13 «a Three Grades of Modal Involvement ‘There are several closely interrelated operators, ealled modal ‘operators, which are charactritio of modal loge. ‘There are the operators of neccssity, poenbility, impossibility, non-necesity. Also there are the binsry operators, or connectives, of strict ‘plication and strict equivalence. These various operators are casly definable in terms of one another. Thus impossibility is necessity of the negation; possibility and non-nocessity are the negations of impossibility end necessity; and strict implicetion ‘and strict equivalence are necessity of the material conditional snd biconditional. In a philosophical examination of mada! logic ‘we may therefore conveniently limit ourselves for the most part ‘to a single modal operator, that of nzcessty. Whatever may be said about noesssity may bo ssid also, with essy and obvious ‘adjustments, about the other modes. ‘There are three different degrees to which we may allow our logic, or semantics, to embrace the idea of necessity, The frst or Teast degree of acceptance is this: necessity is expressed by a semantical predicate attributable to statements az notational forms—hence attachable to names of statements. We writ, ees rom the Precedings of the XI International Congress of Phiosophy, ‘Brule 183, Volume 14 (Amucerdans Nort alld Pubbsting Co), ORADES oF MODAL INVOLVEMENT 167 o Neo > 8, % Neo (Sturm's theorem), @ [Nee ‘Napeleon escaped from Elbe’, in esch ease attaching the predicate ‘Nec’ to « noun, a singular term, which is a name of the statement which is affirmed to be necessary (or necessarily true). Of the above examples, (1) and (2) would presumably be regarded as true and (3) as false; for the necessity concerned in modal logic is generally conceived to be ofa logical or a prior sort. ‘A eeeond and more drastic degree in which the notion of necessity may be adopted isin the form of a statement operator. Here we have no longer « predicate, attaching to names of statements asin (1)~(B), but logical operator ‘ne’, which at= taches to statements themselves, in the manner of the nogation sign. Under this usage, (1) and (8) would be rendered rather as: @ neo (8 > 9), © ee (Napoleon escaped from Eb), ‘and (2) would be rendered by prefixing ‘nee’ to Sturm's eetusl thoorem rather than to its sme. Thus whereas ‘Neo! ia a predieate or verb, is necesary’, which attaches to a noun to form 4 statement, ‘nec’ is rather an adverb, ‘necessarily, which ‘Attaches to a atstement to form a statement. ‘Finally the third and gravest degre is expression of necesity by a sentence operator. This is an extension of the seeond degree, and goes beyond it in allowing the attachment of ‘nee’ not only to statements but also to open sentences, such a8 “z > 5, prepare tory to theultimate attachment of quantifiers: © @) nee (2 > 5), @ Ge) neo 2 > 8), © le = 9.D neo (e > 8) ‘The example (6) would doubtless be rated as false, and perhape (7) and (8) as true T shall be concerned in this paper to bring out the logical and philosophical signfieanee of these three degres of acceptance of a necosity device,168 ‘The Ways of Paradox 1 Teall an occurrence of « singular term in & statement purely rajerentia® (Frege: gerade"), if, roughly speaking, the term serves in that particular context simply to refer to its object. Occurrences within quotation are notin general referential; ex, the statemente: ® ‘Cicero! contains sx letors, 0) ‘9 > 8 contains just three charactors say nothing about the statesman Cicero or the number 9. Frege's criterion for referential oceurrence is substitutvity of identity. Since ay ‘Tully = Cicero, (2) ‘the mumber of planets = 9, ‘whatover is true of Cicero is true ipso facto of Tully (these being fone and the seme) and whatever ie true of 9 is true of the number of planeta, If by putting “Tully” for ‘Cicero’ or ‘the number of planets’ for‘ in a truth, eg, (9) oF (10), we come out with a falethood: as) ‘Tully’ contains sx letters, (14) ‘the mamber of planets > 8 contains just three characters, vwe may be sure thet the position on which the substitution was made was not purely referential. (9) must not be confused with: as) Cicero has a sit-tter name, which dace say something about the man Cicero, and—unlike (9) remsine true when the name ‘Cicero! is supplanted by ‘Tully’ ‘Taking a hint from Russell? we may speak of a context as ‘eferentially opaque when, by putting a statement ¢ into that context, we ean eause s purely referential occurrence in g to be not