Akratic Believing?

Author(s): Jonathan E. Adler

Source: Philosophical Studies: An International Journal for Philosophy in the Analytic
Tradition, Vol. 110, No. 1 (Jul., 2002), pp. 1-27
Published by: Springer
Stable URL: http://www.jstor.org/stable/4321282 .
Accessed: 31/12/2010 08:17

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=springer. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Philosophical Studies: An
International Journal for Philosophy in the Analytic Tradition.

http://www.jstor.org
 JONATHANE. ADLER

AKRATICBELIEVING?*

(Receivedin revisedversion 1 March2002)

ABSTRACT. Davidson's account of weakness of will depends upon a parallel

thathe drawsbetween practicalandtheoreticalreasoning.I arguethatthe parallel
generates a misleading picture of theoretical reasoning. Once the misleading
picture is corrected, I conclude that the attempt to model akratic belief on
Davidson's account of akraticaction cannot work. The argumentsthat deny the
possibility of akraticbelief also undermine,more generally,various attemptsto
assimilatetheoreticalto practicalreasoning.

1.

Davidson's ingenious account of the possibility of weakness of

will depends upon a parallel that he draws between practicaland
theoretical reasoning (Davidson, 1982). I argue that the parallel
generates a misleadingpicture of the goal of theoreticalreasoning
(I follow standardterminology first, in using "weakness of will"
and "akrasia",interchangeably;and second, in calling reasoning
to belief "theoreticalreasoning").Once the misleading picture is
corrected,I conclude that the attemptto model weakness of belief
on Davidson'saccountof weakness of will cannotwork.1
The only authorthat I know of to draw a similar conclusion is
Susan Hurley.In her Natural Reasons, she argues that "there'sno
such thing as evidentialakrasia"(1989, p. 131). On the main points
I am in agreementwith Hurley, though I come at this conclusion
from a differentroute. Hurley'srouteis from the third-personpoint
of view of interpretation(1989, p. 159) and a distinctionbetween
primafacie andpro tantoreasons.My routeis primarilyfrom a first-
person point of view which does not appeal to the primafacie/pro
tantodistinction.
The argumentsthat deny the possibility of akratic belief also
undermine,more generally, various attemptsto assimilate theore-
tical to practical reasoning. However, I develop this theme only,
LA Philosophical Studies 110: 1-27, 2002.
02002 KluwerAcademicPublishers. Printed in the Netherlands.
O
2 JONATHANE. ADLER

roughly,to this extent:whateverthe similaritiesbetween theoretical

andpracticalreasoning,thereis a fundamentaldivide between them
stemming from theoretical reasoning's singular aim of an all-out
judgmentof truth.

2.

Davidson understandsweakness of will as acting or intending to

act against one's better(or best) judgment.I will largely follow his
understanding.2In a typicalcase, the akraticagent- me -judges, all
things considered,thatit is worse to eat a brownieice cream sundae
than to refrain, yet I (deliberately)eat it anyhow. By parity, and
with putativeillustrationsto follow, akraticbelief would be belief
in (recognized)defianceof one's evidence.3
To his credit, Davidson's account preserves akrasiaas motiva-
tionally intelligible, but (first-personally)irrational,4yet the agent
performs it deliberately. There is an effective reason (desire) that
renders intelligible, or makes sense of, why someone would be
motivatedto eat the sundae instead of refraining.The anticipation
of the pleasureof eating the sundaeoverwhelms,withoutreversing,
good judgment.Its causal efficacy is strong,even thoughits rational
efficacy is used up and comparativelyweak. Nevertheless,the action
or intentionis "surd",Davidson says because, presumably,the anti-
cipatedpleasurealreadyentersinto the evaluationof what the agent
should do. So, in being moved by it anyhow,the agent is "double-
counting".The agent deliberatelyacts upon a considerationthat he
has alreadytakenaccountof - used up - in the fuller evaluationthat
his action knowingly defies.5
The key step in Davidson'sargumentdrawsfrom the structureof
theoreticalreasoninglessons for understandingpracticalreasoning.
The cruciallesson is thatthereis a differencebetween a conditional
(attached)probabilityjudgmentand a detachedor all-out one. That
"h is most probableon the total relevantevidence" does not imply
the all-out or detachedjudgmentthat "I ought to believe h to such-
and-suchdegree"or, simply,h.
Correspondingly, Davidson holds that practical reasoning
consists in relationalprimafacie judgments of what is the best, or
better,judgment, given various considerations.But the judgment
 AKRATIC
BELIEVING? 3
that option A is best, all-things-consideredis distinct - does not
imply - the all-outjudgmentthat A is best or that I ought [intend]
to do A. To judge, all things considered, that doing A is better
than doing B is consistent with an out-and-outjudgment that B is
betterthanA (a judgment"detached"from its all-things-considered
evaluativegrounding).6The weak-willedperson then either does B
intentionallyor, at least, intendsto do B.
The consistencyof the all-things-consideredandthe opposed all-
out judgment secures the coherence, and so possibility,of akrasia.
But it is not rational because it violates the "principleof contin-
ence", which requiresthatthe all-things-considered judgmentfix the
all-outjudgment.Davidson (1986, 1999) takes this as a constitutive
demandof rationality.It parallelsthe demandupon the rationalman
to satisfy the "requirementof total evidence" when he concludes
from a conditionaljudgment to an assignmentof credence: "give
your credence to the hypothesis supportedby all availablerelevant
evidence"(1982, p. 41).

3.

The fundamental disanalogy in Davidson's practical-theoretical

comparison surfaces at just the point where the requirementof
total evidence is introducedas a requirementfor the application
of inductive logic. Since the objective of practicalreasoning is to
discover which option is best all-things-considered,then, following
out Davidson's comparison,the objective of theoreticalreasoning
would be to discoverwhich hypothesisis best supported(on the total
availableevidence). But this implicationdoes not hold - theoretical
reasoningaims at what is right(true),not merely whatis best (given
one's evidence).
Acceptance of this implicationin the case of akrasiais at worst
misleading since, in Davidson's argument,theoretical reasoning
serves only as a heuristicfor understandinghis solutionto the puzzle
of akrasia.But the implicationis pivotalfor endorsementsof akratic
belief. Once we dropthe implication,and we comparethe end point
of practicalreasoning with the real end point of theoreticalreaso-
ning, we will understandwhy akraticbelief on the (Davidsonian)
4 JONATHANE. ADLER

model of akraticactionis not possible (This conclusionis not drawn,

however,until Section 5 below).
The objective of theoreticalreasoningis to determinewhethera
hypothesis is true, not whether it is best supported.Knowing that
one's total available evidence supports h better than any of the
alternativesis insufficientfor accepting h as true. Our total avail-
able evidence might be but a fragmentof the evidence requisiteto
acceptingor fully believing thath.
The unconditionalor detachedassignmentof a degree of belief
is not truly the endpointof epistemic or cognitive inquiry.If we do
not judge our evidence adequatefor acceptance,7doubt is not yet
resolved- inquiryought not to close. Whatit is best to do is thatact
which is betterthanall the alternativeson the availablereasons.But
what one can or should believe is only what is genuinely worthy
of belief, not what is currentlybetter than the alternatives(Think
here of the differencebetween poker,where the best handwins, and
rummy,where only the right or properhandcan win).8
Theoreticalreasoning is directed to the content or proposition
believed, and only directedsecondarilyor derivatively,to the agent
or to his attitude.9This structureis obscured if the conclusions
of theoretical reasoning are taken to be of the form "I ought to
believe p". But this is an error.Theoreticalreasoningaims to answer
whetherp is the case, not whetherI ought (ethically?,prudentially?)
to believe it. Of course, it is a naturalassumptionthatif the question
of p's truthinterestsX and X concludes (learns, discovers) that p,
then X ought to believe p. Since a conclusion of theoreticalreaso-
ning is someone's conclusion, then, by this naturalassumption,it
becomes thatperson'sbelief. But no directiveto belief derivesfrom
theoretical reasoning itself. Indeed, the natural assumption actu-
ally misrepresentshow intimate is the tie between conclusions of
theoreticalreasoning and believing. One cannot openly judge the
conclusion of one's theoreticalreasoningto be p, and fail to accept
- fully believe - thatp. For belief is nothingbut an attitudethatthe
contentof the belief is true.
This divergence may be usefully summarizedas a difference
between the relationof the 'a's to the 'b's in each of the following
pairs:'0

