superassertibility.
Claim (III) is also
mistaken. Recent epistemology has
devoted some attention to Ulysses problems in epistemology.[iii] The generic Ulysses problem arises out of the
story of Ulysses and the Sirens, in which Ulysses knows today that he will be
strongly inclined to act irrationally tomorrow, and hence has his crew fasten him
to the mast. Applied to epistemology,
cases arise where we presently know that our cognitive equipment will mislead
us tomorrow. Perhaps we have just taken
a drug that will have its effect tomorrow; perhaps we merely know that in many
years our cognitive equipment will deteriorate so that it will be warranted
then for us, say, that we are not very good at constructing philosophical
arguments. All of these cases are cases
where we have a firm promise of a justification in the future for what we now
know is false (I’m optimistic here on our philosophical behalf).
Wright’s claim
(III) does not succumb to such counterexamples, for his claim includes the
proviso that the promise of a future justification is only a present
justification when one has no reason to doubt the claim. But once one sees the recipe for Ulysses’
problems in epistemology, it is easy to see how to generate counterexamples to
Wright’s more restrictive claim as well.
Suppose the effect of the drug will be to make me believe that there are
an even number of grains of sand on Galveston beach, and will further present
me with strong evidence (which I now recognize would be misleading, but won’t
recognize tomorrow) that this is so.
Since I have no information regarding the evenness of the number of
grains of sand on the beach, it would seem to count as one I have no reason to
doubt. But the promise of justification
tomorrow simply does not constitute present justification.
One might think
this counterexample hinges too much on a special construal of reasons for
doubt. If this reply is taken, however,
one will need a very strong construal of reasons to doubt. For the same kinds of counterexamples can
arise for any claim that you presently are less than justified in
believing. So, for the sake of argument,
assume that you always have reason to doubt everything except what you are
justified in believing. Then principle
(III) claims that the promise of future justification for what you are
presently justified in believing constitutes present justification. On one reading, this claim is perfectly
trivial, for having a promise of future justification for what you are
presently justified in believing entails, in a trivial way, that you
have present justification. So
counterexamples to this version of claim (III) will not be forthcoming. But note that the promise of future
justification here does absolutely no work whatsoever, and so provides a
useless clause in a trivial principle.
Perhaps Wright
would insist that the claim is not trivial, since claim (III) asserts that it
is the promise of future justification that constitutes present
justification, even if there is other present justification available as
well. Here we must heed the lesson of
Ulysses. The lesson of Ulysses problems
is simply and directly this: what you’ll warrantedly believe tomorrow simply
has no implications for what kinds of warrant you have today. Ulysses problems show that you can know that
tomorrow you will warrantedly believe a claim without the promise of future
justification constituting any evidence whatsoever in the present for that
claim.
So claim (III) is
undermined by Ulysses problems, unless it is reformulated as a trivial
principle amounting to the claim that if you have a promise of a future
justification for what you presently believe, then you have present
justification. Since the promise of
future justification plays no rule in accounting for the truth of this
principle, Wright will not be able to use this version of claim (III) to argue
that warrant for p is warrant for the superassertibility of p, for that
argument requires an implication from the promise of future justification to
present justification.
So Wright’s
argument for the claim that superassertibility yields an idempotent operator is
a complete failure. This point alone
constitutes a serious problem for Wright’s project, but we have yet to see the
deepest difficulties for the project. I
will argue in the next section that Wright’s pluralism about truth predicates
and his worry about begging the question against potential truth predicates
makes it impossible for any formal features of truth, such as the idempotence
of the truth operator, to be used to show that a predicate is in fact a truth
predicate.
II.
Idempotent Operators and Formal Features of Truth
Wright
argues that the plurality of potential truth predicates creates a danger of
begging the question against a potential truth predicate by constructing
counterexamples which implicitly rely on another truth predicate. He then claims that the only way to avoid this danger of begging the question
is to focus on completely formal features of the concept of truth when
evaluating a potential truth predicate.
In particular, the capacity for a concept to be a truth predicate turns
on the question of whether that concept, when employed as an operator, is
idempotent. This last claim, I want to
argue is clearly false.
First, this
proposal will let in vastly too many predicates as truth predicates, for there
are many idempotent operators. Consider
the following analogues of (Es), (Fs), and (Gs):
Example 1: Possibility
(Eà)
It is possible that p if and only if p.