purely referential in the whole context Bg, the conte contains just three charneters 2 From a Logical Point of Vin, pp. TE, 1908, 16 per Bina und Bedoong” ‘Whihead and Rowall 4/od, Vol 1, Appendix C. is referentially opaque; for, the oocurrence of 9" in ® > 5" is purely referential, but the oeurence of ‘9 in (10) isnot. Briefly, ‘context is referentially opaque if it ean render a referential ocurrence non-referental. Quotation is the referentally opaque context par excellence. Intuitively, what occurs inside @ referentally opaque context may be looked upon as an orthographic accident, without logical status, like the occurrence of ‘et’ in ‘cattle. The quotational context ‘9 > 5°" of the statement ‘9 > 5° has, perbaps, unlike the context “estle’ of ‘eat’, a docoptively systematic air which tempts us to think of its parts as somehow logically germane. Insofar as this temptation exists, it is salutary to paraphrase ‘quotations by the fllowing expedient. We may adopt names for each of our letters and other charactere, and Tarsk's “ to ex: pres concatenation. Then, instead of naming a notational form by putting that notational form itself bodily between quotation rarks, we can name it by spelling it. Eg, since y”is mu, "is is nu, the word ‘ao? ismuepsilon"nu. Similarly f, if we adopt the leters and "Fas names of the characters 9, ‘>, and ‘5. The example (10) an thus be tanseribed as 16) gf contains just three characters “Here there is no non-referential occurrence of the numeral, for there is no occurrence of it all; and here there is no referentilly ‘opaque containment of one statement by another, beesuse there is no contained statement at all. Paraphrasing (10) into (16), #0 fas to get rid altogether of the opnquely contained statement ‘9 > 5}, i lke paraphrasing ‘eatl’ into ‘kine’ so as to rid it of the merely orthographic ooourrence of the term ‘eat. Neither paraphrase is mandatory, but both are helpful when the irrefer~ entia) oreurrences draw undue attention, ‘An occurrence of a statement as part ofa longer statement is called truth-functional if, whenever we supplant the contained statement by another statement having the same truth value, the containing tetement remains unchanged in truth value, Nate rally one would not expect occurrences of statements within referentially opaque contests, ich ss quotations, to be truth- fonctions. Eg, the truth (10) becomes false when the contained statement ‘9 > 5! is supplanted by snother, ‘Napoleon escaped160 The Ways of Parador from Biba’, which has the same truth value a9 > 8 Again the truth (1) is carried, by that same substitution, into the falsehood (3). One might not expect occurrences of ‘statements within statements to be truth-funetional, in general, even when the contexts aro not referentilly opaque; certainly not when the ‘contexts ere referentally opaque. Tn mathematial logic, however, a poliey of extensionality is ‘widely espoured: a policy of admitting statements within state ‘ments truth-funetionally only (apart of course from such con texts as quotation, which are referentally opaque). Note thet the semantical predicate ‘Nec’ as of (1)-(8) is roconcilable with this policy of extensonality, since whatever breach of extensionality it prima facie involves is shared by examples like (10) and attributable (o the referential opacity of quotation, We ean always ewitch to the spelling expedient, thus rewriting (1) as: an Nee (oD, (17), lke (26) and indocd (2) and unlike (1) and (8), eontaing ‘no component statement but only a name of a statement ‘The statement operator ‘nee’, on the other hand, is & premedi- tated departure from extensionality. The occurrence of the truth ‘9 > 3°in (4) is non-truth-funetonal,sinee by supplanting it by 1 differont truth we ean turn the true context (4) into a falsehood ‘auch as (5). Such occurrences, moreover, are not looked upon ‘somehow spurious or irrelevant to logical structure, like oocur- rengea in quotation oF like ‘eatin ‘atte’. On the contrary, the rodal logic typified in (4) is uevally put forward ae a corrective of extensionality, a needed supplementation of an othervise impoverished logic. Truth-functional ooeurrence is by no means ‘the rule in ordinary Ienguage, as witness occurrences of state- ments governed by ‘because’ ‘thinks that’ ‘wishes that’, ete, as well as ‘necessarily’. Modal logicians, adopting ‘nce’, have seen no reason to euppose that an adequate logic might adhere to a poliey of extensionality. But, for all the willingness