Mind Association

Conservatism and Tacit Confirmation

Author(s): Jonathan E. Adler
Source: Mind, New Series, Vol. 99, No. 396 (Oct., 1990), pp. 559-570
Published by: Oxford University Press on behalf of the Mind Association
Stable URL: http://www.jstor.org/stable/2255020 .
Accessed: 29/12/2010 11:58

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=oup. .

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.

Oxford University Press and Mind Association are collaborating with JSTOR to digitize, preserve and extend
access to Mind.

http://www.jstor.org
 and TacitConfirmation
Conservatism
JONATHAN E. ADLER

i. Methodological conservativismholds thatbelievinga statement is a

reasonto continueto believeit. It also favourshypothesesthat are
introduced earlierthancompetitors. Otherconservative theseshavebeen
defended,but thesetwo alreadycontainenoughto motivatethe main
objectionto conservatism. Since neitherbeingbelievednor the timeof
introduction are relevantto showinggreaterapproximation to thetruth,
conservatism amountsto a prejudice,it is alleged,creatingfundamental
epistemic differenceson non-epistemic grounds.
Nevertheless,I wantto defendconservatism. In partI, I arguethata
statement's beingbelievedor introduced earlierprovidesgoodreasonsto
expectit to havetraditionalepistemicadvantages oversomecompetitors,
equallycompatiblewiththe data. Part II takesup the objectionthat
because of underdetermination manyhypotheses will elude the above
defence. Against this objection I argue that if one accepts the
underdetermination thesisas raisinga seriousmethodological problem,
thenwe shoulddistinguish knowing thatthereareequallygoodcompeting
hypotheses fromactuallyproducing them.Underdetermination does not
itselfguaranteethelatter,and thatis whatcounts.
Standardobjectionsto conservatism appealto intuitions,and concep-
tionsofjustification,whichareresponsive solelyto an interest in truthor
correspondence withreality.The difficultyin defending conservatism is
thatitmustconcedeitsfailure on thesecounts.In PartIII itis arguedthat
thisfailuredoes notfallon it differentially.
Conservatism's claimto our
allegianceshifts
thentoshowingthat,forall itsdefects, itremainsthebest
policywe know.

themainclaimin partI is thatifenquiryis functioning

2. Briefly, well,
statements protectedand preferred by a policyof conservatism are
reasonablyexpectedto havesignificantly
greaterconfirmationthanthose
competing statementsnotso protectedor preferred.
For ease of discus-
sion,I concentratemainlyon thebelief/non-belief
conservativism thesis.
According to theQuine-Duhemthesis(i), statements
arenottestedin

Thanksto L. Jonathan
Cohen,CatherineElgin,Roy Sorensen,and especiallyGeorgesRey for
helpfulcomments.
Mind,Vol.99 . 396 . OctoberI990 ? PressI990
OxfordUniversity
 E. Adler
560 Jonathan
isolation.'A secondthesis(2) of confirmation is thata hypothesis is
'confirmed to theextentthatitrantheriskofbeingfalsified'.2 Putting(i)
and(2) together, we inferthatmanymorestatements receiveconfirmation
froma successful testthanthefocalhypothesis underinvestigation. Those
hypotheses we actuallybelieve,unlikethosehypotheses thatmightequally
well have been believedon the same grounds,serveas the essential
background forderivingpredictions. These background beliefsare then
(thesisi) subjectto revisionshouldthe prediction fail.So shouldthe
predictionnot fail they are confirmed, accordingto thesis 2 (with
qualifications we cometo (sections3 and4)). Of course,thesebackground
beliefsare not viewedas beingtestedat thesetimes,so the resulting
confirmation is a mere by-productof enquiry,going unnoticedand
unrecorded. For thatreason,thiskindofconfirmation willbe called'tacit
confirmation'.
Consideran illustration. We come to believethatP=Jones' murder
tookplace in the livingroom,on the basis of certainobservations and
assumptions. In comingtobelievethatP wethereby ruleoutthatQ= The
murdertookplacein thesecretpassageway betweenthelivingroomand
thebasement.Althoughwe do notappreciateit,Q is equallycompatible
withtheevidenceforP. We hypothesize H thatRoe is themurderer. From
H together withP, otherwell-established facts,and generalassumptions
we drawpredictions suchas thatthemurdertookplacearoundmidnight.
These turnout to be the case, so thatH is confirmed. Should the
predictions fail,H is disconfirmed.
The Quine-Duhemthesis (i), focusingon the lattersituationof
falsification,remindsus thattheinconsistency in thesetofbeliefscan be
removedotherthan by rejectingH. In openingup the possibility of
distributing falsificationamong the otherstatements involvedin the
prediciton,the Quine-Duhem thesisrendersmany more statements
susceptible As a result,given(2), manymorestatements
to falsification.
areconfirmed shouldthetestbe passed.In particular, ifourbeliefthatthe
murdertookplace in the livingroom(P) is threatened withrejection
should any predictionfail, it is then tacitlyconfirmed when these
predictions succeed.WhileQ is compatiblewithall the evidence forP
untilthepointwhereP is accepted,Q does notplaya role,as P does, in
deriving testableconsequences. Sinceitdoesnotthenruntheriskof being
it is nottacitlyconfirmed.
falsified, Of course,it is possiblethathad Q
beenacceptedinsteadofP, it couldhaveled to thesamepredictions as P
or to a different thatare as wellconfirmed.
set of predictions While this