a. p is the case.
 AKRATICBELIEVING? 5
b. I ought to believe p.
a*. Q is what is correctto do.
b*. I ought to do Q.
Theoreticalreasoning is concerned to determinea, not b. As just
observed,judgmenta standardlyissues in belief, so b does no work.
Practicalreasoning is concerned to determineb*, not a*, and, as
discussed further in the next section, a*-type judgments do not
constraindeliberationtowardb*-typejudgments.

4.

The fundamentaldisanalogy is, to repeat, that practicalreasoning

aims at what is best, but theoreticalreasoningaims at what is right.
This contrastcan be questioned,particularlyin regardto practical
reasoning.11But for the purposes of this paper I could just stipu-
late that practical reasoning should be so understood.For that is
Davidson's conception, and then my argumentwill be of condi-
tional form. My claim is that if one acceptshis view aboutpractical
reasoning, one should find his conception of theoreticalreasoning
misleading, in the case of akraticaction, and mistaken,in the case
of akraticbelief.
However,within its domain, Davidson's conception of practical
reasoning is not tendentious, and there is really no need to seek
refuge in a conditionalpresentation.Withelaborationsto follow and
furtherillustrations,all that is referredto is that common pattern
of reasoningby which we figure out what to do and whose results
we violate in typical cases of akrasia.For brevity and to build-ina
slight reminderof what notion is involved I will abbreviate"prac-
tical reasoning"as pr, and correspondingly"theoreticalreasoning"
as tr.
The divide between tr's aim of rightness (truth) and pr's aim
of bestness is familiar enough. We sharply distinguish between
whetherone has judged as well as one can and whetherone's judg-
ment is objectivelycorrect.The distinctionis crucialfor ascriptions
of blame andrelatedresponsibility-likeascriptions.In a typical case
of pr,I wantto know whethertonightI shouldgo to the movies with
Tom or bowling with Sally. Once I settle this, it hardly mattersto
me, qua deliberator,what will turn out to be truly best (right) for
6 JONATHANE. ADLER

me. Even if I believed that I will never, or even could never,know

which is truly best, that would not foreclose my interest in deter-
mining what I should do tonight. Withinthe scope of deliberation,
the deliberatoris not beholden (by rationality)to reasons that are
inaccessible.
Pr cannot aim at what is objectively best (for the agent),
since what is objectively best typically depends upon the action's
consequences, many of which cannot be anticipatedat the time of
deliberation.Assume that I reasonablyjudge that my going out
bowling with Sally is better than my going to the movies with
Tom. However, a fire accidentallybreaks out at the bowling alley
ten minutes after we arrive.So, going to the movies is objectively
better.Obviously, it is no demandon pr that I factor in the fire in
deliberationson what to do. As I envisage the reasoning, there is
no negligence. Similarly,objectiverightnessaboutone's beliefs and
desires in regardto the options is no constrainton pr. Oedipus is
not negligent in overlooking the disgust he would feel at merely
the thought of marryinghis mother.The play does not lapse into
incoherence in presenting Oedipus' unfortunatedecision as not
blameworthywithin the scope of his deliberationstowardaction.

5.

Given that pr does not aim at objective rightness, as does tr, we

can now draw our main conclusion. If the productof successful tr
is all-out acceptanceor full belief, not what is best supported,then
the motivationalintelligibility of weakness of belief, on the model
of weakness of will, vanishes. In reachingthis centralconclusion, I
extend an observationof Williams' (1973).
When beliefs conflict, they weaken one another, since both
cannot be true. When one belief is favored by the evidence, the
disfavored belief evaporates, since it has been determinedto be
false. But when desires conflict, as with desires to pursue careers
both in medicine and in ballet, the conflict need not, and typically
does not, weaken either.When one desire is acted upon, the other
retainsa hold, experiencedas regret.
Let us extend Williams' observationfrom beliefs to the evid-
ence or reasons for them. When evidence is adequate to satisfy
 AKRATICBELIEVING? 7

the objective of tr, and we come to all-out judge h true, then we

accept or fully believe it. Consequently, and this is the crucial
claim, previouslyconflictingevidence (i.e., evidence that supported
a contraryof h) is nullified as underminingh12 (Recall that what
recommendsthe comparisonwith tr for Davidsonis thatreasonsfor
action standardlyconflict. He observes a similarconflict of reasons
within probabilistic reasoning. It is here that Davidson draws
his parallel between the prima facie-operatorand the probability-
operator.However,accordingto our argument,the comparisondoes
not hold for tr with its genuine end of an all-outjudgmentof truth).
For an example:If I pass my colleague David's car in the parking
lot of the local diner,I conclude that David is inside. But if, within
minutesof the observation,I talk to anothercolleague on the phone,
who mentionsthatDavid is in his office, then the evidence of seeing
David's car in the lot is nullified. Presumingthat I have no reason
to distrustthe colleague, I infer thatit was, e.g., David's wife or son
who drove his car to the diner.For if observingDavid's car retains
its originalevidentialforce, then I cannotsimply acceptit thatDavid
is in his office. There is still the serious possibility that he is at the
diner,which is incompatiblewith all-outjudgmentthat he is in his
office (A kind of Moore's Paradox:"Davidis in his office, although
there is evidence that he is at the diner".Or, "Davidis in his office,
althoughits possible that he is at the diner").
To bringout the contrastwith pr from anotherangle, observethat
variousauthorshave viewed a decision or formationof an intention
as having second-orderforce: one generatesa reason to engage in
an action and a second-orderreason not to act on reasons to do
otherwise (Raz, 1990). Consequently,and plausibly, weakness of
will in regardto an intentionshouldbe experienceddifferentlythan
acting againsta balance of reasons or all-things-consideredevalua-
tion. When I eat the ice cream sundae after deciding against it, I
experience it as more painful than were I just to recognize that the
pros outweighs the cons without having reached a verdict. Only in
the formercase does my (akratic)intentionor actioncast doubtupon
me as a self-controlling agent. The second-orderview explicitly
acknowledgesthe continuingpull of contraryreasons.
In tr,to close inquiryis both to claim adequatereasonsfor accept-
ance andto commitoneself to not reopeninginquirywithoutspecial
8 JONATHANE. ADLER

reason.But it cannotbe to impose a second-orderreasonnot to have

one's believing moved by contraryreasons. This is not something
we even could do. If you believe thatDavid is in his office, and then
you spot him at the local diner,you are bound (compelled) to give
up your belief. What you now know is that either your judgment
(belief) or the reasoningto it was mistaken,unlikefor the analogous
case in pr.
To generalize to our main conclusion: if evidence is adequate
to accept a hypothesis, no (previously) conflicting evidence can
retain its (epistemic) force against that hypothesis.13 So no evid-
ence remainsthatcan motivatebelief againstthe hypothesisjustified
by one's evidence, as the desires disfavoredin an akraticjudgment
(e.g., not to eat the sundae)still retaina pull on the agent,and so can
motivatehim to act againstit. So weakness of belief - as believing
in oppositionto one's evidence - is motivationallyunintelligible.14

6.