of modal logicians to flout the policy of extensionality, is there really any differeneo—on the score of extensionality——between their statement operator ‘ne’ fand the extensionally quite admissible semantical predicate ‘Neo’? The liter was excusable, within a policy of extension ity, by citing the referential opacity of quotation. But the @Rapns oF wovar ixvouvewexT 161 statement operator ‘ne’ i likewise excusable, within & paliey of extensionality, by citing the referential opacity of ‘ne’ itself! To see the referential opacity of ‘ues? we have only to note that (4) ‘and (12) are true and yet this is fl a8) nee (the numberof planets > 5). "The statement operator ‘net isin short, on a par with quotation (1) happens tobe written with quotation marks and (4) without, ‘but from the point of view of s policy of extensionality one is no worse then the ather, (1) might be preferable to (4) only on the srore of a possible ancillary policy of trying to reduce referen- tially opaque contexts to uniformly quotstional form. Genuine violation of the extonsionality policy, by admitting ‘non-truth-funetional occurrences of statements within statements without referential opacity, sles easy than one wt Sst pposes xtensionality does not merely recommend itself on the score of simplicity and convenience; it rests on somewhat, more eompel- ling grounds, ae the following argument will reveal, Think of ‘p" 1s shor for some statement, aad think of ‘F(p)’ as short for some containing true statement, euch that the context represented by “Fis not relerentially opaque, Suppose further that the eontext represented by ‘P' is such that logical equivalents are inter- changeable, within it, alud vertate. (This ie true in particular of “nec) What I hall show is that the occurrence of ‘pin ‘F(p)" is ‘then truth-funetional. Le, think of ‘as short for some statement having the same truth value as ‘p'; I shall show that 'F(q)" is, like ‘F ()’, true. What 'p' represents is @ statement, hence true or false (and ovoid of freo‘2’).If'p'is true, then the conjunction 'z = A. pis ‘true of one and only one object z, vis, the empty elass A; whereas if ifalae the conjunction ‘x =. pis true of no object x what- fever. The class #(2 = Ap), therefore, is the unit class A or A itaelf agcording ae ‘p’ i true or fal. Moreover, the equation: ae = Ap) =k is, by the above considerstions, opcaly equivalent to ‘p!. Then, since ‘F(p)’ is true and Iogieal equivalents are interchangeable within it, this will be tru: a) Rite = Ap) =a)2 The Ways of Perador Since ‘p' and ‘qa ake in truth value, the clases #(2 = A. p) ‘and #@¢ = Ag) are both cA oF both A; 50 eo) te = Apo tes Ag. ‘Since the context represented by ‘isnot referentially opaque, the ‘eourrence of ‘202 = Ap)" in (10) is a purely referential occu renee and hence subject to the subetivutivty of identity; so from (19) by (20) we can conclude that Fide = 4.9) = «ah ‘Thenee in turn, by the logical equivalence of #(@ = A.) = at! to, we conclude that Fg). ‘The above argument cannot be evaded by denying (20), as Jong ss the notation in (20) is construed, as usual, as referring to clasts, For clases, properly so-called, ere one and the sume if their members are the same—regardless of whether that sameness bbe a matter of logical proof or of historical accident. But the ‘argument could be contested by one who does not admit class nenes 2( . . )?. Tt eal alto be contested by one who, though ‘admitting such class names, does not see a final criterion of referential orcurrence in the substitutivity of identity, se applied to constant singular terms, These points will come up, perforce, when we tum to ‘nec’ a a sentence operator under quantification. Meanwhile the sbove argument does serve to show that the policy of extensionality has more behind it than its obvious Smpliety and convenience, and that any real departure from the poliey (at least where logicsl equivalents remain interchange- fable) must invalve revisions ofthe logie of singular term. ‘The simpler earlier argument for the referential opacity of the statement operator ‘nee’, viz, observation of the truths (4) and (12) and the falsehood (18), could likewise be contested by one ‘sho either repudiates constant singular terms or questions the triterion of referential opacity which involves them. Short of ‘adopting ‘nee’ as a full-edged eentence operator, however, 20 such searching revisions of classical muthemutical logic are required, We can keep to a classical theory of classes and