1 PierreDuhem, The Aimand StructureofPhysicalTheory,Princeton,I954, ch.VI; W. V. 0.

Quine,'Two DogmasofEmpiricism',in Froma LogicalPointof View2ndedn,rev.,Harvard,I980,
pp. 20-46.
The Anatomy
2 Israel Scheffler, of Inquiry,Knopf,i963, p. 280 quotingp. 50 of David Pears,
his claimto the'generalhypothetical'.
Analysis,I950, pp.49-63. Pearsrestricts
'Hypotheticals',
 Conservatism
and TacitConfirmation
56I

possibilitycannotbe ruledout, it becomesprogressively less likelyas

enquiryproceeds.3
3. The conservatism I amdefending assumesthatenquiryis functioning
well(in thevicinity ofthehypothesis in question).To be morespecific the
assumption amountsto the following: (a) We obtainbeliefs,especially
beliefsthatwereformerly hypotheses, purposefully, neither randomly nor
willy-nillynorjustforthesakeofhavingmorebeliefs.(b) The testingof
hypotheses is on-going.In this processno empiricalstatements have
specialexemption fromfalsification. (c) Thereis a tendency towardgreater
unity,integration, or coherenceamongthecorpusof beliefs.
Our beliefsregularly riskfalsification onlyto the extentthat(a)-(c)
hold.If we obtainedbeliefswithoutregardto theirsubsequentpractical
andtheoretical value,thenwe wouldnothavea righttoinferfromthefact
thatsomestatement is believedto the conclusionthatit is likelyto be
involvedin activities thatlead to predictions. Condition(b) impliesthat
the two crucialfeaturesof enquiry-testingand falsifiability-which
promisetacitconfirmation arenotreserved forspecialtimesorhypotheses.
These two featuresare necessaryforenquiryto be self-correcting. If
enquiryis self-correcting, then the survivalof a belief in enquiry
demonstrates its worthiness of beliefbeyondits originalevidentialbase.
(That is but anotherway to stateour main claimin part I). Greater
unification(condition (c)) increasesthepotential becauseit
forfalsification
is through linksto hypotheses at theforefront oftesting thatbackground
beliefsplacethemselves at risk.
The assumptionthat(a)-(c) hold may be rebuttedor weakenedin
specificcasesorin a moregeneralway.Witha pseudo-scientific belief,for
example,theassumption wouldbe rebutted byshowingthatthebeliefhas
beenprotected fromfalsification in variousad hocways.These beliefsdo
notthenincreaseintacitconfirmation, despitelongevity. In a generalway,
the assumptionis weakerfor individualbelieversthan for scientific
practices.The latteraloneis committed to a single-minded pursuitofthe
improvement ofknowledge, especiallythrough testing.
4. An obviousobjectionwillnowbe addressed:ifan empirical statementis
believed(or accepted)thenitsprobabilityis i, and so it is noteligiblefor
confirmation (as a changein probability).
The mainrationaleforthe acceptanceof hypotheses is thatit ends
enquiryintotheirtruth,and allowsus to investenergyand resources

