Essential Attribution Author(S): Ruth Barcan Marcus Source: the Journal of Philosophy, Vol

Home , Ruth Barcan Marcus, Terence Parsons

Journal of Philosophy, Inc.

Essential Attribution Author(s): Ruth Barcan Marcus Source: The Journal of Philosophy, Vol. 68, No. 7 (Apr. 8, 1971), pp. 187-202 Published by: Journal of Philosophy, Inc. Stable URL: http://www.jstor.org/stable/2024901 . Accessed: 04/12/2013 15:38

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 [email protected].

Journal of Philosophy, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Philosophy.

http://www.jstor.org

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions THE JOURNALOF PHILOSOPHY

VOLUME LXVIII, NO. 7, APRIL 8, I97I

ESSENTIAL ATTRIBUTION * I A SORTING of attributes(or properties)as essentialor inessentialto an object or objects is not whollya fabrication of metaphysicians.The distinctionis frequentlyused by philosophers and nonphilosophersalike without untoward per- plexity.Given theirvocation, philosophers have also elaboratedsuch use in prolixways. But to proclaim that any such classificationof propertiesis "senseless," "indefensible,"and leads into the "metaphysical jungle of Aristotelianessentialism"' is impetuous. It supposes that cases of use that appear coherentcan be shown not to be so or, alternatively,that thereis an analysis that dispels the distinc- tion and does not relyon equally odious notions.It furthersupposes that takingthe distinctionseriously inevitably leads to what Quine calls "Aristotelianessentialism." The latter claim is a nest of pre- sumptions,two of which are that "Aristotelianessentialism," as characterizedby Quine, is a characterizationof Aristotelianessen- tialism,and that any theoriesthat countenance the distinctionare versionsof "Aristotelianessentialism." On the occasion where Quine seems to propose an argument againsta genuinemode ofessentialism, i.e., thecase ofthe mathemat- cal cyclist,2it is seen on alternativeinterpretations that either the argumentis invalid or thereis no groundfor supposing that anyone would accept its premises.3Friendly critics suggest that it was not

* Presented in an APA symposium on Essentialism, December 29, 1970; other symposiastswere David Kaplan and Saul Kripke. My thanks to Terence Parsons for his comments on an earlier version. I W. V. Quine, Word and Object (Cambridge, Mass.: MIT Press, 1960), pp. 199- 200; "Three Grades of Modal Involvement," in The Ways ofParadox (New York: Random House, 1966), p. 174. Wordand Object,p. 199. 3First shown in my "Modalities and Intensional Languages," Synthese,xIII, 4 (December 1961): 303-322, pp. 317-319; reprinted in Marx Wartofsky,ed., I87

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions I88 THE JOURNAL OF PHILOSOPHY intendedas an argument.It was supposed to bewilderus, and it does. But Quine acknowledgesthat it was never his purpose to address himselfseriously to essentialistclaims. Raising spectersof essentialism was ancillaryto the granderpurpose of rallyingfurther reasons for rejectingquantified modal logic. The latter, he says, is "committedto essentialism,"which in turnis, on the face of it, senseless. 14 and Terence Parsons5 have argued-persuasively I believe- that Quine's casual characterizationof essentialismis inadequate. It missesthe point of what seems to be presupposedin coherentcases of use or in Aristotelianessentialism. Furthermore,since the char- acterizationdoes not take us beyond the distinctionbetween necessary and contingentpropositions (which Quine accepts), thereis no cause forperplexity. What is perplexingis that his characterization doesn't fithis case of the mathematicalcyclist. On a more adequate characterization,which in the presentpaper is taken to be minimal,Parsons6 showed that quantifiedmodal logic (QML) is not committedto essentialism(E) in the followingsense: in the rangeof modal systemsfor which Saul Kripke7has provideda semantics,no essentialistsentence is a theorem.Furthermore, there are models forwhich such sentencesare demonstrablyfalse. Modal logic accommodatesessentialist talk. But such talk is com- monplace in and out of philosophy.8It is surely dubious whether essentialist talk can be replaced by nonessentialist,less "prob- lematic" discourse.The offhandremark that "some attributescount as importantand unimportant,... some as enduringand othersas fleeting;but none as necessaryor contingent"9(which Quine takes as synonymouswith 'essential or accidental') suggests that such

Boston Studies in the Philosophy of Science (Dordrecht, Holland: Reidel, 1963). Also in my "Essentialism in Modal Logic," Noaas,i, 1 (March 1967): 91-96. See also Parsons, Plantinga, and Cartwright,fns 5 and 13 below. I In Marcus, Quine, S. Kripke, J.McCarty, and D. Follesdal, "Discussion on the Paper of R. B. Marcus," Synthese,xiv, 2/3 (September 1962): 132-143; reprinted in Boston Studies, op. cit. Also, "Essentialism in Modal Logic." 5 "Grades of Essentialism in QuantifiedModal Logic," Noas, 1, 2 (May 1967): 181-191. Also, "Essentialism and QuantifiedModal Logic," Philosophical Review, LXXVIII, 1 (January 1969): 35-52. For Quine's attempt at formalcharacterization, see fn 22. 6 "Essentialism in Quantified Modal Logic." 7 "Semantic Considerations on Modal Logic," Acta Philosophica Fennica, fasc. 16 (1963): 83-94. 8 If the reader is in doubt, he is urged to check the claim against his own reading. Quine himselfdoes not shun such use. It ranges fromthe metaphorical "Nominal- ism is in essence perhaps a protestagainst the transcendentuniverse" to a precise sortingof a certain propertyof expressions,the propertyof occurringin a sentence, as between essential and inessentialoccurrences. See The Ways ofParadox, op. cit. p. 69, p. 73, p. 103. 9 Wordand Object,p. 199.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions ESSENTIAL ATTRIBUTION I89