singular terms, and even to a poliey of extensionality. We have only to recognie, in the slatement operator ‘ne', a referentially ‘opaque context comparable to the thoroughly legitimate and very convenient context of quotation, We can even look upon (4) and (6) as elliptical renderings of (1) and (3) GRADES OP MODAL INVOLVEMENT 168 Something very much to the purpose of the semantical predi- cate ‘Nee’ is regularly needed in the theory of proof. When, e, ‘we speak ofthe completeness of a deductive system of quantifien- ton theory, we have in mind some eoncept of validity as norm ‘with which to compare the class of obtainable theorems, The notion of validity in such context isnot identifiable with truth, A true statement is uot a valid statement of quantification theory unless not only it but all other statements similar to it in ‘quantitiestional structure are true. Definition of such a notion of validity presents no problem, and the importance of the notion {or proof theory is incontestable ‘A conepicuoue derivative of the notion of qusntifiationsl validity is that of quantifetional implication, One statement quantiicetonally implies another if the material conditional composed of the two statements is valid for quantification theory ‘This reference to quantification theory is only illustrative, ‘There are parallele for truth-function theory: statement is valid for teuth-funetion theory if it and all statements like it in truth-funetional structure are true, and one statement truth- functionally implies another if the material conditional formed of ‘the two statements is valid for truth-function theory. And there are parallels, again, for logie taken as a whole: statement is logically valid if it and all statements Hike it in logical structure are true, end one statement logicslly implies nother ifthe maleril conditions formed of the two statements is logically vali. ‘Modal loge roesived special impetus years ago from a confused reading of ‘9’, the material ‘then’, as implies’: eonfusion of ‘the material conditional with the relation of implication Prop- ceri, whereas ‘5 or f-then’ connects statements, ‘implioe’ is a ‘verb which comieets names af statements and thus expresses. rela tion of the named statements, Carelessoess over the distinction of use and mention having allowed thie intrusion of “implies! as a reading of’, the protest thereupon arose that ‘>’ in its material sense was too weak to do justice to ‘implies’, which connotes some- “Notably in Whitehead end Rossel108 ‘The Ways of Paradox thing like login! implication. Accordingly an effort was made to repair the diserepaney by introducing an improved substitute for ‘5, written ‘2’ and ealed strict implication* The initial failure to distinguish use from mention persted; so ‘', though read ‘implies’ and motivated by the connotations of the word ‘implies’ funetioned actually not asa verb but asa statement connective, & ‘uch strengthened i-then'. Finally in recognition ofthe fact that logical implication is validity of the material conditional, validity, operator ‘noe’ was adopted to implement the definition of ‘p 2” ‘nee (p > 4). Sinco‘3” had boon eft atthe level ofa statement connective, ‘nee in turn was of eourse rendered as an operator . p V @ (ith just those letters). ‘Moreover, the predicates ‘implies’ and ‘Nec’ have thus far been looked upon as true only of statements, not of schemata; eon (27) ‘and (28) they are misapplied (pending some deliberate extension of sag). ‘The letters ‘p’ and ‘in (25) and (20) stand in ple of state~ ‘ments For translation of (25) and (28) into semantiea form, on the other hand, we need some special variables which refer 1 stato- ‘menta and thus stand in place of names of statomenta. Let us use ‘¥,W, ete, for that purpose. Then the analogues of (25) and (23) in semantical form eaa be rendered: 9) ‘4 implies the sltemation of ¢ and ¥, (80) Nee the conditional of ¢ with the alternation of and ¥). ‘We ean condense (29) and (80) by use of a conventional notation which T have elsewhere called guasi-quotation, thus: oo implies" VV, @) Neo D.6 VV" ‘The relationship between the modal logic of statement opera ‘tore and the semantieal approach, whieh was pretty simple and ‘obvious when we compared (21)(22) with (23)-(24), is thus teen to take on some slight measure of subtlety at the stage of * Mathematinl Logi,168 The Ways of Paradox (25)-(26); these correspond not to (27)=(28) but to (31)=(32). It is schemata like (25)=(26), moreover, and not