G. C. STINE

SKEPTICISM, RELEVANT ALTERNATIVES, AND DEDUCTIVE CLOSURE

(Received 2 February, 1975)

Discussionsof ,defined with varyingdegrees of precision,are of courseperennial in philosophy.Some recentdiscussions of the issue1 giveprominence to the notionof 'relevantalternatives', according to which a claimto knowthat p is properlymade in the contextof a limitednumber of competingalternatives to p; to be justifiedin claimingto knowp (or simplyto knowp) it is sufficientto be ableto ruleout alternativesrelevant to that context.This seemsto me to be a correctand hearteningdevelop- ment. Recent epistemologicaldiscussions have also broughtup a rela- tively new subject,which is the validityof the generalform of argument: (A) a knows that p a knows thatp entailsq .,. a knows that q I shall call this the principleof epistemicdeductive closure, or simply,in this paper, deductive closure.2 What is interesting about recent comments on this principleis that it is perceivedto have somethingto do with skepticism- in fact to leadto it - and henceis currentlyof verybad repute. And 'relevantalternatives' views of knowledgevis-'a-vis skepticism are supposedto show us the falsity of the principle. In this paper I propose to do three things. First, to give a qualified argumentfor deductiveclosure. Second, to give a qualifiedargument againstskepticism which will make use of the relevantalternatives . It will be similarto othersin leavingrather indeterminate the way in which the contextdetermines what is takento be a relevantalternative, although I shall distinguishdifferent sources of this indeterminatenessand draw some furtherconclusions. Third, I shall give an unqualifiedargument to the effect that the questionsof the of the principleof epistemic deductiveclosure and skepticismare completely irrelevant to one another, and that in fact properattention to the idea of relevantalternatives tends to confirmthe principle.This, of course,puts me in directconflict with the recenttrend I have mentioned.

PhilosophicalStudies 29 (1976) 249-261. All Rights Reserved Copyright? 1976 by D. ReidelPublishing Company, Dordrecht-Holland

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions 250 G. C. STINE

1. EPISTEMICDEDUCTIVE CLOSURE

I am in principle suspicious of all principles of epistemic on the general grounds that while the logic of a knower who is in some way simplifiedand idealizedmay be usefulfor limitedpurposes, what we are ultimately interested in are actual knowers who can be pretty obtuse and idiosyncratic,yet still lay claim to .For this, among other , I have elsewhere been concerned with epistemic logic which eschews possible worlds semantics imposing strong constraints on knowers.3Certainly, I would rejectthe patternwhich goes:

(B) a knows that p p entails q .. a knows that q

However,the patternwhich I have labeled epistemicdeductive closure does seem to representa certainbare minimum.One looks naturallyfor counter-instances involving failure of where p and q are very com- plicated,but any such case I can imagineturns out to be apparentonly becauseit invariablyraises doubts about the of the secondpremise which are as strongas the doubts about the truth of the conclusion.The principleseems to be on a par with epistemicconjunction, to wit:

(C) a knows p a knows q .,.aknowspandq There have, of course, been problemsin reconcilingthis principlewith commitmentsto rationalbelief in terms of degreesof confirmationand knowledgein terms of rationalbelief,4 but one feels stronglyinclined to the view that the adjustmentmust be made in the area of these commit- ments and not in the principleof conjunction. In additionto failureof belief, one may look for counter-examplesto the principle of epistemic deductiveclosure in the area of failure of evidence or warrant.One's initial reaction to this idea is that if one's evidenceis not sufficientfor knowingq, it is not sufficientfor knowingp, either, wherep is known to entail q. I shall be returningto this subject later, for some to whom I have referred deny this point