3 I assumethattheconfirmation relationsnecessary wouldbe validatedin an

fortacitconfirmation
acceptableinductivelogic.Withina qualitative
confirmationtheory,theassumption is problematic.As
Hempelshows,'theconverseconsequence condition'togetherwithotherconditionslead to triviality.
Whilehe thenrejectsthe converseconsequenceconditionas a generalconditionof adequacy,he
readilyallows that thereare manysound examplesof confirmation that are in the converse
consequence relation.CarlG. Hempel,'Studiesin theLogicofConfirmation', in Aspects ofScientific
Explanation,The Free Press,I965, pp. 3-5I. See section8.
 E. Adler
562 Jonathan
elsewhere.As Peircewrote,'When doubtceases,mentalactionon the
subjectcomesto an end.'4Whatwe haveto challengeis theassumption
thatonce we acceptor believea hypothesis, or once in Peirce'sterms
'mentalactionon thesubjectcomesto an end',therecannotbe epistemic
gainsakinto confirmation.5
While the assignment of probabilityi to acceptedhypotheses is a
helpfulconvenience in an inductivelogic,allowing,forexample,much
ofconditional
simplercalculation probabilities, it represents
no formalor
conceptualtruth.The Peirceanrationaledoes not demandit (we cease
doubtingstatements whoseprobability is less thani), nordoes it follow
from,say, coherence('Dutch Book') arguments.Most importantly,
acceptancetakesplace priorto havingreachedcertainty. Even aftera
hypothesishas been assignedprobabilityi, subsequentrejectionis
possible.6
Once thestrategic natureofacceptance is recognized,
roomremainsfor
co-existencewith tacit confirmation, given our limitedpurposesin
introducing that notion.Conservatism establishesasymmetries among
largeclassesofstatements intermsofsimpleandreadilyidentified features
(e.g. believedornotbelieved)thatoffer a goodreasontoexpectgreater tacit
confirmation. We can have good reasonsfor holdingthat believed
statements havegreater tacitconfirmation thanstatements notso believed
withoutknowinganything aboutactualchangesin confirmational value.
We inferit basedupondifferences ofrolein enquiry.
It is a good thingtoo thatour appeal to tacitconfirmation is only
inferred and comparative, forwe can thensteerclearofthecomplications
ofinductive logic:degreeofconfirmation and relatednotionsofquantita-
tiveconfirmation. In particular,
as statements are morelikelytargetsof
falsificationfroma test,theyreceivegreaterconfirmation shouldthetest
be passed.Sincewhatwe believeis usuallynota targetforfalsification, its
gainintacitconfirmation wouldbe muchlessthantheconfirmation forthe
hypothesis thatis actuallybeingtested.
In generating epistemicasymmetries amonghypotheses, conservativism
is responsiveto theirdifferent roles in enquiry.It is because it is
reasonableto expect,on the basis of these different roles, reliable
differences in tacitconfirmation,
thatwe requireno special'mentalaction'
to justify our conservativepreferences. To be conciliatory,we can grant

4 p. I4, C. S. Peirce,'The FixationofBelief',in EssaysinthePhilosophy

ofScience,Bobbs-Merrill,
I957, pp. 3-30.
S One can viewthe'Paradoxof Ideal Evidence'as supporting theclaimthatrationalchangesin
confidenceneednotbe reflected in changesin probability. See Karl Popper,TheLogicofScientific
Discovery,Basic Books,I959, appendixix.
6 Fora defence ofacceptance, seeHenryE. Kyburg,'The RuleofDetachment in InductiveLogic',
in I. Lakatos (ed.), InductiveLogic,North-Holland, I968, pp. 98-I65. For a different
view,see
RichardC. Jeffrey 'Dracula meetsWolfman:Acceptancevs. PartialBelief, in M. Swain (ed.),
Induction,
Acceptance, andRationalBelief,D. Reidel,I970, pp. I57-85.
 563
and TacitConfirmation
Conservatism
thatoncea hypothesis is acceptedno furthergaininconfirmational valueis
possible.Whatwe insistupon,againstacceptancetheorists, is thatthe
absenceofeffort ininvestigating whatwe believedoesnotimplythatthese
beliefsalways retainthe same epistemicstatusrelativeto eachothernor,
morepertinently, thatthey retainthis same statusrelativeto prioror
possible,unrefuted, competitors.
Tacitconfirmation is usefulforotherepistemological purposes,requir-
ingneither challenge toacceptance norinvolving
theorists thecomplexities
ofquantitative inductive logic.Thus thestatusoffoundational beliefssuch
as 'The earthexistedformorethan2 weeks','There are people',or
'Objects do not vanish'has appearedproblematic becauseno specific
evidencejustifies themand theyare nevertested.So, it is alleged,they
constitutethescaffolding forourconceptual system, thatcannotbe (and,
perhaps,need not be) justified.Rather,they are groundlessbeliefs
representing a decisiontoadopt,oran acquiescing to,a framework. Sucha
viewtooeasilyinvitesscepticalandrelativist conclusions.In virtually all of
our predictions, these beliefsprovidethe backgroundassumptionsor
presuppositions. If any of these are wrong,and wrongotherthan
accordingto scepticalscenarios-anassumption thatinductionrequires
anyway-ourworldwouldbe radically differentfromhowitappearstous.
(It remainsan openquestionwhether we wouldknowwhatto concludein
such bizarrecircumstances.) Since we havenot noticedanywidespread
failuresof our mainstream predictions,we have a vasthistoryof these
(basic) beliefsfailingto be subjectto doubt,so, as a result,theyreceive
overwhelming tacitconfirmation.

