International Phenomenological Society

Knowledge, Evidence, and Skepticism According to Williamson

Knowledge and Its Limits by Timothy Williamson
Review by: Anthony Brueckner
Philosophy and Phenomenological Research, Vol. 70, No. 2 (Mar., 2005), pp. 436-443
Published by: International Phenomenological Society
Stable URL: http://www.jstor.org/stable/40040803 .
Accessed: 04/03/2013 05:53

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.

International Phenomenological Society is collaborating with JSTOR to digitize, preserve and extend access to
Philosophy and Phenomenological Research.

http://www.jstor.org

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
 Research
Philosophyand Phenomenological
Vol. LXX, No. 2, March2005

Knowledge,Evidence,andSkepticism
toWilliamson
According
ANTHONY BRUECKNER
of California,Santa Barbara
University

1. Introductory
Accordingto TimothyWilliamson,where'E' standsforevidenceand 4K'
standsfor knowledge,E=K.! He arguesfor the followingtheses,which
jointlyimplyE=K:

(1) All evidenceis propositional.

(2) All propositional
evidenceis knowledge.
(3) All knowledge
is evidence.

In thispaper,I will arguethatWilliamson'sviews on knowledgeand evi-

dencegenerate an unacceptable of perceptual
conception knowledge andjusti-
fication.The most plausibleway out of the difficulty
ruinsWilliamson's
case againstskepticism.
2. What is evidence?
In ordertobetterunderstand
E=K andthetheses(2) and (3), it will be helpful
to startwith(1). Accordingto Williamson,a bloodyknifeis not evidence.
He agreeswiththe"familiar claim thatonlypropositions can be reasonsfor
belief,andheputsforward severalconsiderationsaimedat makingplausible
theview that"onlypropositions thatone graspscan function as evidence".2
(194) For in
example, inference to thebestexplanation, one seeksto formu-
late hypothesesthatprovidegood explanations forone's evidence.Explana-
tionscan be put in theform'- because...', and only declarativesentences,
expressing can
propositions, be grammatically substituted
in theblanks. This

1
See hisKnowledgeand itsLimits(Oxford:OxfordUniversity Press,2000), chapter9. All
page referencesin thetextare to thisbook.
2
The "familiarclaim"can be foundin Donald Davidson's "A CoherenceTheoryof Truth
and Knowledge",in Truthand Interpretation, ed., ErnestLePore (Oxford: Blackwell,
1986). The locusclassicusoftheclaimis WilfidSellars' "Empiricismand the Philosophy
of Mind",in his Science, Perceptionand Reality(London: Routledgeand Kegan Paul,
1963).

436 ANTHONYBRUECKNER

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
suggeststhatwhen an hypothesisexplains one's evidence,both explanandum
and explanans are propositions.
Williamson's strategyis to findroles played by evidence (inferenceto the
best explanation, probabilistic confirmation,the ruling out of hypotheses)
whichapparentlycan only be filledby propositions. Of course, this strategy
is consistentwith the existence of otherroles played by evidence which are
non-propositionalin character.We will returnto this point later,in discuss-
ing non-propositionaljustifiers.
Suppose thatS's belief that P is justifiedby his evidence e. Williamson
holds that e must be a proposition in order to play that evidential role.
Though Williamson does say that S must grasp the proposition e, he ini-
tially says nothingabout whetherS must, further, believe e. It would seem
to be veryplausible to hold thate can functionas S's evidence for believing
P only if S also believes e. If e is merelya proposition that S grasps but
does notbelieve, and perhapsdisbelieves, then why would e count as part of
S's evidence? Williamson does eventuallybring belief into the picture.It is
"granted"by him that knowledge entails belief. (202) Since he maintains
E=K, his overall position has the consequence, which he notes, that S
believes e: e's evidentialstatusforS entails thatS knows e, and that in turn
entails thatS believes e.