talk is dispensablethrough some uniformsubstitution of words that are clearerand untaintedby metaphysics.But if Socrates was born and died snubnosed, then that propertyand his being a man are equally durable. Are we helped by the furtherassertion that his being a man is moreimportant? In what way? Is it moreimportant than his being a philosopher?And what are we to make of cases whereit is claimed of a certainattribute that it is importantbut in- essential?10Is 'important'less problematicthan 'essential'? Given the apparentcoherence of some essentialisttalk, interpreted systems (g) of QML are appropriatevehicles for analysis. The pres- ent paper continueswith the account of modes of E withinthe frame- workof QML. In particular,it is suggestedthat Aristotelianessen- tialism may best be understoodon a "natural," or "causal," interpretationof the modal operators.But firsta few historicalremarks. The taxonomyof pre-formallogic distinguishedpure and modal propositions.According to W. S. Jevons, The pureproposition simply asserts that the predicatedoes or does notbelong to thesubject, while the modal proposition states this cum modo,or withan intimationof the mode or mannerin whichthe predicatebelongs to thesubject. The presenceof any adverb of time, place, manner,degree, etc., or any expressionequivalent to an adverb,confers modality on a proposition."Error is alwaysin haste," "justiceis everequal," ... are examplesof modalpropositions.1' Furtheron he mentionsthat some logicians have adopted a special view with respect to 'necessarily','possibly', and the like,as in "an equilateral triangleis necessarilyequiangular," "men are generally trustworthy,"where the "modalitydoes not affectthe copula of the proposition"but "consists in the degree of certaintywith which a judgmentis made or asserted" (70). With formalizationcame the faiththat standard functionallogic (SFL), appropriatelyinterpreted, would yield an analysis and dis- ambiguationof modal propositions.Logical grammarwould replace surfacegrammer, yet the sense (if it had one at all) of the original would be capturedby its formal,nonmodal-counterpart. The faith was not ungrounded.Jevons used 'always' and 'ever' as examples. 10The firstchapter of M. B. Hesse, Modelsand Analogiesin Science (Notre Dame, Ind.: UniversityPress, 1966) beginswith a discussionbetween a Camp- bellianand a Duhemistabout whether it is essentialto (oran essentialproperty of) a scientifictheory that it havea model.The Campbellianclaims it is essential.The Duhemistclaims it maybe importantor usefulbut not essential. Nor is thediscus- sion "withoutsemblance of sense." We recognizein it thesame conceptual scheme implicitin suchdistinctions with respect to propertiesof moremundane objects. 11Lessons in Logic (Londonand New York: Macmillan,1884), p. 69.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions I90 THE JOURNAL OF PHILOSOPHY

Interpretations(,4) of SFC take us a long way. The adverb is de- tached from the predicate. The nontemporal'always' and 'some- times' go into quantification.The temporalcases go into 4 of SFC which includes temporalmoments in the domain of 4. Similarlyfor the nonspatial (rare) and spatial 'everywhere'and 'somewhere'.But even in the temporalcases the success is incomplete.The modalities proved recalcitrant,and extensionsof SFL proved useful.Those who frown on such extensions as deviant should remember Rudolf Carnap's admonition: "In logic thereare no morals."'2 II I should like to focuson two modes of essentialismwhich I will dis- tinguishas individuatingand Aristotelian.Consider some cases. We say of Moby Dick that althoughhe lives in the sea he is essentially a mammal, and of Socrates that he is essentiallya man and acci- dentallysnubnosed. We point to a sample of mercuryat room tem- peratureand say that,although it is a liquid, it is essentiallya metal, suggesting that solidity at room temperature is an accidental propertyof metals. These are cases that fit Aristotle'saccount of essences. What is implicithere? The objects are actual objects, and the propertiesthat are being sortedas essential or inessentialcorre- spond to direct,nonvacuous, "natural" predicates. (A more formal and inevitably approximate characterizationof such predicates is deferredfor subsequent discussion.) For Aristotelianessentialism, an essentialproperty is a propertythat an object musthave. It answers to the question "What is it?" in a strongsense; if it ceased to have that propertyit would cease to exist. It is a propertysuch that, if anythinghas it at all, it has it necessarily.The latter conditionis what distinguishesAristotelian from what I call "individuating" essentialism. Consider,for example, Winston'3the mathematicalcyclist. Sup- pose he is an avid cyclingenthusiast. It is an overridingpreoccupa- tion. Although he holds a position on a mathematicsfaculty, his interestin that subject is at best desultory.Arguing against a re- newal of Winston'scontract, a colleague says "Unlike the restof us, Winstonis essentiallya cyclist,not a mathematician."Analogously, Protagorasmight have said of Socrates, "He's essentiallya philosopher, not a politician." The social workersays of a client, "He's essentiallya good boy; just fellin withbad company,"which is after all not too distantfrom distinguishing, as philosopherssometimes do,