II
assumeH andH' areequalin
toconservatism:
5. Recallthemainchallenge
all epistemic exceptthatH is believed,but H' not (or H is proposed
respects
earlierthanH'). Is themerefactofbelief(or age or systemic
importance)
alonesufficientto groundan epistemicpreference? so far
Our argument
has deniedtheinitialassumption.The pointofthatdenialis to question
thepresupposition mustbe distinctfromothertradi-
thatconservatism
tional epistemicvirtues.
However,the appealto tacitconfirmation willnot help us whenthis
challengeis combined, as it oftenis, withtheunderdetermination thesis.
For underdetermination guaranteesthat the assumptionof equality,
including equalityin tacitconfirmation, is oftenmet.
The underdetermination thesisthatis mostrelevantto our topicis
the mild one thatclaimsthathypotheses are underdetermined by the
total relevantevidenceavailableat a given time. 'Transientunder-
determination' (UT), unlike radical underdetermination, allows for
 E. Adler
564 Jonathan
subsequentevidencebeingfoundthatselectsbetweenhypotheses equally
goodon theprevioustotalrelevant evidence.7
Hypothesesposited by TU elude our above defencebecause by
stipulationtheyareequal to ourconservatively preferred hypothesis on all
counts.(Note thatwe continueto treat'hypotheses' as covering notonly
conjectures,butalsostatements believedoraccepted.)Specifically, among
thewaysthatsomeH' can be positedas equal to ouracceptedH mayjust
as well be equalityin tacit confirmation. Our responseto this main
challengeto conservatism turnson thefactthattacitconfirmation is only
tacit.Actuallygenerating thematching hypotheses willthenbe burden-
some8 as we would need to formulate explicitlywhat is now, with
conservatism,takenforgranted. For anyalreadyadmitted hypothesis, the
burdenoffinding thematching hypothesisincreasesas enquiryproceeds,
sinceitstacitconfirmation willgrow.
For TU, unlikeradicalunderdetermination, thematching hypothesis is
chosensubsequentto theevidencebeingfixed.We mightput thepoint
this way: underdetermination guaranteesthatthereare such (endless)
matching hypotheses ontologically,notepistemologically.
Whenwe have onlyexplicitly statedevidenceto capture,as withthe
curve-fitters
problem,the matching hypotheses are readilygenerated in
well-known ways.However,tacitconfirmation is not,andimportantly not,
based on explicitevidence,so no additionalinformation is gathered,
evaluated,or stored.
In orderto generate thematching hypotheses we wouldthenneed,for
eachfavoured hypothesis, elaborateinvestigationsthatconservatism neatly
sidesteps.9We wouldhave to determine theevidencethatprovidedthe
tacitconfirmationforthefavoured hypothesis.Such investigations might
notevenbe capableofdeterminate answersas theyinvolvecounterfactual
questionsas to how threatened a hypothesis wouldhave been if the
hypothesis beingdirectly testedhadfailedthattest.(In theJonesmurder
case, we wouldneed to knowat each stagewherea prediction is made
utilizingour beliefthatthe murderwas done in the livingroom,how
threatened thatbeliefwouldhavebeeniftheprediction had failed.)
6. Now it mightbe grantedthatgenerating the matching hypothesis
involvesextremely high,uncompensated cognitivecosts.But, it will be
added,thatworryis merelyprudential or practical,not (epistemically)
principled.
Perhapsso. But it is worthobserving thatourappealto cognitive costs
7 The phrase'transientundeterdetermination' is takenfrompp. 380-I of LawrenceSklar's
important article,'Methodological Conservatism',
Philosophical
Review,I975, pp. 374-400.
8 See Hilary Putnam,Mind, Languageand RealityPhilosophical Papers, vol.2, Cambridge
University Press, '975, pp. 25-6; Christopher
Cherniak,MinimalRationality, The M.I.T. Press,
I986, especially
ch.3, 5, and 6.
9 The argument makesthestandard assumption
thatonlyprojectible hypothesescanbe confirmed.
See NelsonGoodman,Fact,Fictionand Forecast, Bobbs-Merrill,I965.
 and TacitConfirmation
Conservatism 565
or limitsis significantly weakerthan the corresponding appeal within
recentnaturalistic In thoseepistemologies
epistemologies. thelimitsand
costs invokedare those of individualhuman information-processing
systems.'0So theirappealto costsand limitsto justify economiesgained
through strategies likeconservatism, onlyholdsfora veryrestricted setof
knowers. For beings(or groups)withmuchgreatercognitive capacities,
thesegroundsare correspondingly less forceful.
The costsandlimitations assumedbyourargument, bycontrast, areno
greaterthanthoserequisiteto takingthe underdetermination problem
Thatproblemis takenseriously
seriously. underthebroadestidealizations
traditionalin therationalrecontruction of science,whichmakethebare
assumption thatenquirersare finitein timeand resources.
Ourwarrant forconservatism holdsundergreatvariations in conditions
or capacitiesof knowers, and henceundergreatvariations of costsand
limits.It is thenless bound by contingent featuresof our cognitive
capacities.So, on thatcount,itleansmoretowarda principled ratherthan
a merelypracticalor prudential grounding, and muchmoreprincipled
thanin recentnaturalistic epistemologies. Moreover,ifin factthereis no
realalternativetoassuming somelimitsandcosts,thenitis as principled as
is epistemicallypossible.