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions SKEPTICISM AND RELEVANT ALTERNATIVES 251 which seems, initially,fairly obviousand I shall arguethat their reasons are mistaken. Actually,if instead of (A) we adopt the strongerepistemic deductive closureprinciple: (D) a knowsp a knows q a knows(p *q entailsr) .,. a knows r (A) and (C) may be seen as instancesof a common principle,provided we allow 'a knows (p q entails p q)' as an uncontroversial instance of the third premise.5(D) is, ultimately,what we need, anywayto capture the idea of knowingthe knownlogical consequencesof what one knows, for (A) covers only the known consequencesof the things one knows taken individually,not the known consequencesof one's whole body of knowledge.And although(D) is strongerthan (A), the argumentsfor (A) workjust as stronglyfor (D), and, so far as I can see, thereare no argu- ments that anyone might seriouslyoffer against(D) which do not also applyto (A). However,for the sake of simplicityand conformityto other discussionsin the literature,I shall continueto discussdeductive closure in the form of (A). In summary,I am not absolutelyconvinced of the validityof the prin- ciple of epistemicdeductive closure, as I am not absolutelyconvinced of the validityof the principleof epistemicconjunction, but in neithercase can I think of an objection,and in both cases, apparentproblems they lead to (skepticism,inconsistency) are eitherapparent only or are better handledby givingup other less obvious principles.

2. SKEPTICISM

In Belief, Truth and Knowledge, D. M. Armstrongargues: It is not a conclusiveobjection to a thermometerthat it is only reliablein a certainsort of environment.In the same way, reliabilityof belief, but only within a certainsort of environment,would seem to be sufficientfor the believerto earnthe accoladeof knowl- edge if that sort of environmentis part of his boundary-conditions.6 For example,I know that the stripedanimal I see in the zoo is a zebra.7 I know this despitethe fact that I have no particularevidence that it is

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions 252 G. C. STINE not a mule paintedto look like a zebra (I have not looked for a paint can, tried paint removeron the animal, etc.). In this context - under normalcircumstances, in zoos of integrity,etc. - that an animalon display has been deliberatelydisguised to fool trustingzoo-goers is just not a relevanthypothesis, one that I need trouble myself about rejecting.If the skeptictries to persuademe to his position by stressingmy lack of evidenceagainst such an hypothesis,my properresponse is to turna deaf ear. He has ensnaredme by impropermeans and is more than halfway to (illegitimately)winning his point if he gets me to agree that I must arguewith him, go look for furtherevidence, etc. This view, which I call the relevantalternative view, seemsto me fun- damentallycorrect. It does leave a lot of thingsunsaid. What are normal circumstances?What makes an alternativerelevant in one context and not in another?However, in ordinarylife, we do exhibit ratherstrong agreementabout what is relevantand what is not. But there are grey areas.Alvin Goldmanmakes this point nicelywith the followingexample which he attributesto Carl Ginet: if on the basis of visual appearances obtainedunder optimumconditions while drivingthrough the country- side Henry identifiesan object as a barn, normallywe say that Henry knowsthat it is a barn.Let us suppose,however, that unknownto Henry, the region is full of expertlymade papier-machefacsimiles of barns.In this case, we would not say Henryknows that the objectis a barn,unless he has evidenceagainst it being a papier-machefacsimile, which is now a relevantalternative. So much is clear, but what if no such facsimiles exist in Henry'ssurroundings, although they do in Sweden?What if they do not now exist in Sweden,but they once did? Are either of these cir- cumstancessufficient to makethe hypothesisrelevant? Probably not, but the situationis not so clear. Another area of obscurityresides not in the natureof the case but in the formulationof the view in question.Goldman seems to hold what I regardas the correctversion of it, which is that: (1) an alternativeis relevantonly if thereis some reasonto think that it is true. But there is also the view that: (2) an alternativeis relevantonly if thereis some reasonto think it couldbe true.