2The Logical Syntax of Language (New York: Springer,1937), p. 52. 18So named by R. Cartwrightin "Some Remarkson Essentialism,"this JOURNAL, LXV, 20 (October24, 1968): 615-626,p. 619. Also called "Squiers" by Alvin Plantinga in "De Re and De Dicto," Nog2s,iII, 3 (September 1969): 235-258.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions ESSENTIAL ATTRIBUTION I9I those who are disposed to act rightlyfrom those who merelyact rightlyout of expediencyor the like. Nor is the context of such examples primarilycolloquial. Consider the following:"In the last century,Dedekind, Frege, and Peano asked themselves: what is essentialabout the seriesof natural numbers (forpure arithmetic)? If one thinks of this structureas an object at all, the following propertiesare clear: . . . " or "what propertiesof an object in mathematical experienceare essential?This is well illustratedby Souslin's problem on the continuum.... How essentialare the rationals to the intuitivecontinuum ?"'4 Implicit in such examples is that among the attributesan object must have are not only those whichit shareswith objects of its kind (Aristotelianessentialism), but those which are partiallydefinitive of the special characterof the individualand distinguishit fromsome objects of thesame kind.But mustthere be some set of individuating essentialattributes that wholly distinguish an object fromthose of its kind? (For presentpurposes, we need not go into the questionof the uniquenessof proximatekinds, the hierarchyof kinds,and the like.) Being a snubnosed, henpecked, hemlock-drinkingphilosopher'5 wholly individuatesSocrates without the addition of a uniqueness condition,but although being a philosophermay be essentialto his nature, presumably being snubnosed is not. Perhaps complete individuationis always a matterof what are generallytaken to be inessentialproperties, accidents of circumstance.If we encountered a winged horse,we could not determinethat it was Pegasus unless we knew the circumstancesof his birthand the like. But then,those circumstanceswould be inessentialto Pegasus as well. Numbers,as contrastedwith empirical objects, are supposed to be objects that can be whollyindividuated by theiressential properties, withouttacking on a uniquenesscondition. But the matteris by no means clear. Consider,for example, the counterclaimthat numbers have no essentialproperties at all, forif theydid "it would conflict with the idea that numbertheory can be reduced to set theoryin various ways."'6 One possibleresolution to thisdisagreement is that,

14G. Kreisel,"Mathematical Logic: What Has It Done forthe Philosophyof Mathematics?"in Ralph Schoenman,ed., BertrandRussell: Philosopherof the Century(London: Allen & Unwin,1967), pp. 213-216. 15 The exampleis fromDaniel Bennett,"Essential Properties," this JOURNAL, LXVI, 15 (Aug. 7, 1969): 487-499,p. 487. 16Gilbert H. Harman, "A NonessentialProperty," this JOURNAL, LXVII, 6 (March26, 1970): 183-185.Harman's claim raisesinteresting questions. If one supposesthat numbers are first-orderobjects, that there is at mostone of each, and thatthere are neverthelessequally acceptable alternative choices for the natural- numberstructure, then any "world" W thatincludes natural numbers in its do-