III
troublesome
7. One typeofcase whichhas beenacceptedas particularly
forconservatism envisagestwo separatelyevolvingsystemscontaining
beliefsH and H'. The twosystemscomeintoconfrontation.
conflicting
The questionis whether shouldfavoureachbeliefwithinits
conservatism
ownsystemand ifso, why.1l
To beginwithwe mustdistinguish betweentheconfrontationarising
fromsuccessin actualenquiryas opposedto imaginedenquiry.Onlythe
formeris worrisome.
The assumptionof well-functioning enquiryimpliesthattestswere
devisedso as to maximizethechancesoffalsification againstcompetitors.
Shouldsuchtestsbe devised,naturehas a voicein hypothesis selection.
Consequently, the factthata beliefhas survivedin on-goingenquiry
reflects
its withstandingcriticalcontrols;it is corroboratedcomparedto
competitors.Not so the survivors of merelyimaginedenquiry.For the
survivalofanyonehypothesis, comparedto competitors, is notdue tothe

10 See, forexample,GilbertHarman,Changein View,The M.I.T. Press,I986, ch.4 and 5.

11 See D. Goldstick,'MethodologicalConsrevatism', Quarterly,I97I,
AmericanPhilosophical
pp. I86-9I; and 'More on Methodological Conservatism', Studies,I976, pp. I92-4;
Philosophical
Sklar,'MethdologicalConservatism', op.cit.;Mark Kaplan and LawrenceSklar,'Rationality and
Studies,I976, pp. I97-202.
Truth',Philosophical
566 Jonathan
E. Adler
devisingof testsso thatnaturecan choose. Rather,we just assume,
additional to assumingthatenquiryis good,thatnaturehas favoured that
hypothesis. Clearly,though,any competitor can, on the same (purely
imaginary) basis,be hypothesized to survivethetestsinstead.
But ifwe putasideimaginedenquiry, it shouldbe recognized thatthe
abovesituation ofequallysuccessful, competing systems is hardlylikely.It
dependsupontherarecoincidence ofbothsystems havinggonefarenough
so thattheirrespectivecompetingbeliefsare highlyand equallywell
justified,whileyetneitherhas a basisforchallenging thegroundsof the
conflicting belief.Shouldeitheroftheseconditions notbe met,conserva-
tismwouldnot be invoked.For, eitherwithineach systemtherelevant
beliefswouldno longerbe justified, and so the existenceof conflicting
claimsis readilyadmitted, or,ifthelattercondition holds,eachsystem has
reasonto denigrate thebasisoftheconflicting belief.
The unlikelihood ofthis'perfect match'scenariois relevant becausethe
underdetermination problematic impliesthatwhatever policyis adopted
somehypotheses mustbe eliminated without investigation. The necessity
of some arbitrariness in any relevantpolicy is less offensive if its
anticipated errorsfalloverthosecasesleastlikelyto arise.
In anyevent,we rejectthe commongroundthatconservatism must
rulein theenvisagedsituationof a perfectmatch.For in thatsituation
the (transitory) underdetermination problematicis diminished.The
difficulty hereis notthatthereare too manymatchinghypotheses, but
onlyone or, at most,a few.(Positinga greaternumberof alternatives
wouldeithermakethesituationmorefantastic, so fallingundermyfirst
response,or so weakentheevidenceforanyone hypothesis as to lead us
to justsuspendjudgement). If thatis so, thenthesematching hypotheses
do notpose anyobstacleto testingor enquiryofanything likethesame