As a way to advance and test the argumentjust concluded, let us

consider some examples expressly meant to be cases of akratic
believing. Alfred Mele remindsus of the following example,
Consider the stock example of the woman who judges, solely on evidential
grounds,thatthereis good andsufficientreasonfor her believing thather husband
is having an affair (and for not believing thathe is not having an affair)but who
neverthelessbelieves, owing in part to her wanting it to be the case that he has
been faithfulto her, thathe has not been so engaged (Mele, 1987, p. 114).

Relying also on a Davidsonianmodel of akrasia(p. 377 ftn. 18),

Thomas Scanlon (1998) offers an example, which is meant as a
counter-exampleto Hurley'sopposed claim (p. 377 ftn. 17):
I may know, for example, that despite Jones's pretensionsto be a loyal friend,
he is in fact merely an artfuldeceiver. Yet when I am with him I may find the
appearanceof warmth and friendship so affecting that I find myself thinking,
althoughI know better,thathe can be relied on afterall. ...
Commentingon a contraryview, Scanloncontinues
... Even if it is true that in order to believe something one must take there to
be a reason for thinkingit true ... this would not rule out "akraticbelief". For
example, in the case of the false friend,mentionedabove, thereis somethingthat
I take to be a reason for believing in his genuineness,namely his appearance of
 AKRATICBELIEVING? 9

genuineness.Given all thatI know abouthim, of course, I know thatthis is not a

good reasonin this case, but it can serve as my reasonnonetheless.In this respect
the situationis quite parallel to some examples of akraticaction: even though I
accept the judgment that the pleasure of going for a walk is not a good reason
for missing my appointment,I act on it nonetheless.In each of these examples a
considerationof the kindthatcan sometimes serve as a good reason(for actionor
for belief) is knownnot to be such a reasonin the case at hand,but is nonetheless
taken as the basis for belief or intentionalaction (1998, pp. 35-36).

Scanlon'sexample15is an advanceover Mele's, which seems viable

only throughgarden-varietyself-deception. In Scanlon's example,
we can even imagine thatthe thoughtthatJones is an artfuldeceiver
is presentto Scanlon in the course of his deciding to rely on Jones,
as his avowals may demonstrate("I know thatJones is a bit sleazy,
still and all ..."). No doubt self-deception and akrasia are closely
linked, as various accounts, including Davidson's, propose. Mini-
mally, both depend upon the ease with which the mind can divide
off its contents, in such simple ways as by categorization.Still, to
understandakratic belief as a form of self-deception, at least of
completed self-deception,16is to break with any model of akrasia
thatextrapolatesfrom familiar,straightforward cases.17 The akratic
agent can first-personallyrecognize his intentionas akratic,but the
self-deceived agent cannot so recognize his (continued)belief as
self-deceptivelyformed.
There is, however, a fundamental problem with Scanlon's
example as illustratingthe parallelwith akraticaction, and it is one
that applies to Mele's as well, even if envisaged in the early stages
of self-deception.To take a discreditedreasonas one's reasoneither
to act or to believe would invitethe questionof "why".In the case of
the akraticaction, the answeris thatthe anticipatedpleasurecan be
Scanlon'sreasonto (e.g.) go for the walk and miss the appointment
because of its lingeringdesirability.
What are we to say, though,of Jones's "appearance of genuine-
ness"?Why (how) shouldthis discreditedreasonserve as Scanlon's
reason for believing in Jones' loyalty? Scanlon takes the parallel
(with akratic action) to be that just as a subset of one's reasons
to act, e.g., the pleasure of the walk, can remain a reason to act
on its own, despite its being used up within the agent's all-things-
consideredreasoning, so too the "appearanceof genuineness"can
remain a reason to believe that Jones is a loyal friend, despite its
10 JONATHANE. ADLER

being used up within the agent's reasoning on his total available

evidence. But there is a crucial difference.The pleasurecan render
it moreprobablethat"I [Scanlon]go for a walk"on its own because
it is a pro-attitudetowardthe walk. However, the "appearanceof
genuineness"althoughit rendersit more probablethat Jones is a
good friend on its own, fails to capture any pro-attitudetoward
believing thatJones is a good friend.18Instead,in hearingScanlon's
presentation,we think somethingto the effect that Scanlon,like the
rest of us, is pulled towardJones' salient display because Scanlon
craves either Jones's affection or his friendship.But when we treat
this "craving"as providinga reasonto hold the correspondingbelief
(i.e., as generatingthe pro-attitude),it is precisely an appeal,typical
of self-deception,that goes outside the set of considerations(in tr)
which weigh upon Scanlonin forminghis judgmentthatJones is an
artfuldeceiver.19So this reasoncannotbe the locus of both motiva-
tional intelligibilityand first-personalirrationality,as it could be for
weakness in regardto pr.20
Similarly,consider again Mele's borrowed(and slightly altered)
example of the suspicious wife.21How will Mele preservethe first-
personalirrationalitythat characterizes(Davidsonian)weakness of
will? The difficultythat Scanlon's example faces applies here: The
considerationthat moves the wife against her belief is not itself a
portionof her total evidence by which she forms thatbelief.
That the wife wants the husbandto be faithful enters into the
considerationswhich bearuponthe desirabilityof herbelieving (her
attitude). However, what we deliberate about in tr is the propo-
sition to be believed, not the attractivenessof one's believing.
Consequently,if the wife acts on her prudentialreasoning,she will
suffer no first-personalirrationality:She does not act on a conside-
ration already included in her evaluationof the belief, and which
is dispreferredwithin that evaluation.She is moved by her pruden-
tial motive. This discrepancyunderminesa wide arrayof putative
examples of akraticbelief.

7.

Thereis an obvious way to attemptto remedythis discrepancy(that

the considerationmoving the agent to act is not included in the
 AKRATICBELIEVING? 11

evaluative grounds for judgment). The proposal is just to include

that considerationin the deliberativeassessment.
Such an inclusion is initially plausible anyhow on grounds
assumedin the previoussection. Even if one is not a Humean(inter-
nalist), it is hard to discern any motivationalforce in merely the
bland fact of which hypothesisis favoredby the evidence. It is only
via aims, attitudes,and desires that an agent comes to care about
that evidence and what hypothesisit favors.
But this proposalintroducesa bizarreelement - a form of unin-
telligibility, not in motivation,but in reason. Manifestly prudential
reasons - like the wife's desire for the husbandto be faithful- are
too blatantlyill fit to be enteredinto tr.
The claim of bizarreness assumes that no believer could in
full awareness mistake manifestly prudentialreasons for eviden-
tial or epistemic ones. Any normalbeliever - one who has a vast
range of fairly dull beliefs e.g., Napoleon is dead; Flowers grow -
must recognize, and be competentto apply, a distinctionbetween
epistemic and non-epistemic reasons. For non-epistemic reasons,
ones which yield preferencesamong beliefs, are constantlypresent
for many of our beliefs. Their pull must accordinglybe regularly
resisted or, more typically, ignored if one's corpus of beliefs is to
retainnormalcy(so thatthe believercan treathis corpusas accurate
and thus as a guide to action). For there are numerousbeliefs that
we lack, thoughwe desire them to hold e.g., thatI am a millionaire,
as well as numerousbeliefs we have, which we do not desire e.g.,
that I am too shortto play professionalbasketball.
Although this assumption - that a normal believer requires
rudimentarycompetence with the epistemic/non-epistemicreasons
distinction - appears disputable, I shall not pursue its defense
further.Aside from an extendeddefense being tangentialto the main
issues, the basic claim thatit supportsseems to be conceded. Thus,
it is typically inferredfrom e.g., the alleged rationalityof the wife's
holding the belief that her husbandis faithful despite her grounds,
that she ought to induce herself to believe it (i.e., that she engage
in self-deception). No one proposes that the wife simply enter her
desire or utility calculations,as epistemic reasons, into the evalua-
tion of the propositionthat her husbandhas been unfaithful.The
uniformity of the recommendationthat the wife deceive herself,
12 JONATHANE. ADLER