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions 192 THE JOURNAL OF PHILOSOPHY if somethingcounts as a number,it has essential numericalproper- ties, but they do not whollyindividuate. Perhaps with respect to inquiries like "Who is Sylvia, what is she?" the latter question can be answered in terms of essences (Aristotelian),but individuatingessences can never whollyanswer the former.This leads some philosophersto make a metaphysical shift.They invent objects (individual concepts, forms,substances) called "essences," which have only essential properties,and then worrywhen they can't locate those objects by rummagingaround in otherpossible worlds. It does not seem to me that an account of essential attribution compels us, even with respect to abstract objects, to shiftour ontologyto individualessences. The usefulness of talk about possibleworlds is not forpurposes of individuatingthe object-that can be done in thisworld; such talk is a way of sorting its properties. mainD does so by specifyingwhich object it identifieswith 0, whichwith 1,. Althoughthere may be othersubsets of D withstructures (N1,Si,di) isomorphic to thegiven choice for (N,s,d), where N is theset of number elements, S thesuccessor relation,and d the distinguishedelement, elements of N1 that are isomorphicto elementsof N willnot be thenumbers in thatworld W. The truthof Harman's claim, i.e., that numbers have no essentialproperties at all, revolvesabout what is meantby "reducingnumber theory to set theory."If it meansthat there is no suchthing as thenatural-number structure but onlysome set ofalternative isomorphic structures (N',S',d'), one foreach W', each of which representssome alternative set-theoretic reduction, then his claim is correct.And, as Parsonshas shownin "Essentialismand QuantifiedModal Logic," pp. 44-46, QML maybe extendedto includearithmetic truths without the consequence that numbershave any essentialproperties at all. '9 is necessarilygreater than 7,' for example,comes to: 'In each possibleworld there is somethingthat is 9 and somethingthat is 7 suchthat the first is greaterthan the second'. Furthermore, 'is the numberof planets'is not substitutablefor 'is 9' in accordancewith principles of substitutionin modalcontexts. But ifone takesit thatthere is somethingthat is thenatural-number structure, as Kreiselsuggests in "MathematicalLogic: WhatHas It Done forthe Philosophy of Mathematics?"then alternative reductions are isomorphicto but notidentical withthe series of natural numbers. In a worldwith a von Neumannspecification of theseries of natural numbers, being a memberof 1 willbe an inessentialproperty of 0, but beingless than 1 willbe an essentialproperty. Here, as in the case of the numberof planets, etc., the theory of descriptions-or taking singular descriptions ('the emptyset') as singularpredicates-will be requiredin the analysis of sentenceslike 'the empty set is necessarilyless than 1'. Harman,in exposingwhat he takesto be thefoibles of essentialists, fails to notethat nonessentialists also rely, in theirantiessentialist arguments, on the presumptionthat thereare objectsof reference,i.e., the series of natural numbers, which are independentof any mode of specification.In thatrespect taking 9 as thereferent of 'the numberof planets' is no differentfrom taking 0 as the referentof 'the emptyset'. Thereis ofcourse a deficiencyin theway Harman presents his thesis, of a serious kind.He seemsto suggestthat one can talk of the propertiesof numbers,independentof theirbeing part of the structure that makes them eligible for number- hood. This is analogousto the speciousarguments that mightdevelop about whethersome small carved piece of woodwas thequeen in a chessgame.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions ESSENTIAL ATTRIBUTION I 93

Withinthe possibleworld view of QML the mattermay be put as follows:among all the direct,nonvacuous, "natural" properties17an object has in this world (W), thereare those it must have. Among those it must have are those it has in commonwith objects of some proximatekind (Aristotelianessentialism) and thosewhich partially individuateit fromobjects of the same kind (individuatingessentialism). We see that to say of an object x and an object y that theyhave all essentialproperties in commonis weaker than claimingidentity. But this reflectscommon speech. To say of x and y that they are essentially the same (the same in essential respects) is a weaker claim than saying they are identical. What has gone wrong in recent discussions of essentialism18is the assumptionof surfacesynonymy between 'is essentially'and de re occurrencesof 'is necessarily'.But intersubstitutionoften fails to preservesense. Would Winston's colleague have been understoodif he had said "Winston is necessarilya cyclist?" And would we ever be inclined to use 'is essentially' instead of 'is necessarily'where vacuous propertiesare concerned,as in 'Socrates is essentiallysnub- nosed or not snubnosed' or 'Socrates is essentially Socrates'? If higher-orderobjects are candidates for essential attribution (as Cartwrightfreely allows) then substitutionwill sometimestake us fromtruth to falsity,as in 'p is essentiallytrue' or 'p is essentially correct'.The connectionbetween the two locutionsis not a surface matter.It is best analyzed withinsome model of QML. Which of the Kripke model structures(W,K,R) are suitable for our analysis? Parson's results clearly exclude maximal models (whereR is symmetricand transitiveas well as reflexive).Intuitive considerationssuggest that, so faras Aristotelianessentialism is con-

17 In what follows,by indicatinghow we "give" an interpretationfor purposes of paraphrase (as distinguishedfrom specifying how we make truthassignments to sentences of our language) and by placing certain restrictionson predicates, we have excluded many predicates that Hume and others would have called nonnatural. See Hilary Putnam, "The Thesis that Mathematics Is Logic," in Schoen- man, op. cit., pp. 299-301, fora briefdiscussion of the "natural" and "philosophical" notion of a predicate or property. Carnap, in The Logical Syntax ofLanguage, pp. 308-309, recommendsexclusion from the class of property-wordsof those which correspond to what he calls "transposed properties." From his examples we see that his "transposed properties" overlaps the loose traditional characterizationof "nonnatural properties." Among his examples are the property a city has if its name is the alphabetical predecessor of the name of a city with more than 10,000 inhabitants; given of course a complete list of names of cities. He also includes properties like being famous, or being discussed in a certain lecture, as not being "qualities in the ordinarysense." 18 See forexample, Plantinga, "De Dicto and De Re"; also "World and Essence," Philosophical Review,LXXIX, 4 (October 1970): 461-492.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions 194 THE JOURNAL OF PHILOSOPHY cerned,the chosensystem should perhapssatisfy the converseof the Barcan formula (the domain of this world W is included in the domain of each K' possiblerelative to W) and R should be transitive (i.e., quantifiedS4). But that is merelya suggestion.Nothing we are claimingin thispaper restson that particularchoice of a system. Wheneverour purposeis to use an interpretedformal language for paraphraseand analysisof an ordinarysentence, how we specifythe interpretationg is crucial,if, in additionto preservingtruth in trans- lation, we want somehow to preserve meaning to the maximum extent. It is perhaps gratuitousto emphasize that this is as true of interpretationsof SFL as of MFL. We would not, in an 4 of SFL (which includes numbers in its domain), choose, fromamong all possible names for9, 'the numberof planets'; forthen the ordinary sentence'the numberof planets is (identically)9' goes into 6 as 'the numberof planets is the numberof planets'. We would not make that choice because we are not inclinedto obliteratemeanings un- necessarily.Nor would we choose 'the henpecked,snubnosed, hemlock-drinkingphilosopher' over 'Socrates' as the name we associate with some constantfor designating that individual. Such considerationsare crucial fortranslatability into QML. For the strategemof talk about possibleworlds is that truthassignments of sentencesand extensionsof predicatesmay vary, but individual names don't alter theirreference, except to the extentthat in some worldsthey may not referat all. If, therefore,we take as Socrates's name a singulardescription that picks him out in this world only, our purpose is defeatedat the outset. What we want is that neutral peg on whichto hang descriptionsacross possible worlds.Similarly, the "sense" of sentencesand predicatesis preservedacross possible worlds.For thosewho are quick to argue that ordinarynames cannot always be used in such a purelyreferential way, we can, in givingthe interpretation,expand our lexicon to provide neutral names where necessary. Given some choiceof an appropriatesystem, we specifyas follows: Associated with sentence symbols are ordinarysentences (not de- scriptionsof sentences). Associated with individual symbols are ordinarynames, not singular descriptions.Where ordinarynames are lacking,such namelessobjects are firstgiven "ordinary" names by a suitable convention for avoiding duplication of names. (A lexiconis kept.) Wheremore than one object has the same name,we distinguishthem by a suitable convention.For symmetrywe might add that, when one object has several names, we choose one as its standard name. This would be required for contexts that have