(Fà)
It is true that it is possible that p iff it is true that p.
(Gà)
It is possible that it is possible that p iff it is possible that p.
Example 2: Necessity
(Eo)
It is necessary that p if and only if p.
(Fo)
It is true that it is necessary that p iff it is true that p.
(Go)
It is necessary that it is necessary that p iff it is necessary that p.
Example 3: Hintikka-Knowledge (i.e.,
that kind of knowledge an operator for which satisfies Hintikka’s epistemic
logic; in particular, it satisfies the KK-thesis)[iv]
(EK)
It is H-known that p if and only if p.
(FK)
It is true that it is H-known that p iff it is true that p
(GK)
It is H-known that it is H-known that p iff it is H-known that p.
In each case, we can mimic Wright’s
argument to show that possibility, necessity, and H-knowledge are all truth
predicates. The crucial question in each
case is to ask whether the particular versions of the Equivalence Schema ((Eà),
(Eo), and (EK)) are correct,
and in each case there are obvious counterexamples. But, wait!
Such counterexamples may appear to succeed by begging the question on
the plurality of potential truth predicates.
They succeed by presenting a counterexample to (Fà),
(Fo) or (FK) that relies on
some other truth predicate, and such an argument ignores the plurality
of truth predicates that is a crucial part of Wright’s minimalism. So one can’t approach the question of satisfaction
of some version of the Equivalence Schema directly via the method of
counterexample, but must instead focus on the adequacy of the appropriate
instances of (G). Notice, however, that
possibility, necessity, and H-knowledge are all idempotent operators, so
Wright’s argument establishes them as truth predicates. Clearly, however, none of these are truth
predicates, so Wright’s argument must be fallacious.
Perhaps Wright is
correct about the question-begging worry, and right that a focus on formal
properties of proposed truth-predicates is the proper way to proceed. If so, the mistake above is that Wright has
focused on the wrong formal properties of the truth operator. Instead of focusing on the idempotence of the
truth operator, perhaps there are other properties of it that delineate truth
predicates from the counterexamples above.
Here is one idea[v]:
the truth operator is one that commutes with any other operator, and in
conjunction with the Equivalence Schema, renders truth “transparent” to any
operator. If we let ‘T’ be a truth
operator and ‘Δ’ be a variable for operators, then T is a commutative
operator if and only if it satisfies principle C:
C: TΔp if
and only if ΔTp.
In addition, T is transparent with
respect to any Δ just in case
Tr: ΔTp if and only if Δp.
C licenses the inference from, e.g., it
ought to be the case that it is true that p to it is true that it ought
to be the case that p, and from it is known that p is true to it
is true that it is known that p. And
using Tr, we can infer from the first premise that it ought to be the case that
p, and from the second premise that it is known that p. In these cases, at least, the account
licenses valid inferences.
This account of
some distinctive formal features of truth helps explain away the earlier
counterexamples to Wright’s emphasis on the idempotence of the truth
operator. If we replace ‘T’ in C with
any of à, o, or K, counterexamples can be
constructed. From the fact that it ought
to be the case that p is known, it does not follow that it is known that it
ought to be the case that p; from the fact that it is desired that p is
necessary, it does not follow that it is necessary that p is desired; and from
the fact that it is known that p is possible, it does not follow that it is
possible that p is known.[vi] So the formal account of truth that focuses
on commutativity and transparency seems intuitively plausible and also avoids
the counterexamples to Wright’s own proposal.
Unfortunately,
this emphasis on commutativity is devastating to Wright’s attempt to defend the
claim that superassertibility is a truth predicate. For if we substitute the superassertibility
operator S for T in C, counterexamples abound.
Suppose it is superassertible that p is thought; it can still be the case
that it is not thought that p is superassertible. Suppose it is desired that p is
superassertible; it does not follow that it is superassertible that p is
desired. So even if the commutativity
proposal can withstand scrutiny, it won’t be of any use to Wright in defending
that superassertibility is a truth predicate.
Not only will it do him no good, the proposal is devastating to the
project.