which guided my initial characterizationof the example,constitutes

a tacit concession of the starkimplausibilityof virtuallyany normal
adult treatingtheir reason to want to believe p as a reason of p's
truth.
Instead, the wife's desire to acquire the belief, contraryto her
evidence, operates, not as an epistemic reason of its truth,but as
a non-conscious bias or influence that helps induce her believing
it. But now the disanalogy observed earlier stands intact. There
is no first-personalirrationalityof a kind analogous to that in
the Davidsonian model because the wife is not moved to act by
a considerationwhose force is already and explicitly used up in
deliberation.Consequently,we still do not findthe irrationalitychar-
acteristic of Davidsonianweakness of will. For the self-deceptive
projectonly makes sense if the prudentialmotive - for the wife to
satisfy her desire or, moreplausibly,to save her marriage- is strong
enough to rationalizeobscuringthe truthto herself (e.g., the wife's
obscuringher evidence to herself). But in that case we begin with
a rational motive, even if it cannot operate in full awareness. So
the example loses the first-personalirrationalitythat is a pillar of
Davidsonianakrasia.

8.

Ourargumentso far has restedon the objectiveof tr as the reaching

of a thresholdof evidence or reasons sufficient for acceptance or
full belief. So the argumentleaves open modelling weakness of
partial, or degrees of, belief on Davidson's model. With partial
belief, conflicting evidence retains its force in opposition to the
favoredhypothesis.Moreover,thatoppositionneed not call for self-
deception,and whatis opposedcan be evidence alreadyenteredinto
tr. So first-personalirrationalitycan be restored.
This opening, however, is much narrowerthan it appears.But
let's firstexplore what is promisingin it. Reasoningto partialbelief
is like pr in not having a premise which implies that one's current
evidence is representative.Only if one's currentevidence represents
futureevidence, so far as the propositionin question is concerned,
do we correctly draw the all-out - detached- conclusion of truth.
But, as with pr, when we affirma conclusion issuing in a degree of
 AKRATICBELIEVING? 13

belief, there is always an implicit relativizationto the evidence, as

premisesof the reasoning.The detachmentis for purposesof action,
not truly an all-outjudgmentand not a complete detachment(from
the premises).
Consequently,pr and tr (which ends in partialbelief) are defea-
sible in a way that (successfully completed) tr is not. Adding new
evidence as premises to the previous total available evidence can,
and usually does, alter the assigned probabilityto p without any
implicationthatthe earlier,differentassignmentis erroneous.22But
if I reach an all-outjudgment from tr, and new informationcomes
along thatrequiresa withdrawalof thatjudgment,thenindeedI have
made a mistake(eitherthe reasoningor the conclusion is faulty).
However, this significant disanalogy is obscured when tr is in
the service of pr. There is a demand for action that necessitates
abortinginquiry.Scanlon's example seems to fit this pattern.The
example will not ring true if you imagine it to take place purely in
reflection ("I wonder whetherI could rely on Jones as a friend?").
This tarnishes the example. In reflection we would hesitate to
take Scanlon'sjudgment to be all-out that Jones can be relied on,
ratherthan that Jones is reliable enough, or that I (Scanlon) can
be sure enough of his reliability,for me to trust him now. Once
removed from the context of action, Scanlon's attitudewill revert
to dependenceon his knowledgethatJones is a sly deceiver.Corres-
pondingly,he will not enter an all-outjudgment that Jones can be
trusted.23
When the concludingjudgmentis only of partialbelief, however,
the possibility of defying one's evidence is more plausible. You
are deciding between a vacation in Paris or one in Vancouver,and
after devoting a considerableamountof time to the evaluation,you
conclude that a vacation in Vancouverwould probably be better
now than one in Paris. But, as the moment of decision and action
drawsclose, romanticimages of Parisbecome especially vivid, even
though the import of these images was already included in your
total evidence. Your judged degree of belief favoring Paris does
not conform to your all-evidence-consideredconditionaljudgment
which favorsVancouver.So here is a plausiblecandidatefor akratic
belief on the model of Davidsonianakraticaction.
14 JONATHANE. ADLER

But thereis a seriousqualificationon the rangeof cases to which

this model can apply. In the kind of case just presented,the move
from the conditionalto the all-outjudgment,and so to the aborting
of inquiry,is demandedby the natureof the decision problem.To
wait until you have adequateevidence to all-out judge it true, not
merely probable,thatthe vacationin Vancouveris preferableto the
one in Paris would either be too late (your vacationtime will have
past) or not worthwhile(you have otherthings to do with your time
than continueto deliberateon this matter).
Since there is a temporal gap between actions and intentions,
especially when instituting a plan, we have to retain reasons
(premises)thatexpress our aim. We have to allow for alterationsin
circumstancesor informationas a resultof which we need to adjust
our mediating actions to remain on track of our goal. The reason
that I should take Route 95 north is that I want to visit Kate, and
that is the most efficient way for me to visit her. Having drawnthe
conclusion that I ought to take Route 95 can I wholly detach from
my reasons? No. If I learn, as I am driving, that Route 95 north
is closed, then I must at that point select an alternativeroute. But
that alternativeroute will be determinedby my objectiveor goal or
reason - to get to Kate's. So I ought not lose track of that reason
priorto completingthe action thatis the conclusion of my pr.
Davidson'sconceptionof akrasiaskirtsaroundthis issue because
he takes the conclusion of pr to be an intention.The intentionwill
follow immediately from pr, and so it does not face the temporal
problem.But then Davidsoneliminatesa locus for weakness. Obvi-
ously, we can be weak in intending an action and not carrying
throughon it (Holton, 1999).
But normally,with tr, if a successful conclusion is reached,the
reasons or evidence are not retained.Their usefulness is exhausted
by their role in acceptance (assuming no further, independent
interest in them). Once the colleague tells me that David is in his
office, to recall an earlierexample, I come to believe it. Given that
belief, the interest in how I learned it is exhausted.We could, of
course, easily imagine variousquestionsthatmight be raisedwhich
would be unanswerablewithoutthatevidence- most obviously,how
do you know that David is in his office? But we should not retain
evidence merely on the off-chancethata questionmight be raisedto
 AKRATICBELIEVING? 15