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions ESSENTIAL ATTRIBUTION I 95 greater obliquity than those here considered. We are restricting indirectoccurrences of variables to those which occur within the scope of a modal operatorso interpretedas to permitintersubstituta- bilitysalve veritateof names (not descriptions)of the same object. Associated with predicate symbols are standardized predicates. A standardized predicate is like an ordinarysentence modifiedas follows: Disambiguate names of multiplereference. Replace one or more occurrencesof names by place markers.Quine's "standard English predicate" for example, with the followingextension: the only indirectoccurrences of names replaceable by place markersare those which fall within the scope of a modality translatable into QML. For example, suppose, as Leonard Linsky claims, that "the statements'I did not miss this morning'slecture, but I mighthave' and 'I did not miss this morning'slecture, but there is a possible world in which I did' are full paraphrases of each other."'9 Then, since 'might' goes into 'O', the predicate formed from Linsky's sentence is indirect,i.e., contains an indirectoccurrence of a place marker. There remains the representationof singulardescriptions. If our choice of (W,K,R) containsidentity, then the theoryof descriptions with attentionto scope will work. An alternative,with or without identity,is taking singutlardescriptions (not names) as uniquely satisfiablepredicates.20 A wordof caution here. In specifyinghow we paraphrase,we hope to avoid a few muddles. Plantinga, forexample, has staked several argumentson the claim that being snubnosed in W is a property Socrates has in all possibleworlds that containhim and is, therefore, essential.Are we to suppose that 'Socrates is snubnosed in W' like 'Socrates was born in Athens',is one of those ordinarysentences we associate withsentence symbols of our interpretedQML, that in the domainof our interpretationthere are places, one of whichis Athens and the other the world,which would put W in the domain of W?

"9"Reference, Essentialism, and Modality,"this JOURNAL, LXVI, 20 (October16, 1969): 687-700. We are excludinghere epistemic contexts along withstronger obliquityin theformation of predicates. For example'John' and 'Jill'may both be replacedin 'Johnmight have marriedJill', but only 'John'may be replacedin 'Johnknew Jill left town' or 'Johnwished Jill would marry him'. 20 In my "Modalitiesand IntensionalLanguages," Russell's theory, or alternatively,taking descriptions as unitattributes or unitproperties was proposed.In eithercase extensionallyequivalent expressions (sentences in the theoryof descriptions,predicates in theother) are not intersubstitutablein modal contexts. I have omittedhere any proposalsfor reinterpretation of quantification, along substitutionallines. It is importantto separatethe groundsfor such a view of quantificationfrom its usefulnessin connectionwith substitution in indirectcon- texts.We are,however, presuming a differencebetween names and descriptions.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions I 96 THE JOURNAL OF PHILOSOPHY