In fact, it is
easy to see on textual grounds that Wright would reject the commutativity
approach. For his argument toward the
idempotence approach claims that begging the question is a concern whenever a
truth predicate remains in the formal principle by which we must evaluate a
proposed truth predicate. When a truth
predicate remains, the possibility exists, according to Wright, of implicitly
using some other truth predicate to evaluate the candidate truth
predicate. So, whatever formal approach
we take, Wright would argue, it must be purged completely of truth predicates,
and the proposal above does not satisfy that constraint.
Wright’s own
proposal does satisfy that constraint, but turns out to be unacceptable, as
I’ve already argued. Perhaps there is
some other proposal that satisfies the above constraint, but if there is, it is
very hard to see what it is. In the face
of the lack of obvious alternatives, I’m inclined to think that the worry about
begging the question has been taken too seriously. It has been taken so seriously that there is
simply no way of avoiding the worry without adopting an unacceptable formal
approach to whether a predicate can be a truth predicate.
So, perhaps
Wright’s constraints on a formal approach should be ignored, leaving open the
possibility of accepting the commutativity proposal. Though this proposal has some advantages over
the idempotence proposal, we should not yet adopt it. For there are problems of intentionality that
plague it. Suppose the following states
of affairs are distinct: p’s being the case and “p’s-being-true”’s
being the case. If so, it would seem
to be possible to be thinking about one of those states of affairs and not
thinking about the other. Yet, if that
is possible, then principle C is false.
So, the only way for principle C to survive is if the state represented
by p is precisely the same as the state represented by Tp. To affirm that, however, is on the precipice
of affirming the redundancy of truth with a vengeance; it flirts with holding
that the word ‘true’ and the phrase ‘it is true that’ or the predicate ‘is
true’ plays no more semantic role than a hiccup. To avoid that conclusion, one would need some
semantical account on which the meaning of ‘true’ is divorced significantly
from the null metaphysical contribution of truth to any state of affairs, and
though such theories are surely possible, it is hard to see why one would want
such a theory here. Without it, though,
the result is the complete redundancy of truth.
That strikes me,
at any rate, as too high a price to pay to save principle C, but I won’t argue
the point here. If I’m right, one will
either have to resort to positing more epicycles in order to save the
appearance of the formal approach begun by Wright’s idempotence proposal but
more readily pursued through the commutativity proposal, or one will have to
reject altogether the formal approach to the nature of truth. On both alternatives, counterexamples abound
to the claim that superassertibility is, or can be, a truth predicate, and
Wright’s attempts to forestall the failure of his proposal simply do not work.
I think that more
can be learned, however. For Wright’s
argument that superassertibility is an idempotent operator is a failure as
well, which I will argue in the next section.
III.
Why Superassertibility Can’t be a Truth predicate
What
we have learned is that Wright’s argument that superassertibility is a truth
predicate is seriously defective; in fact, it is hard to find a serious and
important piece of philosophical argumentation that has more wrong with
it. Apart from the defects of
argumentation, however, what should we think of the conclusion? Is the claim that superassertibility is a
truth predicate one that might still be correct? No; the fundamental reason that
superassertibility cannot be a truth predicate has to do with facts about
defeasibility and the susceptibility of warrant to it.
If we return to
the motivation for introducing superassertibility, we find Wright talking about
Putnam’s account of truth in terms of ideal warranted assertibility, warranted
assertibility under epistemically ideal circumstances. Putnam had rejected the idea that truth is
warranted assertibility, on grounds that warranted assertibility is neither
stable (it comes and goes, whereas truth does not) nor absolute (warrant comes
in degrees, whereas truth does not). In
order to account for stability and absoluteness, Wright notes that these
requirements “can be fulfilled only if to have warrant for a proposition under
“epistemically ideal circumstances” involves have a case for it which cannot
be defeated (else we wouldn’t have stability) or improved (else we
wouldn’t have absoluteness) by any further information.” (p. 45) Wright then goes on to characterize what he
doesn’t like about Putnam’s idealization, and how he thinks an alternative can
be developed. He says,
Warranted
assertibility is assertibility relative to a state of information. That can make it seem as if there is only one
direction for a truth-like idealization of assertibility to assume: to wit, we
have to idealise the state of information involved. And then the quest for stability and
absoluteness seems to ensure that a Piercean notion will be the result. But that is an oversight. Rather than ask whether a statement would be
justified at the limit of ideal empirical investigation, ... we can ask whether
an ordinary carefully controlled investigation, in advance of attaining any
mythical limit, justifies the statement, and whether, once justified, that
statement continues to be so no matter how much further information is
accumulated. (p. 47)
Wright here fancies that there is a
way to build a truth-like notion of assertibility without idealizing to the
Piercean limit in the way Putnam does.