3. E=K
Let us digress brieflyand get a betterfeel for E=K. As we just noted, by
E=K, if O is evidencefor S, then S knows O. Conversely, if S comes to
know O, thenO becomes partof S's evidence. O can then be used to justify
furtherbeliefs for S; O can enable S to expand his body of knowledge. In
defendingE=K, Williamson maintainsthata justified,true belief that fails to
amount to knowledge cannot be part of S's evidence. This would support
E- »K. Williamson says,

Ifevidencerequiredonlyjustifiedbelief,thena criticalmassof evidence could set offa kind

of chainreaction.Our knownevidencejustifiesbeliefin varioustruehypotheses; theywould
countas evidencetoo,so thislargerevidencesetwouldjustifybeliefin stillmoretruehypothe-
ses, whichwouldin turncountas furtherevidence....The resultwouldbe verydifferent from
our presentconceptionof evidence.(201)

Undercertainconditions,E=K can also engendera problematicevidential

chain reaction.Suppose that S has evidence e for believing Z (=This is a
zebra): the animal looks just like a zebra, it is in an ordinary-seemingzoo,
and thereis a sign thatreads "Zebra" by its enclosure. Suppose that e enables
S to know Z but does not enable him to know the entailed -CD (=This is not

BOOK SYMPOSIUM 437

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
a cleverlydisguisedmule).3Given E=K, when S comes to know Z, Z
becomespartof S's evidence.Call theexpanded evidencee+. Z is the best
evidencethatS couldhave for-CD, since Z entails~CD. Now S knows
-CD on thebasis of e+, eventhoughhis originalevidencee didnot enable
himto know-CD. Thisis moreof an evidentialspontaneous generation than
an evidential
chainreaction,
butitis problematic
all thesame.4
More generally,E=K licensesevidential
spontaneousgeneration if it can
happenthat(a) one's evidenceenablesone to know O but not *F, and (b)
one's evidenceplusO enablesone to knowXF.
4. Skepticism
In orderto tryto understand Williamson'sview of perceptual knowledge,it
willbe helpfulto discusssomeof hisremarks on skepticism. Let us consider
thegoodcase (G), in which"thingsappeargenerallyas theyordinarily do,
andarethatway",andthebad case (B), in which"thingsstillappeargenerally
as theyordinarily do, but are some otherway". (165) In G, S correctly
believesH (=S has hands).In B, S mistakenly believesH and so does not
knowH; supposethatin B, S is a brainin a vat. Indeed,in B, S mistakenly
believesthathe is in G, and he failsto knowthathe is notin G. Williamson
says(in myfavorite ofhisgraceful "Partofthebadnessof thebad
sentences),
case is thatone cannotknowjusthowbad one's case is". (165)
The skepticwantsto showthatregardless of whether S is in G or in B, S
does not knowH. Williamsonthinksthatall the skepticneedsin orderto
showthisis to establishtheSamenessofEvidenceLemma:

(SEL) S has exactlythesameevidencein G andinB.

Williamsonsays,
The sceptictypically
insiststhatone has exactlythesame evidence in thetwo cases. There-
fore,sinceone believes...[H] withthatevidencein thebad case, believing...
[H] withtheevi-
denceone has in thegood case is insufficient
forthetruthof...[H].(169)

Thisis a ratherunusualtakeon theskepticaldialectic.Manyopponentsof

skepticism grant S's evidencein G does notensurethetruth
that of H, in the
sensethatit is possiblethatS shouldhavethesame evidenceas he has in G
whilein a situationin whichH is false.Such opponentsof skepticismsub-
scribeto thefallibilist
principle:
This is theviewdefendedby FredDretskein "EpistemicOperators"(Journalof Philoso-
phy,1970). He holdsthatS can knowZ without knowing-CD because CD is nota rele-
vantalternative toZ.
PeterKlein endorsesthisview of how -CD becomes knownin Certainty (Minneapolis:
University of MinnesotaPress,1981), and in "Skepticismand Closure: Why the Evil
GeniusArgument Fails" (PhilosophicalTopics,1995). See my"Skepticismand Epistemic
Closure"(PhilosophicalTopics,1985) and "Klein on Closureand Skepticism"(Philoso-
phical Studies,2000) forcriticismoftheview.

438 ANTHONYBRUECKNER

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
 (~E) One's evidence foran ordinaryexternal-world proposition O is such
thatit is possible thatone should have exactly similar evidence in a
situationin which O is false; thus,wherethe evidence in question is
propositional,it fails to entail O.

Typically,(~E) is supportedby the claim that an ordinaryknower's evidence

could be duplicatedin B, in which he is mistaken in his external-world
beliefs. Thus, opponents of skepticism who subscribe to (~E) accept SEL.
So it seems doubtfulthat SEL is sufficientto generatea skeptical conclu-
sion.
Even if SEL is not sufficientfor the skeptic's purposes, it might well be
necessary.One standardway of settingup the skeptical argumentis as fol-
lows:

(A) If S knows H (=S has hands), thenS knows ~SK (=S is not in B).