All that Plantinga's funnysentence (P) mightcome to is that,in our choice of (W,K,R), its truthassignment is T in W and so, therefore, must be the assignmentto OP. If we should also choose QS4 as our basis, it will not followthat zOP is assignedT.21 Furthermore if we are to conformto coherentcases, we will argue below that indirect predicates are excluded fromthe characterizationof essentialism; forwould anybody,including essentialists, ever say that Socrates is essentiallypossibly snubnosed? Implicitin essentialismis that an object has attributesnecessarily that are not necessary to other objects. To say that Socrates is essentiallya man is to take as true (1) ElF(s)* (3x) - nF(x) fromwhich it followsthat (2) (3x) rlF(x) * (3x) - rF(x) (EM) Indeed, (2) may be taken as minimalessentialism, where F is any monadicpredicate that contains no constants.(2) excludessuch tauto- logical predicatesas F(x)v F(x). There is anotherkind of vacuous predicate,the partial instantiationof a tautologicalpredicate, which the essentialistdoes not count as designatingessential attributes; e.g., F(s) F(x) is necessarilytrue of s but not of anythingelse. In order to exclude such cases, as well as forextending our charac- terizationto relationalattributes, Parsons has generalized (2) to Fn in such a way as to sort out those identitieswhich hold between free variables in Fn. For simplicityof presentationI will restrict the discussion to monadic predicates that are general (i.e., contain no constants). To those who argue that the exclusion of vacuous predicates is arbitraryand posthoc we need onlypoint out that Aristotleexcluded them,for such philosophersalso claim that theyhave in mind some version of Aristotelianessentialism. Bennett (op. cit.) sums up the Aristotelianview as follows: Beingan entityis a necessaryproperty of everything,i.e., a trans- cendentalproperty.... Essential propertiessort the entitiesof whichthey are truein somefashion (487). Beingan entity,like being self-identical and beinga unity,failed to sortSocrates from anything. Everything is an entity,self-identical, a unity. Being identicalwith Socrates,on the other hand, sorted Socratesfrom everything. Nothing but Socratesis identicalwith Socrates.Essential properties are not transcendental,and theyare not ... individuative(494). 21We havenot presumed symmetry ofthe alternativeness relation.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions ESSENTIAL ATTRIBUTION 197

As we noted above, an extensionof QML that includestruths like (2) supposes that there are necessitiesother than logical. But this conformsto cases. Aristotle,in his theoryof essences,was afterall concernedwith some kind of natural necessity.Indeed, if it is true, as some claim, that numbershave essentialproperties (that meet the conditionof minimalessentialism), did not Kant classifysuch truths as synthetic,although a priori? We now see that, although Quine is mistaken in claiming that "any quantifiedmodal logic is bound to show favoritismamong the traits of an object" and "must settle for essentialism,"22it is true of a QML that importseven minimalessentialism, for (2) does show favoritismamong the traits of an object. Surely if I were to say of Quine that he was necessarilyeither snubnosed or not snubnosed,he could not accuse me of playing favoritesamong his traits. But if I were to say that he is essentiallya man or essentiallya philosopher, that is playing favorites;but is it on the face of it, senseless? If he stillfinds it so, he is not compelledto rejectQML altogether.He can restricthimself to maximal models of QML, in which essentialist statementsare demonstrablyfalse. The ontologyof such models is one of bare particulars. III In the presentsection I will discuss extensionsof (2), whichfit some modes of essentialism. (A) Where an object has essential attributesit is implicitthat it has attributesthat are not necessary.22This could be representedas * * o (3) (3x)(E: F(x) G (x) *oG(x)) (3x) - F(x) (E*) One mightwant to weaken E* by substituting'o G (x)' for'G (x)' or furtherstrengthen the second conjunct. But our interestis not in spinningout alternatives.What is worthindicating is that, if E* is presumed,we see why statementslike '9 is essentiallycomposite' may strikeus as odd ifwe believe that all the attributesof a number are necessary. However, against a backgroundof set-theoreticre- ductions,one mightwant to claim (fora worldwith a von Neumann reduction) that being less than its successoris an essentialproperty of numbersand being a memberof its successoris not. But, as Harman (op. cit.) has pointed out, there are difficulties here. On a Russellian account, E* has greaterplausibility. If our

22From a LogicalPoint of View, rev. ed. (Cambridge,Mass.: Harvard,1961; New York: Harper& Row, 1963),p. 155. 23 0n the one occasion whereQuine attemptsa formalcharacterization, he choosesfor "Aristotelian essentialism," the first conjunct of (3); see "ThreeGrades of Modal Involvement,"p. 174.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions 198 THE JOURNAL OF PHILOSOPHY characterizationis extendedto higher-orderobjects, thenit is a non- necessaryproperty of 9 that it is a propertyof the set ofplanets. The numberof planets mightbe 8 in some world that is possible relative to W. (B) Individuatingessentialism. An extensionof EM thatconforms to cases of what I have called "individuatingessentialism" is (4) (3x) F (x) - (3x) (F (x) - caF(x)) (EI) El may be extended to include E* by appropriatestrengthening of the firstconjunct. This yields (E*I). (C) Aristotelianessentialism. In contrast with El, Aristotelian essentialismtakes it that, if anythingis a man or a mammal,it is so necessarily.These are not propertiesthat anythingcan have per accidens. The same strong condition extends to properties (e.g., rational-animal)which are definitiveof a kind (e.g., man). Versions of this conditionare: (5) (x) (F(x) 3 c F(x)) (6) (x)o(F(x) oF(x)) (7) ci(x) (F(x) caF(x)) Conjunctionof one of (5)-(6) with Em or E* will give us some mode of Aristotelianessentialism. Since Aristotledid seem to presume accidental attributes,conjunction with E* is plausible, as in (8) (x) (F(x) D oF(x)) * (3x) (oF(x) *G(x) oG(x)) o(3x) - F(x) (E*A) Attributessatisfying some mode of Aristotelianessentialism are seen to be disjoint with what we call individuatingessences. El is a per- plexingthesis, although suggested by use. However, unless we have reason to suppose that EI is false,we can take (4) and (8) [or one of the alternativesto (8)] as sortingessential attributes. (D) Furthermodifications. As we noted above, a furtherrestric- tion on eligiblepredicates, in addition to generality,is that they be nonmodal; i.e., direct predicates. Paraphrasing Jevons, the essen- tialistis not intimatingthe mode or mannerof the mode or manner in which the predicate belongs to the subject. Indeed iterated ad- verbs,modification of modifiersrarely occur sensiblyin ordinarydis- course, althoughwithin the semanticsof QML we can make sense of iteratedmodalities. If, in additionto generalityand directness,which are requiredfor conformityto cases of essential attribution,we furtherrestrict our predicateswith some loose approximationof "natural" predication