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions SKEPTICISM AND RELEVANT ALTERNATIVES 253

Clearly,the force of the 'could'cannot be merelogical possibility,or the relevantalternative view would lose its distinguishingfeature. However, if the 'could'is readin some strongerway, we could still have a versionof the relevant alternativeview. Dretske's 'ConclusiveReasons' 8 paper, espousing a view accordingto which if one knows, then given one's evidence,one couldnot be wrong(he reads'could' as 'physicallypossible') suggeststhat we should consideran hypothesisa live one unless it could not be true,given one's evidence. Hence any alternativewould be relevant, in the sense of blockingknowledge, if one has not the evidenceto rule it out, so long as it is physicallypossible, given one's evidence.Also, the passagein 'EpistemicOperators' where Dretske says: "A relevantalter- native is an alternativethat might have been realized in the existing circumstancesif the actualstate of affairshad not materialized",9is more akin to (2) than (1), althoughso taking it dependson the force of his 'might'.This, I think, is the wrong way to take the relevantalternative view. First of all, howeverunclear it may be as to when there is some reasonto think an alternativeis true, it is muchmore unclearas to when thereis reasonto think it could be true. Certainly,if thereis a difference between(1) and (2), (2) is weaker,allows more to count as a relativealter- native. So possibly Descartesthought there was some to think that there could, in some sense strongerthan logical possibility,be an evil genius.But it seemssafe to say he was wrongif he thoughtthat there was some reasonto think that there was an evil genius.That is, the evil genius hypothesisis not a relevantalternative according to (1) but may be accordingto (2) (althoughI shall qualifythis). But the whole thrustof the relevantalternative position, as I conceiveit, is that such an hypoth- esis is not relevant.To allow it as relevantseems to me to precludethe kind of answerto the skepticwhich I sketchedin the openingparagraph of this section. In truth, Dretske does combine a relevantalternative view with an answerto skepticism.But his accountis tied in with a view of knowledge, which, althoughit does defeat skepticism,does so in a way which gives small confort. On his account,we do know many things, i.e., there are many things about which given our evidence,we could not be wrong. However,he does not merelyreject the viewthat knowingentails knowing that one knows.'0He also seems committedto the view that one rarely, if ever,knows that one knows,for it is well highimpossible on his account

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions 254 G. C. STINE to defendthe claim that one knows,given one's evidence,that one could not be wrong, in his sense of 'could'. Perhapsthis is preferableto skep- ticism, but at best it is going from the fire into the fryingpan. Here some qualificationsof this position that the relevantalternative view providesan answerto the skeptic are in order. In truth, in some sense skepticismis unanswerable.This rathersupports the relevantalter- nativeview, for the uncertaintywhich infects (1) as to whenthere is some reasonto think an alternativetrue explainswhy this is so. The relevant alternativeview does providea kind of answerto the skeptic- the only kind of answerwhich can be given.But the skeptichas an enteringwedge, and rightlyso. It is an essentialcharacteristic of our conceptof knowledge that tightercriteria are appropriatein differentcontexts.11 It is one thing in a streetencounter, another in a classroom,another in a law court- and who is to say it cannot be anotherin a philosophicaldiscussion? And this is directlymirrored by the fact we havedifferent standards for judging that thereis some reasonto think an alternativeis true,i.e., relevant.We can point out that some philosophersare very perversein theirstandards (by some extremestandard, there is some reasonto think thereis an evil genius,after all) - but we cannotlegitimately go so far as to say that their perversityhas stretchedthe conceptof knowledgeout of all recognition- in fact they have played on an essentialfeature of the . On the other hand, a skepticalphilosopher is wrong if he holds that othersare wrongin any way - i.e., are sloppy, speakingonly loosely,or whatever- when they say we know a great deal. And the relevantalternative view gives the correctaccount of why a skepticis wrongif he makessuch accu- sations.

3. DEDUCTIVE CLOSURE AND SKEPTICISM

Proponentsof the relevantalternative view have tended to think that it providesgrounds for rejectingdeductive closure. Although many philos- ophershave recentlytaken this position,Dretske has providedthe fullest publishedargument to this effect. He writes: To know that X is A is to know that X is A within a frameworkof relevantalternatives, B, C, and D. This set of contraststogether with the fact X is A, serve to definewhat it is that is known when one knows that X is A. One cannot change this set of contrasts without changing what a person is said to know when he is said to know that X is A. We have subtle ways of shifting these contrastsand, hence, changingwhat a person is said to know withoutchanging the sentencethat we use to expresswhat he knows.'2

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions SKEPTICISM AND RELEVANT ALTERNATIVES 255