actual state- rents ike (21)-(22), that Bl the pages of works on modal logic. However, be that as it may, it is in gotual statements such a6 (21)-(24) that the point of modal loge les, and itis the come parison of (21)-(22) with (28)~(24) that reflects the true rela tionship between the use of statement operators and that of semantion) predicates. Schemata such as (25)~(26) are mere heuristic devices, useful in expounding the theory of (21)~(22) nd their ike; and the heuristic devices which bear similarly on (23)-(24) are (31)~(22). Seeing how modal statement operators can be eonverted into semantieal predicates, one mny of course just note the conversion ‘8 & principle and leave it undone in practice. But there are five reasons why i is important to note it in principle, One is that the inclination to condemn ‘5’ unduly, through a wrong association of ‘if-then’ with implies, is thereby removed. A second reason is ‘that iti at the semantidl or proof-theoretie level, where we tal ‘bout expressions and their truth values under various substitu tions, that we make clear and useful senge of logiel validity tnd itis lopiel validity that comes nearest to being a clea ‘elication of ‘Nec, taken as a semantical predicate. A third reason ie that in using ‘Nec’ as a semantical predicate we Saunt 0 {milisr reminder of referential opaity, inthe form of quotation marks. A fourth reason is that the adoption of ‘nes’ at ‘statement operator tempts one to goa step further and use it a8 a sentence operator subject to quantification. The momentousness of this further step—whereof more anon—tends to be overlooked save as one expressly conceives of the ‘nee’, in its use as statement operator, ae shorthand for the semantical usag 'A fifth reason bas to do with iteration. Sinee ‘nes’ attaches to = statement, and produces a statement, ‘nee’ ean then be applied ‘agnin, On the other hand ‘Nee! attaches to.» name and yields a statement, to which therefor, it cannot be applied again. An iterated ‘ne’, eg 3) ee neo(2) (xis red 3 is red), ean of course be translated by our regular procedure into semantieal form thus: 64 Nee ‘Nee (2a is red D 2 is red)’ GRADES oF MODAL INVOLVEMENT 167 ‘and we are thereby reminded that ‘Nec’ can indeod be iterated if We insert new quotation marks as needed. But the fect remoains ‘hat (84) is, in contrast with (33), an unlikely move. For, suppose we have made fair sense of ‘Nec’ as logical validity, relative say to the logic of truth functions, quantifation, and perhaps elastes, The statement (5) ()(eis red D zis red), then, is typical of the statements to which we would attribute such validity; 80 @) Nee (2) (eis red D x is red’, ‘The validity of (25) resides in the are all other statements with the eame quantieational and truth-funetional structure es (95). Thus itis that (96) is true. But if (86) in turn is also valid, itis valid only in an extended senso with which we are not likely to have been proviously ‘concerned: a sense involving not only quantifieational and truth- functional structure but also the semantieal structure, somehow, ‘of quotation and ‘Ne’ itself Ordinarily we work in « metalanguage, as in (36), treating of an object language, exemplified by (85). We would not rise to (24) except in the rare case where we want to treat the metalanguage by means of tel, and want furthermore to extend the notion of validity beyond’ the semantics of logic to the semantis of seraantics, When on the other hand the statement ‘operator ‘ne is used, iteration as in (98) i the most natural of steps; and i is significant that in modal loge there has boen some ‘question as to just what might most suitably be postulated regarding such iteration.* "The iterations need not of eourse be consecutive, Inthe use of ‘modal statement operators we are led also into complex iterations such a cc) p3qanaaen, short for: 8) nce nee (p D 4) D nee (~g > ~168 ‘The Ways of Paradox 0, to take an actual exemple: 69) (2a has mass) 3 (e)(z has mass). ‘~ G@e)(c has mass) 3 ~ ()(c has mass), 40) ree {noe [(2}(z has mass) > (B2)(x has mass)] > nee (~(3e)(e has mass) ~ (2)(2 has mas} In terms of sementical predicates the correspondents of (39) sand (40) are: co) **G2)(e has mass)’ implies (32) (2 has rows)’ implies "'n(@o)(e us mas) implies ‘~(2) has mass)", 2) ‘Nos ‘Neo ‘(2)(¢ bas mast) > (@2)(z has mass)" > ‘Nee '~(Ge)(e has mass) ~ (2)(2 has mass)" But (41)~(42), like (34), have singularly litle interest or moti- vation when we think of necessity semantieely. Tis