which it will supply the answer.Memoryis not cheap or unlimited,

and it has been well arguedthatwe shouldnot retainour reasonsor
evidence given our needs for cognitive economy (Harman,1986).
These arguments,though, do not go far enough. The more far-
reachingreasonthatthe terminusof successful tr is real detachment
from its premises is that to take it as a demand upon oneself to
retain one's reasons, after a belief is formed (or rejected), will be
to disrespector deny one's authorityas a believer.
As corroboration,let us appealto the parallelbetween assertion
and belief (assertion as the expression of belief).24 Asserting is
governed by a default rule - accept a speaker's sincere assertion,
unless you (the hearer) have special reason to object (Brandom,
1994, Chapters2 and4; Burge, 1993). The defaultrule removes any
burdenon speakersto hold on to reasons whose only functionis to
back their assertions.To do so automaticallywould be unfeasibly
prolix. Speakers must offer their backing only if properly chal-
lenged. It is the challengerupon whom the burdenfalls to show that
her challenge is proper.The challengerincursthe burdento present
reasons for why her challenge constitutes a serious possibility of
errorin the speaker'sassertionor his position to offer it. So to insist
on the speaker'sretainingthe reasons that he would offer to back
his assertion,subsequentto a properchallenge, would be to require
that retainingone's reasons is a normalcost of assertion.But such
a requirementrendersassertiona much more effortfulundertaking
and so diminishesthe ease with which speakersarewilling to inform
hearers.More crucially,it would imply thata speakeris not (default)
entitledto standbehindhis words.
The disanalogy we now find with the Davidsonian model is
this: Within tr, if relevantevidence for a hypothesis is not shown
misleading or superfluous, it needs to be retained as inquiry
proceeds, until an all-outjudgmentof truthis reached. So in tr we
will not normallyform a detachedjudgmentof probabilityissuing
in a correspondingdegreeof belief. We will stick with a conditional
- nondetached- judgment, which overtly displays that inquiry is
incomplete. Again, then, except if tr is in the service of pr, and
so under exigencies to guide action, there is no place for akratic
(partial)believing on the model of Davidsonianakraticaction.
16 JONATHANE. ADLER

9.

When tr is in the service of pr, as just discussed,it does not proceed

to reach its fundamentalobjective - inquiry is aborted, prior to
reachingan all-outjudgment.The abortedprocedureis uncharacter-
istic of tr notjust in ending with a partialratherthana full belief, but
in a related,less strikingway. To partiallybelieve that p is thereby,
if one's degrees of belief conform to the axioms of probability,to
partiallybelieve that not-p (to a complementarydegree).25But to
not-believe that p is not thereby to believe that not-p, nor does it
imply it. Developmentof this claim leads to furthercontrastswith
pr.
A characteristickind of weakness in believing (and tr) is to
continue to deliberateor inquire,despite convincing evidence. But
Davidsonian akratic action requires an all-out judgment. I have
followed Davidson'smodel, along with the literature,in considering
only the case of believing againstone's evidence. I hope to redress
this oversightbriefly here to the extent that it relates to our central
claims.
Shakespeare'sHamletcomes quickly to mind as almostparadig-
matic of the suggested weakness in believing. However, Hamlet
appearsindecisive, placing too much weight on avoidingerrorover
obtainingtruth(James, 1951), because of the practicaldemandon
him to act, as discussed in the previous section. I doubt that we
wouldjudge Hamletweak-willedwere his governinginterestor task
not action, "butmerelythe acquisitionof knowledge",as Descartes'
expresses his goal in MeditationI.
Hamletdisplays a hesitancyto act thatcould not be Davidsonian
akrasia.When Hamlet refuses to judge that his mother and step-
fatherkilled the king, his father,Hamlet'sweakness of will consists
in keeping inquiryopen, not in violating an all-outjudgment.
The role of the need to act and so to abortinquiry,as a constraint
on deliberation,is a premise in pr, even if tacit. Pr is reasoning
about what one should do, given one's beliefs and desires. The
move from the all-things-consideredto the all-out judgment in pr
is effectuated by the tacit premise that the options do not permit
furtherinquiry.26The premise provides the linchpin in the move
from "giventhese reasons,it is best to do A" to "I intendto do A".
Withthatlinchpin,no objectionarises from an admissionthatgiven
 AKRATICBELIEVING? 17

more time to inquire,further,potentiallycontraryreasons,may - or

even likely would - obtain.
Admittedly,tr is also a boundedactivity.But the boundingis not
inherentin any particularconclusion to be reached.It is because of
our finiteness and limited resourcesthat inquiriesare closed if they
reach a thresholdof evidence or reasons adequatefor acceptance
or full belief. We need not continueinquiryto achieve ever greater
certainty. We count evidence as adequate when a contextually
restrictedor relevantrangeof counter-possibilitiesis eliminated,not
all (logically or conceptuallyor even physically) possible ones.
But these constraints,which shape the structureof reasoning
toward belief, are not premises of that reasoning. There is no
premise, within any instance of tr, which implies that we have to
settle the truth of a hypothesis in a specified time. Were such a
premise involved in derivingthe conclusion, that conclusion could
not be an all-outjudgmentof truthor falsity,but only of bestness.
By keeping inquiry open, despite (let's assume) adequateevid-
ence, Hamlet suffers the 'bad faith' of trying to resist the demand
to believe and to act, accordingly.However, the logical structure
of not-decidingand not-believingare quite different.Not-deciding,
like deciding-not implies, in normal circumstances,not-doing (or
not-intending-to-do)27 (Hereis a truthin the saying "notto decide is
to decide").Hamlet'snot deciding to avenge his father'smurder(at
the early stages of the play) has the same consequence (for action)
as his having decided not to avenge it. When the seemingly same
option presents itself again later,Hamlet is in a differentinforma-
tional stateunderdifferentdemandson action. In particular,thereis
a much lower risk of error.
But not-believing does not imply believing-not. Not coming
to the belief that his step-fatherand mother are guilty is widely
different from coming to the belief that they are not guilty. You
mightrespondthateitherwould have prettymuchthe samepractical
consequences. But this does not hold generally. It depends upon
the peculiar exigencies of Hamlet's charged circumstance.If not-
believing is the currentposition in a cognitive inquiry,then inquiry
remains open and to be pursued.Even if we impurify inquiry by
addinga call for action,thoughless demandingthanthatfor Hamlet,
the not-believing/believing-notdifference, remains wide. Imagine
18 JONATHANE. ADLER

that the question to Hamlet is only whether the stepfather and

mothercauseda hit-and-runaccident.If Hamletresolves inquiryand
comes to disbelieve that they did it, then he would obviously take
no furtheraction. But if he retains favorable,though inconclusive,
evidence thatthey did cause the accident,and he recognizes thatno
furtherinquiry of his is likely to be feasible, then there is a good
chance that he will take action by e.g., confrontingthem or even
calling the police.

10.