magnitudeas the (methodological) underdetermination problem.Test-
ingis feasible,sincetheequal confirmation is fora propersubsetofthe
totalpossibleevidence.
So ifthequestionis whether thebeliefthatH is justified-understood
as thequestionofwhether thereis anyreasonable groundfordoubtthatH
or anysurviving seriouscompetitor to H-then theanswerseemsto be
'Yes' evenfromwithintheagent'sownsystemof beliefs.For fromeach
agent'spointofview,theother'sbeliefhas play-ed and surviveda similar
riskyroleinenquiry, anda goodenquirer wouldhavesomerespectforthe
judgements ofothergoodenquirers.
But we mustdistinguish betweenthe questionof whether an agent's
beliefis justifiedfromwhether he shouldrationally continueto believeit.
Havingsomereasontodoubthisbelief,whichimpliesa negative answerto
theformer question,is compatible withan affirmative answerto thelatter
question.On this the criticshould agree,since if the questionsare
conflated an anomalyfora theorywouldrequiretherenouncing of that
 567
and TacitConfirmation
Conservatism
theory.But thereis generalagreementthat a (so farsuccessful)theory12
should be retaineduntila viable alternativecomes along. This agreement
One should accept the
is in thespiritofconservatismabout belief-revision.
loss of surrendering
cost in destabilizationand informational a beliefonly
if a compensatingbenefitis offered.13
These argumentsconcerningthe perfectmatchscenarioremindus that
conservatismconcerns change of belief with new information.The
objection and difficultyunder discussion diverts our attentionto the
synchronicquestion of justificationat a particularmoment.By freezing
our gaze at a particular moment in an enquiry-thereby implicitly
construingconservatismas susceptibleto counter-example-we obscure
conservatism'slimitedambitionto realize the most (not complete)overall
success, assessed over conditionsexpectedto arise.
8. It is sometimessuggestedthatconservatismis a barrierto innovation.If
this could reallybe shown, the objection would be serious. Others have
givenreasonto doubt the forceofthisobjection.'4 Our own limitedclaims
forconservatismgo some waystowardunderminingthisline of objection.
For us, conservatismis more justifiablyinvoked in better enquiries,
overridableforgood reason,and does not usurptherole oftestingbetween
genuinecompetitors.
What concernsus here is that the objectionas usually presenteddoes
notreallyofferan alternativepermittinggreaterinnovation.Thus consider
these suggestionsdirectedagainstconservatism:
... Sometimesshiftingto a new framework
proves highlyfruitful....'5
... The overthrow ofacceptedopinionandthedictatesofcommonsenseareoften
essentialto epistemicadvance.Moreover,an epistemic adventurermayarriveat
beliefsthatare notonlynewand revelatory, but also better thanthose
justified
morecomfortably heldby others.16
The emphasizedterms'Sometimes' and 'may' suggestthat the objection
ofconservativism,
actuallybeingraisedis to the (readilygranted)fallibility
masqueradingas a concernforconservatismas a barrierto innovation.
What givesthesecriticismsintuitiveforceis thattheyimplicitlyposit a
backgroundthatis preferablebecause moreneutral:judge each case on its
merits;do not exclude a frameworkwithoutexamination(since it may