important to note that we must not translate the schemata (37)-(@8) into semantical form in the manner: “pimples "implies te To do o0 would be to compound, to an altogether horrifying degre, the error noted earlier of equating (25)~(26) to (27) (28), ‘The analogues of (87)~(88) in sementical application should be rendered rather: (43) impos ¥* implies *—9" implies “—g™, as) Neo "Neo" D WD Neo'¥D ~#7, subject to some special conventions governing the nesting of ‘quasi-quotations. Such conventions would turn on certain subtle ‘considerations whieh will not be entered upon here. Suice it to recall thatthe sort of thing formulated in (33)=(34) and (37)— (44) is precisely the sort of thing we ate likely to see least point in formulating when we think of necessity etrielly ae a ematical predicate rather than a statement operstor, It impressive and significant that most of modal logie (short of quantified modal logic, to which we shall soon turn) is taken up with iterated cages like (83) and (87)~(40) which would simply not recommend themselves to our attention if necessity were held to the status of 1 sementical predicate and not depressed to the level of a state- ‘ment operator. (ur reflections have favored the eemantical side immensely, GRADES oF MODAL INVOLVEMENT 160 but they must not be allowed to abseure the fact that even as @ semantical predicate necessity ean raise grave questions. There is no dificulty as long as necesity is construed as validity relative fay tothe logie of truth funetions and quantifiestion and perhaps classes, If we think of arithmetic as reduoed to class theory, then ‘such validity covers aleo the truths of arithmetic. But one tends to include further territory alll; eases such as ‘No bachelor is rarried, whose truth ie supposed to depend on “meanings of tems" or on “synonymy” (eg, the syponymy of ‘bachelor’ snd ‘gaan not marvied’). Tho synonymy relation on which such ca depend is supposedly narrower relation than that of the mere cooxtensiveness of terms, and it isnot known to be amenable to ‘any satisfactory analysis. In short, necessity in semantical ‘application tende to be identified with what philoeophera call ‘aualytiity; and snalyticty, T have argued elsewhere,” is a peeutdo-concept which philosophy would be better off without. ‘As long as necessity in somantical application is construed simply as explicit truth-funetional validity, onthe other hand, or ‘quuntificational validity, or sat-theoretie validity, or validity of Any other well-determined Kind, the logic of the semantical necesity predicate is a significant and very central strand of proof theory. But itis not modal logic, even unquantified modal Jogie, sa the latter ordinarily presenta itself; for it isa remarke- bly ‘meager thing, bereft of ll the complestios which are encouraged by the use of ‘neo! as a statement operator. It is tunguantified modal loge minus all principles which, explicitly or implicitly (via, ete), invelve iteration of nevesity; and plu, it we are Kteral-minded, a pair of quotstion marks alter exch "Nev. ‘Having adopted the operator ‘~" of negation as applicable to statements, one applies it without second thought to open sentences 43 well: sentences containing free variables ripo for ‘quantification, ‘Thus we can write not only ‘~ (Socrates is ‘mortal)’ but also '~ (zis mortal’, from whieh, by quantifieation Two dogmas of empiri.”170 ‘The Ways of Parador and further negation, we have ‘~ (2) ~(z is morta)! or briefly *(1z)(z is mortal)’. With negation this i as it shouldbe. As long. as ‘nee’ is used esa statement operator, ou a par with negation, the analogous course suggest itself again: we write not only “nee (9 > 5) but also ‘nee (e > 5), from which by quantifiation we an form (6)-(8) and the like "This step brings us to ‘nee’ as sentence operator, Given ‘nec as statement operator, the step is natura. Yet it is drastic one for it suddenly obstructe the earlier expedient of translation into ‘tomas of ‘Nee! as semantical predicate. We ean reeonstrue (4) and (5) at will as (1) and (8), but we eannotreconstrue: a5) nee (2 > 5) correspondingly a a) Neo's > 8. ‘Nee’ has been understood up to nov 26 a predicate true only of statements, wheross (46) attribute it rather to an open sentence and is thus trivially false, at least pending some deliberate cxtension of usage. More important, whereas (45) is an open sentence with fre ‘x, (46) has no corresponding generality; (48) js simply a statement about a specific open