Considerthe followinginstance of (A): (E) John knows that the animalis a zebra John knows that [the animalis a zebraentails the animalis not a mulepainted to look like a zebra] . . John knows that the animalis not a mule paintedto look like a zebra In Dretske'szoo example,the animal'sbeing a mule paintedto look like a zebra is not a relevantalternative. So what one meanswhen one says that John knows the animalis a zebra,is that he knows it is a zebra,as opposedto a gazelle,an antelope,or other animalsone would normally expectto find in a zoo. If, however,being a mule paintedto look like a zebrabecame a relevantalternative, then one wouldliterally mean some- thing differentin sayingthat John knowsthat the animalis a zebrafrom whatone meantoriginally and that something else maywell be false.Now, normally,in saying that one knows that p, one presupposes(in some sense) that not-p is a relevantalternative; hence one does not know p unlessone has evidenceto ruleout not-p.This is in fact Dretske'sview, for he holds that one does not know that the animalis not a mule paintedto look like a zebrabecause one has no evidenceto rule out the possibility that it is. However,according to Dretske,so long as the animal'sbeing a mule paintedto look like a zebrais not a relevantalternative, the fact that John does not know that it is not does not count against John's knowingthat it is a zebra.Hence, deductive closure fails (we areassuming that John's knowingan animal'sbeing a zebra entails his knowingthat it is not a mule); i.e., (E) and hence (A), are invalid. I submitthat there is anotheraccount of this exampleon the relevant alternativeview which does not entail giving up deductiveclosure. On this account, to say that John knows that p does normallypresuppose that not-p is a relevantalternative. This is, however,a pragmatic,not a semanticpresupposition.13 That is, it is the speaker,not the sentence(or )itself, who does the presupposing.Thus, the falls in the category of those which Grice labels 'cancellable'.14It is possible for 'John knows that p' to be true even though a pragmatic presupposition,that not-p is a relevantalternative, is false. I would say that we may createsome sort of specialcircumstance which cancelsthe normalpresupposition when we utterthe sentencein the courseof making

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions 256 G. C. STINE a deductiveclosure argument. After all, the utterancehas got to be an odd case wherewe are given that not-p is not a relevantalternative to begin with - we can expect somethingunusual to happen, other than being forced to admit that it is a relevantalternative, after all. For even if we would not normallyaffirm 'John knows that p' in such a situation,we would not normallysay that John does not know that p, either. Or it may happen that stating a deductiveclosure argumentaffects normal presuppositionsin anotherway. If we hesitateto say "John knows that the animalis not a mulepainted to look like a zebra",we maywell hesitate to affirm"John knows the animalis a zebra".If this is so, not being a mulepainted to look like a zebrawill havebecome a relevantalternative - we will have decidedthere is some reasonto think it true - with respect to the latter sentenceas well. Perhapsthe mere utteranceof the former sentenceis enoughto make us loosen up our notion of what counts as a relevantalternative. Eitherway, my accountholds the set of relevantalternatives constant from beginningto end of the deductiveclosure argument.This is as it should be; to do otherwisewould be to commitsome logical sin akin to equivocation.If the relevantalternatives, which have after all to do with the truthor falsityof the premisesand conclusion,cannot be held fixed,it is hard so see on what basis one can decidewhether the argumentform is valid or not. And if the set of relevantalternatives is one thing for the firstpremise and anotherfor the conclusion,how do we determinewhat it is for the secondpremise, and how doesthis affectthe truthof the second premise?There is no reasonfor my accountof the matterto make skep- tics of us all. The skepticalargument goes: If you know it is a zebra,and you know its being a zebraentails its not being a paintedmule, then you know it is not a mule paintedto look like a zebra.But you do not know the last, so you do not know the first- i.e., you do not know it is a zebra. With our account in hand, let us see how the skeptic is to be treated. There are two possibilities.First, the skeptic may be up to something legitimate.He is beginningby suggestingthat beinga mulepainted to look like a zebra is a relevantalternative - i.e., that there is some reason to think it is true. We point out to the skeptic that undernormal circum- stances,given what we know of peopleand zoos, etc., this is not the case. The skepticmay, however,persevere, playing on the looseness of 'some reason to think true'. At this point, while we cannot argue the skeptic