(B) S does not know ~SK.

So (C) S does not know H.

(A) is supportedby a closure principleforknowledge,such as

(CL) If S knows that O, and S knows that O entails VF, then S knows
that¥.

(B) can be supportedin the following way. First, this Underdetermination

Principleis put forward:

(UP) If S's evidencefor O does not favorO over a competing incom-

patible hypothesisW, thenS is notjustified in believing O.

Suppose thatthe skepticcan establish

(*) S's evidence does notfavor~SK over SK.

(UP) and (*) imply that S is not justified in believing ~SK. Given that
knowledgerequiresjustification,(B) would therebybe established.5
But what is the skeptic's justificationfor (*)? This is where something
like SEL seems to be required.If S has the same evidence regardlessof
whetherhe is in B or insteadin G, thenhis evidence (regardlessof whetherhe
in factholds it in G or in B) does not favor~SK over SK.
Williamson thinksthatSEL is false, and he criticizes what he takes to be
the skeptic's only argumentfor SEL.6 Working throughWilliamson's cri-
tique of that argumentwould take us too farafield. For presentpurposes, I

5 See my"The Structure of the SkepticalArgument" (Philosophyand Phenomenological

Research(1994) ) fora discussionof thiswayof conceivingtheskepticalargument.
6 See sections8.4-8.6.

BOOK SYMPOSIUM 439

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
wouldliketoraisea questionandthenconsider, in thenextsection,William-
son's likelyanswer.Whyisn'titjustobviousthatSEL is true?Suppose that
it is indeedpossibleforsomeoneto be in B in virtueof beinga brainin a
vat.Thenisn'titjustpartof thedescriptionof such a bad case thatthebrain
has experiencesindistinguishablefromthoseof a normalsubjectof experi-
ence, togetherwithbeliefsand reasoningabilitiesthatmirrorthose of an
ordinary subject?7If the vat thoughtexperiment is so constructed,then,
accordingtothecurrent lineofthought,
SEL seemsclearlycorrect:theenvat-
tedbrain'sevidenceforH is justlikethatof someonein G. We will consider
Williamson'slikereplybelow.
5. Perceptualknowledge
Now we can pick up themainthread of the paper.Williamsondiscussesa
pair of cases: S sees a mountain in normalcircumstances and correctly
believesthatitis a certainshape,andS sees a mountainin unfavorable cir-
cumstances and is undertheillusionthatitis a certainshape(it is some other
shape). Whatis S's evidencein each case? The evidencewill consist of
believedpropositions, on Williamson'sview.In thisgood case, S's evidence
is thetrueproposition expressedby his utterance of 'It is thatshape'. Wil-
liamsonspecifiesthisas theproposition thatthemountain is thatshape(call
thisM). But this propositioncannotbe S's evidencein the pertinent bad
case, sinceS mistakenly believestheproposition in thebad case. The propo-
sitionis falsein thebad case, hencenot knownby S in the bad case, and
hence(in lightofE=K) nota candidateforbeingS's evidencein thebad case.
Williamsonsays,
If perceptual
evidencein thecase of illusionsconsistsof truepropositions,
whatare they?The
obviousansweris: theproposition
thatthingsappearto be thatway.(198)

Thus,S's evidenceinthebad case, accordingto Williamson,is theproposi-

tionthatthemountain appearsto be that
shape(call thisMA).
So theremarks at theendofthepreceding sectionabouttheplausibilityof
SEL presuppose a mistakenconceptionof evidence,Williamsonwill say. In
G, thenormalsubject'sevidenceconsistsof propositions like M, whereasin
B, theenvatted
subject'sevidenceconsistsof evidencelike MA. SEL is false
forthisreason,Williamsonwillsay.

Let us considera bad case in whichthecontents of thesubject'sthoughtsare the same as

in thegood case. We supposethatthesubjecthas livedformanyyears in a normalenvi-
ronment buthasjust recently beensurreptitiously
envatted.So we are notconsideringthe
specialized vat hypotheses describedby HilaryPutnamin Reason, Truthand History
(Cambridge:CambridgeUniversity Press,1981),in whichthebrainsin a vat neverhave
anyordinary causal contactwithtrees,hands,and so on. In sucha scenario,an external-
istaboutintentionalcontent mayholdthatthecontents ofthebrain's thoughts differfrom
thoseof a normalthinker.