12 Responding to a worryvoicedby Sklar,'Methodological Conservatism', op.cit.and Michael

Slote,'Confirmation and Conservatism', American PhilosophicalQuarterly, I98I, pp. 79-84.
13 For discussion see Sklar,'Methodological Conservatism', op.cit.;Harman,Changeof View,
op.cit;Cherniak, MinimalRationality, op.cit.,and Levi, TheEnterprise of Knowledge, The M.I.T.
Press,I980, ch.2 and 3.
14 See Goodman,Fact, Fictionand Forecast, op.cit., ch.4, sec.4.; and Harman,op.cit.,ch.7.
15 p. 340, my emphasis,Alan H. Goldman,'The Force of Precedentin Legal, Moral, and
EmpiricalReasoning',Synthese, I987, pp. 347-54.
16
p. I84, firstemphasismine,KeithLehrer,Knowledge, OxfordUniversity Press,I974.
 E. Adler
568 Jonathan
provebetterthantheone we have);do notexcludewithoutexamination
riskier models(fortheymayarriveatbetterjustified
inferential belief).But
thisbackground is inadmissible, oncetheunderdetermination problemis
accepted.For thereare innumerable alternativeframeworks or riskier
inferentialpoliciesthatare unrefuted at any giventime,and so any
plausiblecandidateforhandling thatproblemmustexcludesomeofthese
withoutexamination.
If,however, theobjectionis meanttosuggestthatifspecific groundsare
offered thata newframework willprovefruitful,orthatan adventuresome
approachwillbe moreworthwhile in a particularcase,thenthisis advice
compatible withconservatism. A goodreasonforoverruling conservatism
in thatcase has beenoffered.
If,totryanother interpretation,theclaimis thatoccasionally we should
shiftor be adventuresome, this mightvery well work. But it has
disadvantagesas a generalpolicy comparedto conservativism. The
disadvantages areonthedimensions ofeaseanduniformity oflearning and
application.We adopt(whether through evolutionor sociallearning)the
policyof conservatism withoutdifficulty or notice;and the policyis
7 It justinvolves
appliedeffortlessly.' ourneither proposing norexamining
numerous(wild)alternative hypotheses More-
(or revisionpossibilities).
over,we all pretty muchagreeon our conservative preferences.We can
thusexpect(broad)convergence on whatis worthinvestigating and what
is obviouslynot,(or whatrevisionsare admissibleand whichare to be
ignored). Enquiriescanbe betterco-ordinated, hencemoreproductive and
less redundant.
Those who makethe abovecriticisms do not tellus how theymight
capturetheseadvantages; nor,ifnot,whatbenefits theirproposalsmight
offeras compensation. Primafacie,though,being'new' or 'adventure-
some'byitselfarenotpolicieswhich,ifadopted,promiseconvergence or
agreementamong investigators. A policy of 'adventuresomeness' or
'newness'aimsat novelty anddiversity. Bothproposalsmaybe valuableas
occasionalvariations on a basicallyconservativepolicy.Butsuchvariations
acceptconservatism as basic,and so do notconstitute objections.
9. The outcomeof thislast sectionis thatthe impliedcriticism in the
question'If twohypotheses
rhetorical are equal on all epistemiccounts,
exceptfor the featurepickedout by conservatism, then can one be
preferable
epistemically to the other?'is as misguidedas the parallel
questionforthe policywhichpicksout onlythoseholdinga Ph.D. as
candidatesfora job: 'If twocandidatesare equallygood on all relevant