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions SKEPTICISM AND RELEVANT ALTERNATIVES 257 out of his position,we are perfectlywithin our rightsin refusingto adopt the skeptic's standardsand can comfort ourselvesby feeling that the skeptic,if not flatly wrong, is at least very peculiar.On the other hand, the skepticmay be up to somethingillegitimate. He may be tryingto get us to doubt that we know it is a zebrawithout going throughthe hard work of convincingus that being a mule paintedto look like a zebra is a relevantalternative. The skepticseeks to persuadeus of his conclusion by gettingus to admit that we do not know it is not a mule paintedto look like a zebra becausewe do not have evidenceto rule out the pos- sibilitythat it is. This is what Dretskebelieves and this is why he we must give up deductiveclosure to defeat the skeptic. I think this a wrongmove. We do know it is not a mule paintedto look like a zebra. Let us grant temporarilyfor the sake of this argumentwe do not have evidence.But Dretskeis deludedby the fact that manyknowledge claims requireevidence on the part of the knowerinto thinkingthat all knowl- edge claims requireevidence. Normally, as I have admitted,saying 'a knows that p' presupposesthat not-p is a relevantalternative. And it does sound odd to say that we know it is not a mule paintedto look like a zebrawhen its being one is not a relevantalternative. But the fact that it sounds odd - is indeedperhaps misleading or even improperto say - does not meanas we have seen that the presuppositionis not cancellable, and that the propositionin question is not true. We often get results which sound odd to say when we drawvalid conclusionsfrom true pre- mises the utteranceof which does not sound odd. 'John knows that it is raining may be true and quite in order to say to convey its literal meaning.But on the assumptionof minimallogical competence on John's part and deductiveclosure, it entails'John knowsthat it is eitherraining or not raining'.But this sentence,if utteredat all, is most likelyto be used to suggestthe negationof the first sentence.We might, in fact, say that the speakerpresupposes it. Given knowledgeof the fiist sentence,the latteris too obviouslytrue to botheruttering at all, exceptfor purposes of sarcasm,ironic effect, or some purposeother than conveyingthe in- formation expressedby the literal meaning of the words. Yet, for all that, it is literallytrue. Or take a case with perhapsmore analogiesto our example.This is an examplefrom Grice.15'My wife is in the kitchen' implies'My wife is in the kitchenor in the bedroom'.Yet, the utterance of the latter, in normal circumstances,presupposes the speaker'signo-

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions 258 G. C. STINE rance of the formerand is thus an improperor at best misleadingthing for him to say if he knows the former.But for all that, the latteris true if the formeris, and the presuppositionis cancellable. The logical consequencesof knowledgeclaims which the skepticdraws by deductiveclosure of the sort Dretskediscusses, are the sorts of prop- ositions which, in normal circumstances,are such that their negations are not relevantalternatives. Thus they sound odd to say and often have the effect of suggestingthat the circumstancesare abnormal.It is indeed improperto utter them in normal circumstancesunless one explicitly cancelsthe relevantalternative presupposition which they carry,because one misleads.Nevertheless, they areliterally true in normalcircumstances. I endorsehere a view which I believeto be Austin's.16This view is ad- umbratedin the followingpassage:

If, for instance,someone remarksin casual conversation,'As a matter of fact I live in Oxford', the other party to the conversationmay, if he finds it worth doing, verify this assertion; but the speaker,of course, has no need to do this - he knows it to be true (or, if he is lying, false).... Nor need it be true that he is in this position by virtue of having verifiedhis assertion at some previous stage; for of how many people really, who know quite well wherethey live, could it be said that they have at any time verified that they live there? When could they be supposed to have done this? In what way? And why? What we have here, in fact, is an erroneous doctrine... about evidence.'7