Such a notion, Wright hopes, can be built out of ordinary warrants for
assertion. Some such warrants, developed
out of “ordinary carefully controlled investigation,” attach to claims that
will continue to be justified even in the face of further learning. Such warrants are still warrants for
assertion, and the resulting superassertibility of a statement thus arises out of
ordinary assertibility.
There are two
fundamental difficulties with this notion being a truth predicate. The first is that it is incompatible with the
obvious truth that knowledge implies truth, and it fails to handles Putnam’s
demand that truth be both stable and absolute.
On the first point, note that knowledge can be present even if it will
be soon lost. This is so because some
defeaters are misleading defeaters, a point well-explored in the
archives of the Gettier problem.[vii] Wright recognizes something of this
phenomenon when he says,
I do not deny
that, in suitable circumstances, an agent may know something on the basis of
information which can in fact be defeated.
But if his knowledge claim is not to be undermined by the availability
of such defeating information, it is surely required that the negative effect
of that information, once acquired, could itself be stably overturned. (p. 58)
The significance of this remark may
be overlooked if one focuses too much on the cross-temporal feature of the
example, signalled by the phrase “once acquired”. By focusing on that feature, Wright’s point
is that if d defeats S’s warrant for p but S
nonetheless knows p, then should S come to learn d, there
must be something else S could also come to learn that would stably
overturn the defeating effect of d.
That is true, but it has nothing to do with diachronic features of the
case. It is rather built into what it
means for S’s warrant to be defeated by d and S know p
nonetheless. For if S knows p
in spite of there being such a defeater, then there must be some overrider
present to account for how S knows p in spite of the
defeater. Such cases are quite
common. Suppose you see your friend Tom
steal a book from the library, and know that he did as a result. These facts do not preclude the possibility
of defeating information. For example,
Tom’s mother may be telling the police right now that it wasn’t Tom, but rather
his identical twin brother Tim who stole the book. The police don’t take the claim seriously;
they don’t need to. Tom has a long
record, and his overly-protective mother is known to be willing to say anything
to try to get him cleared of any charge.
So there is defeating information (what Tom’s mother tells the police),
but that information doesn’t undermine your knowledge because of the
co-presence of the overrider (his mother is a pathological liar who will say
anything to protect her son).[viii]
Here is the
relevance of such cases to Wright’s claim that superassertibility is, or can
be, a truth predicate. Tom’s mother may
be pathological enough that she does her best to convince everyone who saw Tom
steal the book that Tom didn’t do it.
So, tomorrow, she will show up at your door and tell you the story she
told the police. Not knowing what the
police know, your warrant is undermined and you no longer know that Tom stole
the book, even though there is further information you could acquire that would
restore your knowledge. So today you
know that Tom stole the book, tomorrow you won’t, even though you might find
out further information that would reinstate your knowledge.
Such a case
involves a warrant that is presently undefeated, but does not constitute a
warrant sufficient for superassertibility.
So you presently know that Tom stole the book, but it is not
superassertible that Tom stole the book (not for you, and perhaps not for
anyone, since we could construct the case so that you and only you have the
warrant in question). But if
superassertibility is a truth predicate, then what we have is a case where you
know that Tom stole the book, but it isn’t true that Tom stole the book.
Wright could
respond here as he does in the text to other difficulties with the claim that
superassertibility is a truth predicate.
Namely, he could claim that I’m begging the question against his
position by ignoring the plurality of possible truth predicates. He could do that, but this case is much more
difficult to address than the cases he considers in the text. For those cases all involve claims that are
imagined to be true, but undiscoverably so, leading Wright to claim that they
implicitly use a strongly realist notion of truth. Nothing of the sort occurs in the above
case. The case does not involve
undiscoverable truths, and it is not even essential to the case that claims
about the future have truth values now, even though I described the case that
way. Furthermore, as we saw earlier,
Wright’s way of rescuing his position in the face of counterexamples is thoroughly
unacceptable, so there is little rhetorical advantage to be gained here by
repeating discredited arguments.
