orotheranentire network of interwoven hypotheses
inwhich Hisingredientrather than the separate testing of the component H,(ii) No one constituenthtypothesisH can everbeextricated from the ever-present web of collateral assumptions so as tobeopen to decisive refutationbytheevidence as partofan explanansofthatevidence, justasno suchisolation is achievable for purposes of verification. This conclusion becomesapparent byaconsiderationofthe two partsof theschemaofunavoidably
inconclusive falsifiability, which are:(a) itisan elementary fact of deductive logic that if certain observationalconsequences0are entailed by theconjunctionofHandaset A of auxiliaryassumptions,thenthe failureof 0 tomaterializeentailsnot the falsity of Hbyitself butonlytheweaker conclusion that Hand A cannotboth be true;thefalsifiabilityofH is therefore inconclusivein the sensethat the falsityofHisnotdeductively inferablefrom thepremiss [(H-A)
0],(b)theactual observationalfindings 0',whichareincompatible with 0,allow that Hbetrue while Aisfalse, because they always permit the theoristtopreserveHwithimpunityasapartof theexplanansofO'bysomodifyingA that theconjunctionofHand the revised version A' of Adoesexplain(entail) O'.ThispreservabilityofH is to be understoodasaretainabilityinprincipleand does notdependon theabilityof scientists topropoundtherequiredset A of collateralassumptionsatany giventime.Thus, thereis aningressionof a kindofapriorichoiceintophysical theory:atthe priceof suitablecompensatorymodificationsinthe remainderof thetheory, anyone of its
lhypothesesHlmaybe retained in thefaceofseemingly contrary empirical findingsasanexplanansof thesevery findings.Andthisquasiapriori preservabilityofHissanctionedbythefar-reachingtheoretical ambiguityandflexibility ofthelogicalconstraintsimposed by theobservationalevidence.'Let us now considerthetwoparts (a)and(b)oftheschema which thestronger form (ii)of the Duhemianthesis claimstobetheuniversalparadigmofthelogicoffalsifiabilityinempirical science. Clearly, part (a)isvalid,beinga modustollens argumentinwhichthe antecedentofthe conditional
The Aimand Structure of Physical Theory (tr.P. P.Wiener), Princeton,1954,Part II, chapter VI,esp. pp. 183-190. Duhem'sexplicit disavowalof bothdecisivefalsifiabilityand crucial verifiabilityofan
will not bear K. R. Popper'sreadingofhim[9, p.78]:Popper, who is anexponent of decisive falsifiability, misinterpretsDuhemasallowingthat
tests of a hypothesismay be decisively falsifyingand as denyingonly that they maybe cruciallyverifying. NotwithstandingPopper's exegeticalerror,we shallfindpresentlythat histhesisof the feasibility ofdecisivelyfalsifyingtests can bebuttressedbyatellingcounterexampleto Duhem's categoricaldenial of thatthesis.A defense of theclaim that isolatedpartsof physical theorycanbeconfirmedisoutlined byI-I. Feigl in his "Confirmabilityand Confirmation,"RevuteInternationale dePhilosophieV(1951),pp. 268-279, whichisreprintedin P. P. Wiener (ed.), ReadingsinI'hilosophy of Science,NewYork, 1953, esp. pp.528-529.