The point is that one does know what one takes for grantedin normal circumstances.I do know that it is not a mule painted to look like a zebra.I do not needevidence for sucha proposition.The evidencepicture of knowledgehas been carriedtoo far. I would say that I do not have evidencethat it is a zebra, either.I simplysee that it is one. But that is perhapsanother matter. The point I want to make here is simplythat if the negationof a propositionis not a relevantalternative, then I knowit - obviously,without needingto provideevidence - and so obviouslythat it is odd, misleadingeven, to give utteranceto my knowledge.And it is a virtueof the relevantalternative view that it helpsexplain why it is odd. Thereis anotherway in which (E) could be defended.This line could be to claimthat John does, afterall, in his generalknowledge of the ways of zoos and people, etc., have evidencethat the animal is not a mule paintedto look like a zebra.The samewould hold for otherconsequences of knowledgeclaims which the skepticdraws by deductiveclosure. This would involve a notion of evidenceaccording to which having evidence

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions SKEPTICISM AND RELEVANT ALTERNATIVES 259 is not just limitedto cases in which one has a specificdatum to whichto point. Malcolmexpresses this point of view when he says:

... The reason is obvious for saying that my copy of James's book does not have the characteristicthat its print undergoes spontaneous changes. I have read millions of printed words on many thousands of printed pages. I have not encountereda single instance of a printedword vanishingfrom a page or being replacedby anotherprinted word, suddenlyand without externalcause. Nor have I heard of any other person who had such an encounter.There is overwhelmingevidence that printedwords do not behave in that way. It is just as conclusiveas the evidencethat houses do not turn into flowers. That is to say, absolutelyconclusive evidence'8 (underscore mine).

It is true that in the last sentenceof this passageMalcolm talks about evidencefor a universalproposition to the effect that printedwords do not behavein a certainway, but the thrust of his argumentis such that he commitshimself to the view that he also (thereby)has evidencethat the printedwords on his particularcopy of James'sbook will not behave that way. I am not inclinedtowards such a view of what it is to have adequateevidence for the propositionthat the printof my own particular copy of James'sbook did not undergoa spontaneouschange. I am in- clinedto rejectMalcolm's view, and othersakin, in favorof the Austinian sort of one previouslydiscussed - that is, that in such a case, evidenceis not requiredto supporta knowledgeclaim. I mentionthe view only as a possible alternativeview of defendingepistemic deductive closure in a way consonantwith the relevantalternative riew.

4. SUMMARY

My view is that the relevantalternative position should be conceivedof as in two parts: (1) With respectto manypropositions, to establisha knowledgeclaim is to be ableto supportit as opposedto a limitednumber of alternatives- i.e., only those which are relevantin the context. (2) With respectto many - in particularthose which are such that their negationsare not relevantalternatives in the context in question- we simplyknow them to be true and do not need evidence,in the normalsense, that they, ratherthan their negations,are true. So conceived,the relevantalternative view neithersupports the aban- domnent of deductiveclosure, nor is such abandonmentin any way

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions 260 G. C. STINE needed to provide the relevantalternative view with an answerto the skeptic,insofar as he can be answered.19