I have argued against the possibility of akratic believing on the

Davidsonian model of akraticaction, whose main ingredientsare
the distinction between all-out and detached judgments, and the
retentionof motivationalintelligibility and first-personalirrationa-
lity. In so arguing I have set out various disanalogies between tr
and pr. I do not deny that there are illuminatingparallelsbetween
them - I thinksuch illuminationis found in a numberof Davidson's
comparisons,as well as those of Velleman.28Nevertheless,most any
serious attemptto view tr and pr on a similar model is bound to
be misguidedand misguiding,even when the comparisonsselected
hold. Very roughly, tr is reasoning to remove uncertainty;pr is
reasoningunderuncertainty.
Sometimes,pr is reasoningundercertaintye.g. I need money,and
the only accessible location for money is the ATMacross the street.
So, I should certainlygo to the ATM.But this is just happenstance.
Sometimes, as well, pr involves reasoning to diminish uncertainty
- we try to figureout what we really want and how best to achieve
it. But the lessening of uncertaintyis a means, not a goal, and then
only a goal within the bounds permittedby practical constraints.
However, the elimination, not just the lessening, of uncertaintyis
the goal, and not just a means, in tr.
To review: The fundamentaldisanalogy is that the goal of tr is
all-out or full belief, and so the (threshold)acceptanceof a propo-
sition. When tr reachesthatgoal, contraryor underminingevidence
is nullified. So there is no evidence to play the role of conflicting
desires in drawingthe agent away from his betterjudgment.
 AKRATICBELIEVING? 19

Next, we observed that treatingakraticbelief as a form of self-

deception immediately breaks with Davidsonian akratic action.29
If the self-deception is motivatedby desires that do not serve an
evaluative role in judgment, then we surrenderthe first-personal
irrationalityof akrasia. But if we imagine instead that they are
entered in an evaluative,ratherthan merely a causal, role then the
proposal is bizarre, since overt reasoning cannot accord with it.
These conclusionsexposed a broadbarrieragainstakraticbelieving.
A statement which reports on motives (reasons) that pull belief
against the direction of the total evidence is not a part of that
evidence. So, again, we surrenderthe crux of the first-personal
irrationalityin Davidsonianakrasia.
The disanalogiesset out so far focused on full belief, as ordinary
belief, leaving open the possibilityof Davidsonianakrasiafor partial
or degrees of belief. However,within tr, partialor degrees of belief
must be viewed generally in a Peircean way as unstable (unsatis-
factory) states seeking resolution in a terminus of inquiry with
acceptance(full belief) or rejection.In thatcase, partialor degreesof
belief would not issue in an all-outjudgment,which is necessaryfor
Davidsonianakrasia.Rather,the degree of belief would represent
a conditionaljudgment, which displays openness both to further
inquiryand,correspondingly,to modificationwith furtherevidence.
Finally,there is no place in tr for an inferencefrom not-believinga
particularproposition,as when suspendingjudgment,to any belief,
specifically,to believing-not.However,given the constraintsunder
which pr occurs, to not choose to do any of a set of options is, like
choosing not to, to do-not all of them.
These various disanalogies are all unified by the fundamental
one - of tr aiming for rightness and of pr aiming at bestness. The
unificationis tied to a glaring difference between pr and tr not yet
explicitly mentioned.We have free will in regardto actions, as we
do not in regardto belief.
Free will is, of course, a presuppositionof any responsibility-
involving activity,and in this respect, there is no special reason to
highlightit here. But once the comparisonbetween akrasiaof action
and of belief is proposed, the free will presuppositionis forced to
the fore. We simply cannot freely decide to come to a belief or
to suspend judgment. Williams has famously argued against the
20 JONATHANE. ADLER

possibility of simplybelieving at will (1973, "Decidingto Believe").

The core of Williams' argumentis that were a belief so formed we
could not recognize it as a belief because our decision cannotmake
the belief true.
It is this claim to truthupon which the fundamentaldisanalogy
rests and from which I drewvariousconsequences.It is the concept
of belief, and its claim to truth, that informed the denial in the
previous section of any inference from not-believingto believing-
not (i.e., to any attempt,as with James [1951], to treat believing
that p and believing thatnot-p on the model of the law of excluded
middle).
When one acts akratically,one is in oppositionto one's judgment
of what is best for one to do. But were one to believe akratically,
one would not just oppose one's judgment as to what it is best
to believe. One is in opposition to belief itself. One can avowedly
act badly or will akratically- "I know I shouldn'teat that brownie
sundae,but I'll enjoy it so much. So to hell with good judgmentjust
this once .. ." But one cannotavowedlybelieve akratically- thatis,
believe in overt defiance of one's evidence. For we each know that
the only basis to satisfy belief's claim of truth,at least for a wide
range of ordinarybeliefs, is throughevidence. So to be in explicit
defianceof one's evidence would be to contradictthatclaim.
Can I corroboratethe claimed difference in coherence between
akraticaction and believing? The corroborationis available,but it
requiressome massagingof the data.The difficultyis thatone can be
conscious of a complex thought without attendingsimultaneously
to all its salient parts- specifically,the conjunctsof a conjunction.
To highlight the point, I use an extreme case (because it involves
mental disturbance).A girl sufferingfrom anorexianervosacan be
imaginedto be entertainingsome thoughtto the effect that I despe-
rately need to lose weight, but it is evident, as I look in the mirror,
that I am thin and do not need to lose weight. The thought seems
an instance of akraticbelieving, yet, she does have the thoughtand
so, trivially,it is possible. But what is not possible, and thus does
corroboratethe claimed incoherence, is that she cannot attend to
both conjunctssimultaneously,which is the conditionthatassertion
readily satisfies.
 AKRATICBELIEVING? 21

The conditionthatwe needed to impose to bring out the contrast

suggests an explanationfor how the many putativecases of akratic
believing can be denied, when the deniers accept the same charac-
terizationof akraticbelief as those who offer these cases. Defenders
take the data at face-value, which is fair enough, since, initially
at least, the cases ring true. However,the failure of these cases to
genuinely count as instancesof akraticbelieving becomes apparent
only with the contrivanceor idealizationof full awarenessandatten-
tion (underwhich these cases eitherdissolve or turnout to be forms
of self-deception).30Since these conditionsare not regularlymet in
everyday life, the consequences of imposing them are easily over-
looked. Nevertheless,these conditionsare appropriate,as they focus
us on the first-personalthoughtsthat should correspondto akratic
action and believing, respectively,and it is the possibility of these
thoughtswhich is at issue.
Absent, however, an independentcheck on whether the condi-
tions of full awareness and attention are met, the proposed test
cannot be successfully appliedhere. Still, we can focus directly on
the coherence claim, which is, anyhow, the source of confidence
in the proposedtest. Contrastassertionsof a form to express one's
weak-willed act
(1) I intend to do R, even though I judge doing S better all things considered
(includingthe reasonsto R).

with those to express one's (alleged) weakness of belief

(2) p, even thoughthe evidence indicatesnot-p.

e.g. Jones is a good friend,thoughthe evidence indicatesthathe is not.

Assertions of instances of (2), but not (1), are heard as suffering

a Moore's Paradox-typecontradictionthat is explainedas due to a
correspondingincoherencein thought.31
If assertionsof the form of (2) are Moore's Paradoxical,then it is
unsurprisingthat akraticbelief cannotbe modelled on Davidsonian
akraticaction. And it would be surprisingif any model of akratic
action that retains first-personalirrationalitycan serve as a model
of akratic belief. For the first-personalthought correspondingto
the admission of akraticbelief would be not merely irrational,but
incoherent.32
22 JONATHANE. ADLER