440 ANTHONYBRUECKNER

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
 Williamson's views about evidenceand knowledgehave problematiccon-
sequences regardingperceptualknowledge. To see this, let us consider my
belief of the propositionthatmy cup is red (call this C). Let us suppose that
this is an instanceof perceptualknowledge. It is naturalto ask whether,for
Williamson, my visual experienceof the red cup plays a role in my knowl-
edge of C. Since experiencesare not propositions, they cannot functionas
evidence forC, on Williamson's view. The most that Williamson is inclined
to grantregardingtheepistemicrole of experienceis thatin some cases,

...the evidence foran hypothesish consists of propositionsel,..., en, which count as evidence
for one only because one is undergoing a perceptual experience e. (197)

Williamson does not explicitlyendorse this view, saying only thatit is "con-
sistentwithE=K". (197)
My belief of C, we are assuming, is an instance of knowledge. Let us
suppose thatmy belief of C is justifiedon the basis of evidence, which, on
Williamson's view, will consist of one or more believed propositions.
Which? In the mountainexample, S's evidence in the good case consists of
the propositionthatthe mountainis thatshape. Apparently,Williamson will
maintainthatin the presentexample, the propositionthatmy cup is red con-
stitutesmy evidence for my belief of C (supposing that that belief is indeed
evidentiallybased). In orderto functionas my evidence,I must believe the
evidentialpropositionin question.8Further,Williamson may grantthat that
evidentialpropositionattainsthe statusof evidenceonly because I am under-
going a visual experienceof the red cup. So now we have arrivedat the fol-
lowing position: my belief of C is justifiedin virtueof my belief of the evi-
dential proposition that my cup is red. That is to say, my belief of the
propositionthatmy cup is red is justifiedin virtueof my belief of the propo-
sitionthatmy cup is red!
This is an unacceptable view of the structureof perceptualknowledgeand
justification.9Further,insofaras Williamson's rejectionof SEL is based on
this view (see the earlierdiscussion of S's evidence in the mountain exam-
ple), his main answer to the skeptic is vitiated.That is, Williamson can no

8 As we saw earlier,Williamson'sE=K entailsthis,since one knowsone's evidence and

knowledgerequiresbelief.
9
It is a curiousfeatureof Williamson'sconceptionof evidence thatif e is part of S's
evidence,thenforS, e is evidence foritself.This is because Williamsonholdsthate is
evidenceforh forS ifand onlyifS's evidenceincludese, and the conditionalprobability
of h givene is greater thantheprobabilityofh. The secondrequirement willbe metif e's
probability is less than1, since Prob(e/e)=l. But thisfeatureof Williamson'soverall
positiondoes nottakethestingout of thecurrentdifficulty. He says of thiscuriousfea-
tureof his view:
Certainly...[thefeature]does not make it trivially
easy to have evidence for
e, fore is evidence foritselfonlyif S's evidence includese. By E=K, that
requiresS to knowe, whichmaynotbe easy.(187)

BOOKSYMPOSIUM 441

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
longerdistinguishS's evidencein G fromhis evidencein B by holdingthat
theformer consistsoftruepropositionsabouttheexternalworldwhilelatter
consistsof truepropositions
abouthowthingsappear.
6. A possiblemove
Even thoughWiliamson'sremarks aboutthemountainexamplewarrant the
attribution
of theunacceptable view to him,perhapshe can avoid it while
maintaining of his overallposition(such as E=K andtherejec-
key features
tionof SEL). One movewouldbe to giveup theviewthatperceptual knowl-
is
edge evidentially based. He would balk at this, given the following
remarks.He raises the question,"...is all justifiedbelief justifiedby
evidence?"(207) He answersas follows:
It is farfromobviousthatanybeliefis justifiedin the truth-directed
sense [relevantto knowl-
edge] withoutbeingjustifiedby evidence. ...E=K supports plausibleequationof truth-
the
directed withjustification
justification byevidence.10(207-8)