So
superassertibility is not a truth predicate, because if it were one, knowledge
would not imply truth. Moreover,
superassertibility cannot be a truth predicate because it cannot account for
Putnam’s requirement that truth is both stable and absolute. A first point to note is that Wright never
once addresses the question of whether his alternative to Putnam’s idealization
can preserve both the stability and absoluteness of truth. To be sure, Wright goes to great efforts to
fashion superassertibility so that it is stable--that is what all the talk of
permanence in the face of enlargement and improvement is meant to accomplish. But he never takes on the question of
absoluteness at all; what he does instead is merely to assert that
“superassertibility is, plausibly, both absolute and stable.” (p. 49)
But things get
worse. For Wright not only fails to
argue for the absoluteness feature, he points out unequivocally that it fails
for superassertibility: “the superassertibility of a sentence carries no
implication about the strength of the available evidence, which though
positive may be enduringly weak.” (p. 57) What is amazing about this remark is
that Wright seems not to have noticed that he just gave away the store. The remark is simply an admission that the
warrant involved in superassertibility comes in degrees, unlike the kind of
warrant claimed by Putnam to arise in the idealized limit of inquiry. Thus, if warranted assertibility is
problematic because it cannot explain the absoluteness of truth, so is
superassertibility.
The path to
missing this point goes through a confusion about two related concepts
involving the notion of defeat. One is
the concept of indefeasible justification, the kind of justification one has
that cannot be undermined by any further information whatsoever and cannot be
enhanced by additional learning.[ix] The other concept is the concept of
undefeated justification, the kind of justification one has when no defeaters
are present (or where any defeaters that are present are overridden by further
information). By confusing the two, one
can attribute to superassertibility, which includes the notion of (a kind of)
undefeated rather than indefeasible warrant, the kind of absoluteness that
comes only with indefeasible warrant.
What Putnam hopes to get by idealizing to the Piercean limit is
indefeasible justification, justification which cannot be defeated nor improved. Wright simply assumes without argument that
his weaker notion has the same virtues, contrary to the facts.
The lesson is
that if Wright wants superassertibility to be absolute, he’ll have to
understand it in terms of indefeasible justification, for only that kind of
justification is immune from increased strength due to further learning. But if we define superassertibility in terms
of indefeasible justification, first we lose Wright’s claim that it is merely a
special kind of, or constructible out of, ordinary justification. Wright hangs the uniqueness of his own
account as compared with Putnam’s on the fact that Putnam idealizes to
justification possessed only at the limit of inquiry, whereas his notion of
superassertibility arises in a mundane way as merely a special case of ordinary
assertibility. But the only kinds of
warrant underlying ordinary assertibility or belief are essentially defeasible
ones--that is the lesson of the history of epistemology. A further implication of this lesson is that
if we identify truth with what has an (unidealized) indefeasible justification,
there are going to be very few, if any, truths.
For all these reasons, the conclusion that must be drawn is that
superassertibility simply cannot be a truth predicate.
Notes
[i]. I do not here assert the soundness
of Wright’s argument, but only to its prima facie appeal. For rejoinders, see Neal Tennant, ???, and
Michael Hand, ????.
[ii].Wright says as much in the
discussion notes to the chapter.
[iii].Van Fraassen and Foley, in
particular.
[iv].Don’t worry that perhaps no human
knowledge counts as Hintikka-knowledge.
The most that would be needed here is the coherence of the concept, and
if an omniscient being is possible, that shows that the concept is coherent.
[v]. I owe this point to Michael Hand, from
discussion with whom I have benefited immensely.
[vi].Bracketing, of course, anti-realist
conceptions of truth according to which, for any proposition whatsoever, it is
possible that it is known.
[vii].For the most sustained and detailed
attempt to understand the nature of defeat and its relationship to knowledge,
see Peter Klein, Certainty: A Refutation of Skepticism, (Minneapolis,
1981).
[viii].This example is drawn from Keith
Lehrer and Thomas D. Paxson, Jr., “Knowledge: Undefeated Justified True
Belief,” pp. 146-154 in George S. Pappas and Marshall Swain, eds., Essays on
Knowledge and Justification, (Ithaca, 1978).
[ix].I use the term ‘indefeasible’ here
so that it entails truth, and that is why it cannot be enhanced by further
learning. In order to achieve this
result definitionally, we need only countenance the falsity of the belief as
itself a possible defeater.