17 W. V. 0. Quineand J. S. Ullian,The WebofBelief,RandomHouse, I978, p., 67. Quineand

is found.Space does not
Ullian are concernedmainlywithbeliefrevisionwhenan inconsistency
permit metodealwiththattopicdirectly.However,thesamedefenceofconservatismapplies.Forthe
areeachcapableofrestoring
revisions
problemofbeliefrevisionis thatinnumerable
first consistency.
 569
and TacitConfirmation
Conservatism
counts,exceptthatone has the Ph.D. and the otherdoesn't,does that
showthatone has moreacademicmeritthantheother?'
Of course,thequestionremainsas to whether conservatism, whatever
its epistemicand practicalmotivation, failsto accordwithour (realist)
ideals forepistemicjustification. (Similarly,whateverits warrantand
practicalnecessity, we can stillask whethera policywhichexcludesall
thosewithouta Ph.D. degreedoes yieldthecandidatewiththegreatest
merit.)Those ideals are met whena statement's acceptanceis shown,
roughly, to be forreasonsthatreflectsolelyan interestin truth.Butifthat
wereour sole interest, therewouldalwaysremaincompetitors equally
good(in thoseterms)as theoneaccepted,thathadnonethelessnotbeen
accepted.18
But it is the underdetermination problemitselfwhichgeneratesthe
scepticalworry, nottheconservatism solutionin particular.
Whatwould
be distinctivelyor selectively
relevantto conservatism wouldbe ifitcould
be shownthatthespecific featuresthatconservatism picksoutexacerbate
our deviationfromthese ideals intolerably. The criticismswe have
surveyedare marshalledto drawthisconclusion,but thesearguments
neverreallyprovidegroundsmorespecificthanthegeneralworryabout
reconciling underdetermination and realism.
Moreover, thisscepticalworryis too broadin anotherway.It failsto
acknowledge justhowmuchconservatism concedesto theserealistaspira-
tions.Conservatism allowsthat,ideally,choiceamongcompeting hypothe-
ses shouldbe decidedby testing. Underdetermination rulesout complete
fulfillmentof thisideal,so thequestionbecomeswhatwaysare available
whichdeviateminimally fromthisideal.19As we sawin theabovesection,
conservatism need not usurpthe role of testingbetweenactual,equally
successfulcompetitors withintheirrespective systems ofbelief.
io. Our arguments challengedefencesofconservatism forconcedingtoo
muchto itscritics.The defencesaretoonaturalistic, failingto appreciate
traditionalepistemicvirtues(i.e. tacit confirmation) of conservatism.
Criticsof conservatism have framedthe problemas one of whethera
statement's being believed(or proposedearlier)can itselfprovidea
rationalgroundforcontinuing to believeit. That framepresupposes that
thereasonsconservatism for
offers continuing to believemust be separate
fromotherepistemic Defenders
virtues. ofconservatism shouldrejectthat
wayof framing theissue.

18 For recentdoubtsaboutthisgeneralworry see W. V. 0. Quine,'On empirically equivalent

systemsoftheworld',Erkenntnis, 1975, pp. 313-28.
'9 Simplicitysimilarlysuffersfromthedemandfora realistjustification. Space does notpermit
but in focusingon conservatism
discussionof simplicity, I do not mean to excludethe role of
The approachto simplicity
simplicity. thatis closestto ourdefenceofconservatism is thatofRobert
as Fall-Out',in HowManyQuestions:
Nozick,'Simplicity EssaysinHonorofSidneyMorgenbesser, L.
Cauman,I. Levi, C. Parsons,and R. Schwartz, HackettPublishingCo., I983, pp. 105-19.
 E. Adler
570 Jonathan
We can then divide up criticsof conservatism along two dimen-
sions-thosewhodemanda realistjustification or not;thosewhoaccept
methodological underdeterminationor not.To thosewhoreject(metho-
dological)underdetermination,thispaperhas nothing to say.We tookthe
seriousnessof thatproblemas established.To those who accept the
underdetermination their
problem,but insistupon a realistjustification,
objectionsare not worrisome to conservatism, as we have defendedit,
differentially.
Finally,forthosewhoacceptunderdetermination and do notdemanda
theburdenis on themto providean alternative
realistjustification, that
worksbetter.To workbetterit will need to do betteron balance(or
provideoff-setting advantages)in ease and uniformity of learningand
application;as well as in the selectionof a (non-evidential) basis for
preference betweencompeting hypotheses thatis morereliableand more
efficientin lightoftheexpectedcourseofenquiry.20

ofPhilosophy
Department JONATHAN E. ADLER
College,CUNY
Brooklyn
Brooklyn,
NY II2IO
USA

20 Paul Feyerabendis exemptfrommy criticisms of opponents(and many defenders)of

conservatismfortheirmisrepresenting demand.He triestoshowin detailwhy,
theproperjustificatory
scientific
and withcase studiesthat,farfromfacilitating progress, is a barrier.
conservatism Further,
Feyerabendattempts to defendat lengtha genuinealternative. See, forexample,his 'How to Be a
Good Empiricist-APlea forTolerancein MattersEpistemological', in B. Brody(ed.),
reprinted
ReadingsinthePhilosophy ofScience,Prentice-Hall,
I970, pp. 3I9-42; andAgainst Method,London,
New Left Books, I975, especiallyhis discussionsof 'the consistency condition'.Examininghis
argumentswouldrequirea lengthy and separatetreatment,and hisworkhas alreadybeenthesubject
ofextensivestudy.