However,supposethatWilliamsonacceptsthe suggestedmove: perceptual

knowledgeandjustification are not evidentially
based. Then the questions
arise:Whatis thesourceofthejustification of perceptualbeliefs,andin vir-
tue of whatdo suchbeliefsconstitute knowledge? We saw earlierthatWil-
liamsonis willingto countenance theviewthatin some cases, a proposition
e constitutes evidenceforan hypothesis onlybecausethe subjecthas some
experience e. PerhapsWilliamsoncouldmaintainthatmy experience of the
redcup (i) is whatjustifiesmybeliefofC, andthus(ii) enablesthatbeliefto
amountto knowledge.However,this wouldbe a case of justification and
knowledge that does not involvereasonsforbelief,which,on Williamson's
view,can only be propositions.Further, Williamsonexplicitlyrejectsthe
proposedmove:
The threateningalternative [to my view] is that [experience] e can itself be evidence for h,
withoutthe mediation of any such [evidential propositions] el,..., en. (197)

Even so, Williamsoncouldmaintain,as some have, thatexperiences have

content,in thattheyrepresent
propositional the worldas being a certain
way.11So eventhoughthejustifieris not a proposition,it is nevertheless
proposition-like.

The second sentenceis implausible:E=K is consistentwiththerebeing non-evidential

Whereverthereis evidentialjustification,
justification. one may grant,the evidence is
known;and wherever thereis knowledge,it mayfunction as evidence; butsomejustifi-
cationthatyieldsknowledge, one mayhold,is non-evidential.
See, forexample,JamesPryor's"The Skepticand the Dogmatist"(Nous 2000), Michael
Huemer's Skepticismand the Veil of Perception(Lanham: Rowman and Littlefield,
2001), and Bill Brewer's Perceptionand Reason (New York: OxfordUniversity Press,
1999).

442 ANTHONYBRUECKNER

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions
 This way of avoidingtheunacceptable view of perceptualknowledge and
justificationthatWilliamson'soverallpositionentailshas a majordrawback
forhim.S's experiences in G andin B havethesame propositional content:
in B, S mistakenly believesH (=S has hands)in virtueof hisexperienceas of
havinghands.S's experience inB represents theworldas containing a handed
S, justas does hisexperiencein G. On thesuggestedmove,then,thatwhich
(allegedly)justifiesS's beliefof H in G is his (non-evidential)
experience as
ofhavinghands.In B, S hasthesameexperience as ofhavinghands.G and B
do notdiffer inrespectof S's evidenceforH, sinceS has none,on thesug-
gestedmove. Further, the elementwhichprovideswhatever justification S
has forbelievingH in G is also presentin B. Thus, Williamson'sdenialof
SEL is blockedif he adoptsthesuggestedmove. G and B do not differ in
respectofthatwhichwouldservetojustifyS's beliefof H (if anything does
in factjustifythatbelief).The skepticcan thereforeappealto variantsof (UP)
and(*) whichtreatof justifyingelementsratherthanevidenceand thereby
showthatS is notjustifiedin believing~SK (and thusfailsto know~SK):

(UP*) If S's putativejustifierforO does not favorO over a competing

incompatible hypothesis *F,thenS is notjustfified
in believingO.

(**) S's putative does notfavor~SK overSK.

justifier
So on thesuggestedmove,Williamson'sanswerto theskepticis unavailable
to him.
7. Conclusion
Therearetwo alternatives forWilliamson.Eithermy perceptual knowledge
thatmycup is red(=C) is evidentially based,or it is not.12On thefirstalter-
native,his overallpositionenableshim to stymiethe skepticby denying
SEL. But hisoverallpositionis unacceptable, as itimpliesthatmy beliefof
C is justifiedin virtueof my believingtheevidential propositionC. On the
secondalternative, Williamsonneedsto holdthatmybeliefof C is notjusti-
fiedbyevidencebutinsteadbyexperience. But on thisalternative,he can no
longerdenySEL, and then hiscase the
against skeptic thusevaporates.13

12 Another whichWilliamsondoes notdiscuss,is to optfora coherencetheory

alternative,
on whichtheepistemicstatusof mybeliefof C dependson itsbeingthe memberof a
coherentsystemof beliefs.See, for example,Laurence BonJour'sThe Structureof
EmpiricalKnowledge(Cambridge:HarvardUniversity Press,1985).
13 Thanksto ArashNaraghiforhelpfuldiscussionof theseissues.

BOOKSYMPOSIUM 443

This content downloaded on Mon, 4 Mar 2013 05:53:21 AM

All use subject to JSTOR Terms and Conditions