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions ESSENTIAL ATTRIBUTION 199 in mind,an interestingresult follows with respect to the relationbe- tweende re and de dictomodalities. In our formallanguage we can form"artificial" predicates25 in the veryimprecise sense that transla- tion back into colloquial speech ranges fromextremely awkward to impossible.Consider the predicateformed from so simplea sentence as 'Someone offeredSocrates poison.' which goes into '(3x)P(xy)'. To what propertyof Socrates does it correspond?Is it the property of being offeredthe lhemlockby someone? Yes; but we can also see that predicatesthat contain quantifierseven when theyare just beyond minimalcomplexity border on the inexpressible.The same is true of predicatesthat contain sentenceparts. What propertydoes an object have if it satisfies'All ravens are black. x = x'? If we take as our stock of eligibleE-predicates those whichhave no quantifiers and no sentence parts, and are direct and general (N-predicates), tlhen,as Parsons27has shown,if EM is truewhere F is as above, then, forany nonmodal sentencesS, if S is not already a theorem,oS is not entailed by EM. Those, like Plantinga, who imagine that with sufficientcunning they can "reduce" the essentialist'sde re truthsto de dictotruths have not been sufficientlyattentive to these results. In our specificationof N-predicateswe could of coursesimply have required that they be built up out of atomic predicatesand truth- functionalsentential connectives. But thatwould have appeared post hoc. Our purpose was to framethese restrictionswithin the context of reasons foraccepting them. Here, as elsewherein the paper, when we say we are "characteriz- ing" Aristotelianessentialism and the like, we are not suggesting that such characterizationsare complete. There is good reason to believe that a completecharacterization of Aristotelianessentialism (if it is possible) would furtherrequire the introductionof temporal modalities.For otherwise,how would we say of an object that when it ceased to have its Aristotelianessence it would cease to exist altogether?Similarly, we are not supposingthat our N-propertiesfully correspondto naturalproperties (if thereis such a characterization).

24Since only the coniverse of the Barcan formula holds, (7) is stronger.We have omitteddiscussion of an importantquestion as to how,if we wereto importsuch truths,to apply the ruleof necessitation. 25Parsons' requirement of generality simplified the characterization of essential- ismfor nth-degree cases. Withan abstractionoperator (as in my"Essentialism in Modal logic,"and his "Gradesof Essentialismin QuantifiedModal Logic,") the ultimategenerality of the predicate can be preserved;e.g., 'aeXlx = a' can be trans- formedinto '(a,a)exy/x = y'. The generalityrequirement in additionfits some loosenotion of "natural"predicate. 26 See footnote17. 27 "Essentialismin QuantifiedModal Logic," pp. 47-48.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions 200 THE JOURNAL OF PHILOSOPHY

For, 'being a numberor else a philosopherand a cow' would count as an N-predicate. Further specificationwould require, perhaps among extensionsof QML with meaningpostulates, something like a theoryof categories.Still, the predicate'being a number ...'is at least intelligiblyexpressible. Let us returnnow to the question of the "commitment"of QML to essentialism.For Kripke it was sufficientto definea model in termsof a set of assignmentsof truthvalues to sentences,extensions to predicates,along withspecification of the domain (a subset of the union of domainsof membersof K) foreach K' or K. For any model short of a maximal model, there will be some object and some P-assignmentsuch that, in any K in which the object exists,it will be in the extensionof P. In this sense, we mightsay that, fornon- maximal models, QML is committedto essentialism,although no instance of minimalessentialism is a theorem.But, as I have suggested, essentialisttalk is frequentlyunproblematic. With careful specificationof how we paraphrase such talk in QML, we can characterizesome modes of essentialism.And, as I will claim forat least one mode, Aristotelianessentialism as here characterized,it is firmlyentrenched in the logic of causal statements.28 IV Considersome familiarexamples. I say of a sample (s) that if I put it in aqua regia (R) it would dissolve (D). We do not take such a claim to be unintelligible.Suppose we interpret'oj' of our QML as causal or naturalnecessity. Then our example may be representedas (9) o (R (s) n D (s)) Suppose I say of anothersample of some differentmaterial (u) that if I immersedu it wouldn'tdissolve, from which it would followthat (10) EJ(R(u) D (u)) and, therefore,that * (1 1) (3x) o (R (x) n D (x)) (3x) -o (R (x) n D (x)) which is an instanceof minimalessentialism. We may thinkof s as having the essentialattribute of eithernot being immersedin aqua regia or dissolving,in all worlds causally possible relativeto W. 28 Recent discussions of essentialism have focused on numerical statement and the like. Parsons showed that in QML, extended to include meaningpostulates and arithmetictruths, it can be done in such a way as wholly to avoid any essentialist consequence. Furthermore,there is an intuitive plausibility to these nonessentialist alternatives for construing analytic statements in the broad sense of 'analytic'. On a "natural" interpretationof the modalities, essentialismdoes not appear to be avoidable.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions ESSENTIAL ATTRIBUTION 20I