NOTES

* For their helpful comments, I thank Michael Cholbi, Chris Gowans, Arthur
Kuflik,Sidney Morgenbesser,David Owens, Amelie Rorty,Michael Stockerand
especially Ruth Chang and TamarGendler.I am also grateful to a referee for
suggestions for developing the article.The argumentin the early sections draws
upon Adler (2002, Chapter3).
1 The most explicit is Heil (1984). Mele (1987, Chapter8) accepts the basics
of Heil's model, while dissentingfrom Heil's deemphasison akrasiaas a loss of
self-control.See also Scanlon (1998, Chapter1) and Millgram(1997, Chapter6).
Davidson's (1986) explicitly builds upon his (1982) in developing comparison
with "weaknessof the warrant".
2 For exceptions see especially Sections 8 and 9 and for pursuitof the theme of
othervarietiesof akrasia,see Holton (1999).
3 For more carefulformulationssee Heil (1984) and Mele (1987, Chapter8).
4 Recently, McIntyre(1993) and Arpaly (2000) have arguedthat akraticaction
can be rationaland so their view appearsincompatiblewith Davidson's. But the
views are reconcilable.
The basic argumentthat it could be rationalto act against one's best judg-
ment turnson the possibility thatone's bestjudgmentreflects an unrepresentative
sample of one's reasons. Consequently,it could be rationalfor the agent, even if
not from his currentpoint of view, to suffer weakness of will in regardto that
judgment.In a compressedversion of Arpaly'sexample, a studentSam comes to
believe thathe ought to become a hermit,so as to pursuehis studiesbetter.But he
does not so act - he suffersweakness of will - because, say, he has a deep-seated
prejudiceagainstsuch a life. Still, it was rationalfor him to act againsthis better
judgment, since he has good reason for not being a hermit,given his goals e.g.,
he actuallystudies worse withoutthe occasional distractionsof social life.
We can reconcile these views if we adhere strictly to the characterizationof
Davidsonianirrationalityas first-personal.Aroundthe time of action, Sam will
regardhis failureto abandoncivilizationfor the hermit'slife as his acting against
reason. However, later on, when he obtains a more matureperspective,he will
regardthe weakness as what was trulybest for him at the earliertime. The retro-
spectivejudgment still distinguishesSam's akrasiafrom the usual kind, since in
the usual kind one is not retrospectivelygrateful.
5 The agent must violate the Principleof Continence,since it is constitutiveof
being an agent. On this and other insights into Davidson's views of practical
reasoningsee Lazar(1999).
6 Davidson explains why theremust be unconditionaljudgmentsthat one action
is better- "otherwisethere would be no such thing as acting on a reason"(1982,
p. 39).
7 On acceptance as the goal of inquiry see, among others, Harman(1986) and
Levi (1981). However, there is opposition to the notion of acceptance,particu-
larly due to Jeffrey (1983, 1970). To my mind this dispute has been resolved
to favor acceptanceor full belief as the goal of inquiry- see especially Kaplan
 AKRATICBELIEVING? 23

(1996), which explicitly defends an acceptance structurewithin a Jeffrey-like

framework.
8 There is, of course, this disanalogy:The cards in the poker hand are not evid-
ence thatit is a winning hand.They'reconstitutiveof it.
9 This divergenceis a forceful premisefor Williams (1985, Chapters4 and 8).
10 The representationwas suggestedto me by Gendler.
1 For the questioning,I am indebtedto Chang.
12 The "nullification"may create an
explanatorygap as to why the seeming, but
not genuine, counter-evidenceexists. But this explanatorygap is just a further
puzzle, not a barrierto acceptance.
13 The conclusion here runs counterto Millgram's(1997, p. 130 ftn. 37) ground
for defendingakraticbelief.
14 We can apply our criticism to the most exacting attempt to model akratic
belief on Davidsonianakraticaction. Mimicking Davidson, Heil (1984) begins
with two principles.These seem to contradicta thirdthat supplies the conditions
for incontinentbelief:
RI: WhereP andP' areepistemicallyincompatiblefor S [S's holdingP constitutes
a reasonfor him not holding P'], if S holds P to be more warrantedthanP', S will
hold P if he holds eitherP or P'.
R2: WhereS holds R andR' andtakesthese to be all thatis relevantto the warrant
of P and P' respectively,then if S holds R epistemicallyto outweigh R', he will
hold P to be more warrantedthanP'.
RI parallelsDavidson's condition which moves from preferencesin wanting to
preferencesin intentions (or actions); while R2 parallel's Davidson's condition
which moves from preferencein judgmentto preferencesin wants.
Heil defines S's holding a belief P' incontinentlywhen:
(i) S takes P and P' to be epistemicallyincompatible;
(ii) S holds R and R', and takes these to be all that is relevantto the epistemic
warrantof P and P' respectively;
(iii) S takes R epistemicallyto outweigh R', hence P to be more warrantedthanP'
given R and R';
(iv) S holds P' (and does not hold P).
However, again following Davidson, Heil hopes to reconcile (i)-(iv) by distin-
guishing between the reason thatjustifies one's belief and the reason for which
one believes. The reasonthatJim takes to justify his belief thatMarylikes him is
that she smiled at him duringlunch. But the reason he actuallydoes believe it is
wishful thinking.If R warrantsP for S, then
(a) S takes R to warrantP, and
(b) R is (partof) what causes (sustains,supports)S's holding of P.
Incontinentbelief occurs when (a) holds, but not (b).
To briefly indicatehow our criticisms in the text apply: (1) Heil's two condi-
tions cannot representcompletion of reasoningtowardbelief because P's being
24 JONATHANE. ADLER

more warrantedthanP' is insufficientfor holding P. (2) If we try to answer(1) by

imaginingthatthe strengthof R is sufficientfor acceptance,then R' will lose any
of its force in favorof P'.
15 Similarly,Millgram's(1997, pp. 129-130).
16 "Complete self-deception"involves the loss of the offending belief, so the

