Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
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.
Duke University Press and Philosophical Review are collaborating with JSTOR to digitize, preserve and extend
access to The Philosophical Review.
http://www.jstor.org
1. Introduction
Pascal's Wageris simplytoo good to be true-or better,too good to be
sound. There must be something wrongwith Pascal's argumentthat
decision-theoretic reasoning shows thatone must (resolveto) believe
in God, ifone is rational.No surprise,then,thatcriticsoftheargument
are easilyfound,or thattheyhave attackediton manyfronts.For Pascal
has giventhemno dearthof targets.
Virtuallyall of the Wager's criticshave directed theircampaigns
againstitspremises.Otherauthorshave ralliedto itsdefense,buttress-
ingthosepremises.I willargue thattheyare fighting a lostcause: devel-
opingargumentsbyJeffrey (1983) and Duff (1986), willcontendthat
I
theWageris simplyinvalid.This motivatesa searchforreformulations
of the originalargumentthatare valid,whileupholdingitsspirit.I will
offerfour such reformulations, each of which finessesthe decision
matrixof the Wager,and in particularits problematicinvocationof
"infiniteutility."
Yet thesereformulations falltoo, albeitfora different
reason.This,in turn,mightpromptadvocatesof theWagerto conduct
anothersearch forstillfurtherreformulations. However,I will argue
thatsuch a searchis likelyto be futile.When we examinewhatis at the
rootofthefailureoftheoriginalWager,and ofthe reformulations that
I offer,
we realizethattheirfailuresare symptomatic ofa deep problem
thatanyvariantof theWagermustovercome.I willpresenta dilemma
forall such variants,and conclude thattheirprospectsforsuccessare
dim.
Pascal's Wager
27
Rationalityrequiresyou to wagerforGod.2
Conclusion:
Some clarificationis in order.Accordingto Pascal, "wageringfor
God" and "wageringagainstGod" are contradictories,as thereis no
avoiding wageringone way or another: "you must wager. It is not
optional." (Unless otherwisestated,all quotationsof Pascal are from
his 1948, Trottertranslation,?233.) The decision to wager for or
againstGod is one thatyoumake at a time-at t,say.But ofcourse Pas-
cal does not thinkthatyou would be infinitely rewardedforwagering
forGod momentarily, thenwageringagainstGod thereafter; nor that
you would be infinitely rewarded for wageringfor God sporadically-
only on everyother Thursdayafternoon,for example. What Pascal
intendsby"wageringforGod" is an ongoing action-indeed, one that
continuesuntilyourdeath-that involvesyouradoptinga certainsetof
practicesand livingthekindoflifethatfostersbeliefin God. The deci-
sion problemforyou at t,then,is whetheryou should embarkon this
course of action; to failto do so is to wageragainstGod at t.3
I understandPascal as regardingyoursalvation,withitsinfiniteutil-
ity ("an infinityof infinitelyhappy life"), as thebestthingpossibleforyou.
I take thisto be in keepingwithCatholic tradition,4but more impor-
tantlyitis supportedbyPascal's owntext.In hispreambleto theWager,
he writes:"Unityjoinedto infinityadds nothingto it ... theadditionof
a unit can make no change in itsnature."In particular,infiniteutility
is not augmentedbythe additionof a unitof utility.The pointis made
even moreemphaticallyin theWageritself,and here itwillbe usefulto
consult the originalFrench text:"si vous gagnez vous gagnez tout."
Trottertranslatesthis accuratelyas: "If you gain, you gain all." We
mightjust as accuratelytranslateit as: "If you gain, you gain every-
28
29
30
3.2 Regularity
Let us pursue thisline of attackstillfurther.
Call a probability
function
regularifand onlyifit assignsprobability1 onlyto logical truths(and
31
32
33
4. Reformulating
theWager
So it is all over forthe Wageras it stands.Can we do better?An ade-
quate reformulationof the Wager must meet the followingrequire-
ments:
RequirementofOverridingUtility
The utilityof salvationmustcompletelyoverrideanyof the other
utilitiesthatenterinto the expected utilitycalculations,thusren-
deringirrelevantthe exact value of the probabilityone assignsto
God's existence.(We impose thisrequirementin orderto uphold
the spiritof the originalargument-for otherwisewe would not
have a reformulationof it,but rathersome quite different argu-
ment.)
Requirement
ofDistinguishableExpectations
We mustbe able to distinguishin expectationoutrightwagering
forGod fromthevariousmixed strategies(based on the coin toss,
die toss,and so on) discussedin section3. In particular,thesmaller
theprobability ofwindingup wageringforGod, thesmallershould
be the expectation,so thatone is rationallycompelled to make
thatprobabilityas high as one can.