Wayne State University

NOTES

1 am partial to J. L. Austin's approach in 'Other ' (PhilosophicalPapers, Oxford, 1961). and ChapterX of Sense and Sensibilia (Oxford, 1962). Other more recent and more explicitly developed accounts include those of Fred Dretske, most importantlyin 'EpistemicOperators', Journal ofPhilosophy, LXVH (1970), 1007-1023, but also in 'ContrastiveStatements', Philosophical Review LXXXI (1972), 411-430; D. M. Armstrong,Belief, Truthand Knowledge(Cambridge, 1973); , 'Discriminationand PerceptualKnowledge', presented at the Annual PhilosophyCol- loquium, Universityof Cincinnati,1973; James Cargile, 'Knowledge and Deracination', presented at the Annual Philosophy Colloquium, University of Cincinnati, 1973; Norman Malcolmin 'The VerificationArgument' in Knowledgeand Certainty(Prentice- Hall, 1963) is more concerned with certainty than knowledge but his discussion of when a proposition is 'possible'is very much in accord with considerationswhich go towards making a proposition a 'relevantalternative'. 2 Dretske, 'EpistemicOperators', Op. cit.; Cargile,Loc. cit; Goldman,Loc. cit. 3 Cf. 'QuantifiedLogic for KnowledgeStatements', Journal ofPhilosophy LXXI (1974), and 'Essentialism,Possible Worlds,and PropositionalAttitudes', Philosophical Review LXXXII (1973), 471-482. 4 Cf. discussionsof the place of a principleof conjunctionin an account of rational belief in, for example,Isaac Levi, GamblingWith Truth, Knopf (1967);and in Induction, Acceptance,and Rational Belief, ed. by Swain, Reidel (1970) the following papers: MarshallSwain, 'The Consistencyof Rational Belief', Henry Kyburg,'Conjunctivitis', and Keith Lehrer,'Justification, Explanation, and Induction'.This case for conjunction holding for rationalbelief is, of course, more problematicthan the case for knowledge. 5 I owe this point to David Kaplan. 6 Armstrong,Op. cit., p. 174. 7 The example is Dretske's in 'EpistemicOperators', Loc. cit. 8 In AustralasianJournal of Philosophy49 (1971), 1-22. 9 'EpistemicOperators', loc. cit., p. 1021. 10 This view has been criticized, for example, by Ronald DeSousa in 'Knowledge, Consistent Belief, and Self-',Journal of Philosophy(1970), against de- fenders of it such as Jaakko Hintikka in Knowledgeand Belief (Cornell, 1962) and Keith Lehrerin 'Belief and Knowledge',Philosophical Review (1968). The view is also rejectedby Armstrong,Op. cit. (p. 146),and at least implicitlyrejected on such accounts of knowledgeas, for example,those of Alvin Goldman,'A CausalTheory of Knowing', Journalof Philosophy64 (1967), 357-372; Brian Skyrms,'The Explicationof 'X Knows that P"', Journalof Philosophy64 (1967), 373-389; and Peter Unger, 'An Analysis of Factual Knowledge',Journal of Philosophy65 (1968), 157-170. 11 Here I take a view directlyopposed to that of PeterUnger, 'A Defense of Skepticism', PhilosophicalReview (1971), according to which knowledge is an 'absolute' concept, like the flatness of geometers. 12 'EpistemicOperators', Loc. cit., p. 1022.

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions SKEPTICISM AND RELEVANT ALTERNATIVES 261

13 Here I distinguish pragmatic from semantic presuppositionsin the manner of Robert Stalnaker,'Pragmatics', in Semanticsof Natural Language,ed. by Davidson and Harnan (Reidel, 1972), 380-397. Attributingthe notion of a semanticpresuppo- sition to ('Singular Terms, Truth Value Gaps, and Free Logic', Journalof Philosophy63 (1966), 481-495, and 'Presupposition,Implication, and Self Reference',Journal of Philosophy65 (1968), 136-151, Stalnakersays (p. 387): Accordingto the semanticconcept, a propositionP presupposesa prop- osition Q if and only if Q is necessitatedboth by P and by not-P. That is, in every model in which P is either true or false, Q is true. According to the pragmatic conception, presupposition is a , not a semantic relation. People, rather than sentences or propositions are said to have, or make, presuppositionsin this sense. ... In general, any semantic presuppositionof a proposition expressed in a given context will be a pragmaticpresupposition of the people in that context, but the converseclearly does not hold. To presupposea proposition in the pragmaticsense is to take its truth for granted, and to assume that others involved in the context do the same. 14 H. P. Grice, 'The Causal Theory of ',Proceedings of the Aristotelian Society, Suppl. Vol. XXXV (1961). 15 Ibid. 16 In 'Other Minds', Loc. cit., and ChapterX of Sense and Sensibilia,Oxford (1962). 17 Sense and Sensibilia,pp. 117-118. 18 'The VerificationArgument', Loc. cit., p. 38. 19 A slightlydifferent and shorterversion of this paperwas read at the EasternDivision meetings of the American PhilosophicalAssociation, December 1974.

This content downloaded from 128.122.149.154 on Wed, 11 Sep 2013 01:11:58 AM All use subject to JSTOR Terms and Conditions