sentenoe. For, it must be remembered that ‘¢ > 5” in quotation marks is « name of the specie quoted expression, with fixed letter ‘2, ‘The ‘2’ in (46) ‘eannot be reached by a quantifier, To write: oO) (e)(Nee's > 53, G)(New's > 51) is lke writing: (48) (@)(Goerntes is mortal), x) (Socrates is mortal); ‘the quantifier is followed by no germane oteurrence of its ‘arable, In a word, necessity as sentence operator does not go cover into terms of neoessity a8 semantical predicate, ‘Moreover, acceptance of necessity as sentence operator implies an attitude quite opposite to our earlier one (in §§1-II fbove), which was that ‘nee’ ae statement operator i referen- tially opaque. For, one would clesrly have no business quantify- ing into a referentialy opaque context; witness (47) above, We ean reasoasbly infer (3z) neo (x > 5)” from ‘nee (@ > 5)" only i we regard the latter as telling us something about the object ®, 1 number, via. that it necesurly exceeds 5. If ‘nee (.. . > 5)’ ‘an turn out true or false “of” the number 9 depending merely on hhow that number is referred to (as the falsity of (18) suggests), then evidently ‘nee (x > 5)’ expresses no genuine condition on ‘objects of sny kind. If the oceurrence of '9” in ‘nee (9 > 5) is not purely referentisl, then putting ‘efor 9" in ‘nee (9 > 6)" makes rho more sense than putting ‘x’ for ‘nine’ within the contest ‘But jan’ it settled by the truth of (4) and (12) and the falsity of (18) thatthe occurrence of 9” in question is irreferental, and more generally that ‘neo’ is referentially apaque, and henoe that ‘nee! as a sentence operator under quantitirs is « mistake? No, not if one is prepared to accede to certsin pretty drastic departures, as we shal see "Thus far me have tentatively condemned necessity as general ‘entence operator on the ground that ‘nce’ i referentially opaque Ite referential opacity has been shown by breakdown in the ‘operation of putting one constant singular term for another which names the same object, But it may justly be protested that constant singular terms are a notational accident, not newded at ‘the level of primitive notation For itis well known that primitivly nothing in the way of sin- gular terms is needed except the variables of quantification thom- selves, Derivatively all manner of singular terms may be intro- duced by contextual definition in conformity with Russell's theory of singular descriptions, Class names, in particular, whieh figured in the general argument for estensionality in §1 above, may be got either by explaining #8..." a8 short for the contextually defined description “(y) (2) (rey... )” or by. ‘adopting a separste st of contextual definition forthe purpose.” ‘Now the modal logicien intent on quantifying into ‘nee! sentences may say that ‘ne isnot referentielly opaque, but that it merely interferes somewhat with the eontertual definition of singular terms. He may argue that ‘(3z) nee (> 8)' is not meaningless but true, and in perticular thet the number 9 is one of the things of which ‘nee (x > 8)’ is true, He may blame the zeal oF apparent diserepancy in truth value between (4) and (18) simply on a queer behavior of contextually defined singular forms. Specifically he may hold that (18) is true if construed as: MCL my Methode of Love, 36-38; Mathematica! Logi, $24, 6v2 The Ways of Parador (49) Ga)fthere are exactly ¢ planets. nee (e > 8)] ‘nd fale if construed as (60) neo @x)(there are exsely 2 planets. x > 5), and that (18) as it stands is ambiguous for lack of a distinguish- jing mark favoring (49) or (80). No such ambiguity aries in the contextual definition of a singular term in extensional logo ( Jong as the named object existe), snd our modal logician may ‘well deplore the complications which thus issue from the presence of ‘nee in his primitive notation. Still he ean fairly protest that the erratic behavior of contextuelly defined singular terms is no reflection on the meaningfulnes of his primitive notation, incud- ing his open ‘nec’ sentences snd his quantification of thom. Looking upon quantification as fundamental, and constant singular terms as contextually defined, one must indeed concede ‘the inconelusiveness of a erteran of referential opacity that rests ‘on interchanges of constant singular terms. The objects of theory are not properly describuble as the things named by the singular terme; they are the values rather, of the variables of ‘quantification. Fundamentally the proper criterion of referen- tial opacity taras on