I willnow considerfourstrategiesthatI thinkmeet these require-
ments.In each case, a somewhatdelicate balance is struck:on the one
hand, reapingthe benefitof attributingan infiniteutilityto salvation
(namely,the swampingeffectin the expectationcalculationthatren-
34
35
WagerforGod (0 f2
WageragainstGod f3
fl
Assume p is some positive, finite (as opposed to infinitesimal)
7 forGod's existence.The expectationofwageringforGod
probabilityl
is
36
37
38
39
1.p+0.(1- p) =p > 0.
WageringagainstGod has zero heavenlyexpectation,and so itis auto-
maticallytrumped by wageringfor God, whateverpositivevalue p
has-even infinitesimal.
Thus, the Requirementof OverridingUtility
40
41
42
WagerforGod 1 0
WageragainstGod 0 0
43
WagerforGod f f2
WageragainstGod f, f3
fp + f2.(1- p)
44
whileyourexpectationofwagering-against
is
+ f3.(1 - P).
fl.p
fp is so large thatthe formervalue exceeds the latter.
By hypothesis,
Furthermore,a mixed strategy, withprobabilityq > 0 forwagering-
forand probability(1 - q) > 0 forwagering-against,
has expectation
45
46
47
48
6. Conclusion
It seems, then, that Pascal's Wager and all the reformulations of it that
I have considered face a serious problem. Moreover, I believe that it is
a problem that runs deep, not one that will go away with some clever
tinkering.For I see no prospects for characterizing a notion of the util-
ityof salvation that is reflexive under addition withoutbeing reflexive
under multiplication by positive, finiteprobabilities, or reflexiveunder
multiplication by numbers greater than 1 withoutbeing reflexiveunder
multiplication by positive, finiteprobabilities. Yet it seems that nothing
less will salvage Pascal's reasoning. So we are leftwith a dilemma. If the
utilityof salvation is reflexive under both addition and multiplication
by positive, finite probabilities (as in Pascal's original argument),
wagering for God will be just one of many equally rational courses of
action, and our choice among them will be arbitrary.If the utilityis not
reflexive under either addition or multiplication by positive, finite
probabilities (as in my reformulations of the argument), salvation will
be so farfrom being the best thing possible as to be unsuitable for Pas-
cal's theology. Iwager that any futureversion of the argument will suc-
cumb to this dilemma.
CaliforniaInstituteof Technology
References
49
50
51
Notes
Unbounded thanksto the manypeople who have helped me duringthewritingof
this paper, especially:John Barker,John Broome,John Conway,Kevin Coyle,Jamie
Dreier,PhilipEhrlich,Paolo Ghirardato,Bill Harper,Ned Hall, Allen Hazen, David Hil-
bert,Carl Hoefer,RichardJeffrey,JimJoyce,Ehud Kalai, KevinKonty,BrigitteLegare,
ReneeLegare,DavidLewis,PeterMenzies,RalphMiles,AndrewMilne,CollinMosh-
man, Daniel Nolan, Graham Oddie, JohnHawthorne,JimPryor,Gideon Rosen,Teddy
Seidenfeld, Brian Skyrms,Roy Sorensen, Peter Vallentyne,Bas van Fraassen, Peter
Vanderschraaf, JimWoodward,and anonymousreaders forthe Philosophical Review.In
particular,commentsby Alex Byrne,Fiona Cowie, Mahmoud El-Gamal,Christopher
Hitchcock,GrahamOppy,MikeThau, and LyleZyndaled to significant revisionsofear-
lier drafts.(Anotherthankyou to GrahamOppy forintroducingme to Conway'swork,
and to Alex Byrneforsuggestingthe essentialsof the second reformulation.)Thanks
also to audiences at the AustralianNational University, Universidadde Buenos Aires,
University of Bristol,Cambridge University, CaliforniaInstituteof Technology,Carn-
egie Mellon University, Universityof Colorado at Boulder, Universidadde Cordoba,
Indiana University at South Bend,JagiellonianUniversity, Cracow,La Trobe University,
London School of Economics,MIT, Monash University, Montana StateUniversity, Uni-
versityof Melbourne, Princeton University, Queens College (Melbourne), Sheffield
University, St. Stephen's College, Delhi, Texas Technical University, TrinityCollege
(Melbourne), University College, London, and University ofWashingtonat Seattle.Part
of thispaper waswritten whileI was a visitingfellowat theResearchSchool ofSocial Sci-
ences at the AustralianNational University, and I am gratefulforthe generositythatI
was shownthere.
1Pascal's presentationof the Wager is somewhatobscure, even quirkyin various
ways,frustrating a definitivereconstruction. I do wantto insist,however,thatI am adopt-
ing a perfectly standardinterpretation of ?233 in the Pensies,ifcastin the anachronistic
terminology of modernBayesiandecision theory(and thatcastingtoo is standard).
2You maysaythatPascal's conclusionis really"Rationality requiresyou to believe
in
God." But perhaps one cannot simplybelieve in God at will; and rationalitycannot
requirethe impossible.Pascal is wellawareof thisobjection:"[I] am so made thatI can-
52
53
54
55
56