typical tension of self-deception is removed. However, I am doubtful whether

complete loss, as contrastedto hiding or obscuringthe offending belief, is the
appropriategoal. For in the former case one will not generatedefenses against
coming across further problematic evidence. On the matter of whether self-
deceptionmust involve a tension see Mele (2001, pp. 52-56).
17 See Davidson's comparisonof weakness of the warrantand self-deception
(1986, p. 85).
18 If the alleged parallelof akraticbelief with akraticaction is to be maintained,
it cannotrely on the possibility of the formerarisingby mistakeor accident (the
mistakeor accidentof treatinga subset of one's evidence on its own).
19 It may be that Scanlon would respond in terms of his broad opposition to
desires as having independentreason-giving force. As one indication of this
theme, Scanlon affirms,subsequentto the above passages, that
Akratic actions (and irrationalthoughts) are cases in which a person's rational
capacities have malfunctioned,not cases in which these capacities are over-
masteredby somethingelse, called desire (1998, p. 40).
I leave aside this broadertheme of Scanlon's, in partbecause it would lead us to
the related,but differentand treacherous,topic of internaland externalreasons.
But also, and more importantly,nothingin my criticismturnson desires as moti-
vational sources of akraticaction or belief. The pertinentassumptionis just that
eitherrequiresmotivationand that a purelyfactualbelief, withoutany attitudinal
or emotionalor normativecomponent,cannotsupply such motivation.
20 A similaranalysis applies to Scanlon'sexample (1998, p. 40.)
21 As standardlytold, the self-deceptive project is not initiated because of the
wife's desire (that her husbandbe faithful), although she undoubtedlyhas that
desire. Contraryto Mele's retelling, the project is undertaken,on the standard
rendition,because the wife takes the expected disutility of falsely ruining her
marriageto be far worse than that of the opposed threatof mistakenlyjudging
him faithful. This divergence is related to another:on the standardversion the
wife does not judge that "thereis good and sufficientreason to believe that . . .".
She only thinks her groundsmake it somewhatprobablethat .... For discussion
relatedto the latterconstrualsee section 8 below.
22 Here especially we should emphasize,what holds throughout,thatthe proba-
bility in questionis epistemic, not objectiveor chance.
23 For relateddoubtsaboutScanlon'sexample, see Wallace(2001, p. 12).
24 The parallelis outlinedin Adler (2002), Chapter10 appendix.
25 For discussion of the implausibilityof the negation law as a law of belief or
supportsee Cohen (1977), Chapter7 and Kaplan(1996), Chapter4.
26 I do not deny that reasoningtowardbetterdesires (ends) can be a form of pr
 AKRATICBELIEVING? 25
(A view vigorously pursuedin Richardson[1994]). What I deny is thatinstances
of pr are incompletewithoutit.
27 The doing-omitting,active-passiveand related, distinctions are not germane
here since those distinctions are in point only when the consequences are the
same. That is not generally true of the blunt doing-not/not-doingdistinction
invoked here. However, there are subtle complexities in representingnegation
in act or intentionsentences that I do not address.For discussion see Vermazen
(1985) and Davidson (1985).
28 Velleman (2000), Chapters8 and 11. Velleman tries to assimilate practical
reasoningto theoretical,whereasmy criticismsaredirectedmainlyto the converse
assimilation.
29 Recently, Davidson clarifies the aim of "How is Weakness of the Will
Possible?"as given by its title:
assumptions that seemed intuitively plausible led to a contradiction.These
assumptions promptedPlato's Socrates to deny that akrasia was possible, and
made Aristotle and many other philosophersdeny, in effect, that it was possible
for an agentto act freely andintentionallycontraryto his own bestjudgmentwhile
realizingthatthis was what he was doing (1999, p. 402).
Wereakrasiato requirehiddennessto oneself of one's judgment,as a kind of self-
deception, the first-personirrationalityclaim would be sacrificed.The action or
intentionis no longer opposed to the governingjudgment.
30 Thus, I take many of Rorty's (1983) rich set of examples to be cases of self-
deception,despite her explicit denials (p. 179).
31 It may be wonderedwhetherakraticactionis really possible undersuch condi-
tions, since it is plausible to hold that irrationaland deliberativeaction does
not allow for full clarity of thought. But even grantedthis plausible claim, the
disparityremains.Contrast:
(4) 1 intendto do R by not focusing on my judgmentthatdoing S is better.
with
(5) p, and I will get myself to (continue to) believe it by not focusing on the
counter-evidenceto it.
(5) is, again, incoherent,and, it shouldnot be confused with
(6) 1 intendto bringit aboutthat I (continueto) believe thatp by not focusing on
the counter-evidenceto it.
Similarly,the contrastbetween (1) and (2) is obscured if one takes the proper
formulationof what (2) means to express as
(3) I believe thatp, even thoughthe evidence favorsnot-p.
For the meaning conveyed in the assertionof (6) or (3), respectively,is not the
properparallelfor (1). The properparallelis of the detachedjudgments:"I intend
to do R" and p, respectively.
32 Mele (1987) considers a claim that "strict incontinent believing ... is
26 JONATHANE. ADLER

impossible"(p. 114), and with Heil (1984), he rejects any such view as "overly
optimisticabouthumanrationality"(p. 114) But if the impossibilityis conceptual,
it has nothingto do with optimism(or pessimism) abouthumanrationality.

REFERENCES

Adler,J. (2002): Belief's OwnEthics, Cambridge:The MIT Press.

Arpaly,N. (2000): 'On actingrationallyagainstone's best judgment',Ethics 110,
488-513.
Brandom,R.B. (1994): Makingit explicit,Cambridge:HarvardUniversityPress.
Burge, T. (1993): 'ContentPreservation',The Philosophical Review 102, 457-
488.
Cohen, L.J. (1977): The Probable and the Provable, Oxford:Oxford University
Press.
Davidson,D. (1982): 'How is weakness of will possible?', Essays on Actions and
Events (pp. 21-42). Oxford:OxfordUniversityPress.
Davidson, D. (1985): Reply to Bruce Vermazen(pp. 217-221), Vermazenand
Hintikka.
Davidson, D. (1986): 'Deception and Division', in J. Elster (ed.), The Multiple
Self (pp. 79-92), Cambridge:CambridgeUniversityPress.
Davidson,D. (1999): Reply to Ariela Lazar(pp. 401-405), Hahn.
Hahn, L.E. (ed.) (1999): The Philosophy of Donald Davidson, Chicago: Open
Court.
Harman,G. (1986): Changeof View,Cambridge:The MIT Press.
Heil, J. (1984): 'Doxastic Incontinence',Mind 93, 56-70.
Holton, R. (1999): 'Intentionand Weaknessof Will', The Journalof Philosophy
XCVI, 241-262.
Hurley,S. (1989): NaturalReasons, Oxford:OxfordUniversityPress.
James, W. (1951): 'The Will to Believe', in A. Castelli (ed.), Essays on Prag-
matism(pp. 88-109), New York:HarnerPublishingCo.
Jeffrey,R. (1970): 'Draculameets Wolfman:Acceptance vs. PartialBelief', in
M. Swain (ed.), Induction,Acceptance, and Rational Belief (pp. 157-185),
Dordrecht:D. Reidel.
Jeffrey,R. (1983): TheLogic of Decision Second Edition,Chicago: The Univer-
sity of ChicagoPress.
Kaplan, M. (1996). Decision Theory as Philosophy, Cambridge:Cambridge
UniversityPress.
Lazar,A. (1999): Akrasiaand the Principle of Continence(pp. 381-401), Hahn.
Levi, I. (1981): TheEnterpriseof Knowledge,Cambridge:The MIT Press.
McIntyre,A. (1993): 'Is AkraticAction always Irrational?'in 0. FlanaganandA.
Rorty (eds.), Identity,Character,and Morality(pp. 379-400), Cambridge:The
MIT Press.
Mele, A. (1987): Irrationality,Oxford:OxfordUniversityPress.
 AKRATICBELIEVING? 27

Mele, A. (2001): Self-Deception Unmasked, Princeton: Princeton University

Press.
Millgram,E. (1997): Practical Induction,Cambridge:HarvardUniversityPress.
Raz, J. (1990): Practical Reason and Norms, 2nd edition, Princeton:Princeton
UniversityPress.
Richardson, H. (1994): Practical Reasoning about Final Ends, Cambridge:
CambridgeUniversityPress.
Rorty,A. (1983): 'AkraticBelievers',AmericanPhilosophicalQuarterly20, 175-
183.
Scanlon, T.M. (1998): What We Owe to Each Other, Cambridge: Harvard
UniversityPress.
Velleman, J.D. (2000): The Possibility of Practical Reason, Oxford: Oxford
UniversityPress.
Vermazen,B. (1985): NegativeActs (pp. 93-104), Vermazenand Hintikka.
Vermazen,B. and Hintikka, M. (1985): Essays on Davidson, Oxford: Oxford
UniversityPress.
Wallace, R.J. (2001): 'Normativity,Commitment, and InstrumentalReason',
Philosophers' Imprint1, 1-26 <www.philosophersimprint.org/001003>.
Williams, B.A.O. (1973): 'EthicalConsistency',Problemsof the Self, 166-186.
Williams, B.A.O. (1973): 'Deciding to believe', Problemsof the Self.
Williams,B.A.O. (1973): Problemsof the Self, Cambridge:CambridgeUniversity
Press.
Williams, B.A.O. (1985): Ethics and the Limits of Philosophy, Cambridge:
HarvardUniversityPress.

BrooklynCollege and the GraduateSchool

CUNY
E-mail:jadler@brooklyn.cuny.edu