Anotherexample. Shylock tells us that if you prick him he will bleed and ifyou tickledhim he would laugh,and ifyou poisonedhim he would die and if you wrongedhim he would revenge. Each of Shylock's assertionscould be representedas in (9). If he had added, the same is not trueof a stone,then we have essentialism,and it is all perfectlycoherent. Let us suppose (9) is true and someone were to ask "Why?" Why would s dissolveif it was immersedin aqua regia? Why would Shylockbleed ifpricked, die ifpoisoned, laugh iftickled? An appropriateanswer to the firstquestion is "Because s is gold." To the second, if the question is not divided, "Because Shylock is a man." The answersto thesequestions begin with 'because', but what followsis not an event descriptionbut a statementthat attributes a kind propertyto the object. How is it that an object's being of a certainkind is a ground (and apparentlycausal) forits havingsome essentialproperty, or forthere being some causal or necessarycon- nectionbetween its nonessentialproperties, e.g., between the pairs (being immersedin R, dissolving), (being tickled,laughing), (being pricked,bleeding) etc.? Define the correspondingcausal conditionalas follows:

(12) Sc P =df O (S n P) then it must be that (9) followsfrom (13) (G(s).R(s)) -*0D(s) which in turninstantiates some generallaw. But which one? There are alternatives,since the modal operatormay be inside or outside the quantifier.If our QML choice is QS4 with the converse of the Barcan formula,then the weaker alternativeis (14) (x) ((G (x) *R (x)) -* D (x)) The differencebetween (14) and otherversions is illuminatingwith respect to questions of invariance of laws, but we will defersuch considerations.Our question now is, how do we get from(13) to (9), whichwe can rewriteas (15) R(x)-*0D(s) since only weakened exportationholds29 in QML? Restricted ex- portationon (13) gives us (16) G(s)--* (R(s) n D(s)) 29Failure of unrestrictedexportation is a characteristicof conditionalssuch as strictimplication and Andersonand Belnap'sentailment, which are strongerthan the materialconditional. The causal analogue of an unrestricteddeduction theoremis surelycounterintuitive.

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions 202 THE JOURNAL OF PHILOSOPHY

This is wlhereAristotelian essentialism comes in. Being gold or being a man is not accidental.Then theymust conformto one of (5)-(7). Therefore,from G(s) we get oG(s), which,together with the modal principle (P-- Q) - (oP oQ)I and (16), gives us (15). No metaplhysicalmysteries. Such essences are dispositional propertiesof a very special kind: if an object had such a property and ceased to have it, it would have ceased to existor it would have changed into somethingelse. If by bombardmenta sample of gold was transmutedinto lead, its structurewiould have been so altered, and the causal connectionsbetween its transientproperties that had previouslyobtained would so have changedthat we do not reidentify it as the same thing. Suppose forcomparison, G(s) had been exportedin (13) instead of R(s), (17) R(s) (G(s) D(s)) Since R (s) is not one of those special dispositionalproperties, no causal connectionobtains betweenbeing gold and dissolving;simply being gold would not count as a cause of dissolutionany more than simplybeing a man would alone be a cause of bleeding.On the otlher hand, although immersionin aqua regia (R(s)) is an accidental property,being aqua regia is not. Th-e general law would be in- stantiated as (18) (A (a) *G (s) *R (sa))c D (s) and if a were a sample of aqua regia and s a sample of gold, then it would followthat if s were immersedin a, then s would dissolve. Given that there are objects that do not dissolve wlhenthey are immersed,wNTe have (19) (3x) (3y)o (R (x,y) D (x))- (]x) (]y) - (R (x,y) D D (x)) wlhich,in a perfectlyintelligible way, commits us to essential relations. Rather than leadinginto a metaphysicaljungle, it seemsto me that formulatingour analysiswithin QML is suggestiveand illuminating. Differentmodes of essentialismwill place differentinterpretations on the modalities.But so far as the causal interpretationgoes, it suggests interestingrelations between laws like (14) and the paradox of confirmation.It allows forsolutions to problemsabout substitution in causal contexts. It raises interestingquestions about the generalityof laws. RUTH BARCAN MARCUS Nortlhwestern University

This content downloaded from 150.209.80.69 on Wed, 4 Dec 2013 15:38:24 PM All use subject to JSTOR Terms and Conditions