quantiietion rather than naming, and is thie: a referentilly opaque context i# one that eanzot properly be quantified into (with quantifier outside the context and variable inside). Quotstion, again, i the referentislly opaque context par excellence; ef. (47). However, to object to necessity ss sentence operater on the grounds of referential opacity #0 {defined would be simply to beg the question. Frege's criterion of referential ovcurtence, vis, substitutivity of identity, underlay the notion of referential opacity as devel- ‘oped in §1 above, The statements of identity there concerned were formed of constant singular terms; ef. (11), (12). But there fs more fundamental form of the law of substituivity of identity, which involves no constent singular terms, but only variables of quantification; viz. wo We = ¥.D.Fe = Fy. "This law is independent of any theory of singular terms, and cannot properly be challenged. For, to challenge it were simply to Ths Sian, 0 From a Lojizl Pint of View, p. 12, Tt, 12-110, 13, 68 OxApes oF wopaLixvonvewey? 175 use the sign ‘= in some unaccustomed way irrelevant to our inquiry. In any theory, whstever the shapes of its symbols, an ‘open sentence mhose fre variables are ‘2’ andy! isan expression of identity only in ease it fulills (61) in the role of ‘z= y'. The igenerality of F” in (51) is this: ‘Pr’ is to be interpretable as any ‘open sentence of the system in question, having ‘r’ as free (quantifiable) variable; and ‘Fy’, of course, is to be a eorre- sponding context of 'y It ‘nee is not referentially opaque, F2’ snd ‘Fy’ in (61) ean in particular be taken respectively as ‘neo (2=2)" and ‘neo (c= y)". From (61), therefore, since surely ‘nee (z=) is true forall z, we have: @) We = v.D nee (@ = WL Le, identity holds necessarily if it hols at all. Let us not jump to the conclusion, just because (12) is true, that 69) noe (the nuraber of planets = 9) ‘This does not follow from (12) and ($2) except with help of a law of universal instantiation, allowing us to put singular terms “the number of planets’ and ‘9 for the universally quantified “2 and “y’ of (52). Such instantistion is allowable, certainly, in ‘extensional logic; but it is a question of good behavior of ‘constant singular terms, and we have lately observed that euch bbchavior is not to be counted on when there is a ‘nee’ ia the oodpile So our observations on necessity in quantifieational application are, up to now, as follows. Necesty in such application is not prima facie absurd if we ecept some interference in the contextual definition of singular terms, The effeot of this interfer~ enco is that constant singular terms cannot be manipulated with ‘the customary freedom, even when their objects exet In partieu- lar they cannot be used to instantiate universal quantiications, unless special supporting lemmas are at hand. A further ffect of necessity in quantifcstional application is thet objects come to De necessarily identical if identical at all ‘There is yet a further consequence, end « particularly striking fone: Aristotelian essntialis, Tie ie the doctrine that some of the attributes ofa thing (quite independentiy of the language in1% The Ways of Paradox which the thing is referred to, if at all) may be essential to the ‘thing, and others accidental. Eg, a man, ot talking animal, o featherless biped (for they are in fact all the same things), is essentially rational and aesidentally two-legged and talkative, not merely qua man but qua itself. More formally, what Aristotelian essentalism cays ie thet you ean have open sen- tences—which I shall represent here as ‘Fo! and ‘Gz'—sueh that a) x)(nce Fr. Gr, ~ nee G2). An example of (64) related to the falsity of (68) might be: (@z)[oce(e > 8) there aro just 2 plancts. ‘nes (there are just x planets), such an object z being the number (by whatever name) which is variously known as 9 and the numberof planets. How Aristotelian excentialiem as above formulated is required boy quantified modal logic can be quickly shown, Actually something yet stronger can be shown: thet there are open suntenees Ps" and ‘Gr’ fulfilling not merely (64) but: (2)(nee Fr. Gz. ~ nee 02), (@) noc Fr. (2) Ge (2) ~ nee Gr. ‘An appropriate choice of ‘Fs’ is easy: ‘x = 2". And an appropriate choige of ‘G2 is ‘x = 2. where in place of ‘p” any statement is chosen which is true but not necessarily true. Surely there és such ‘statement, for otherwiso ‘ne’ would be a vacuous operator and ‘hore would be no point in modal loge. ‘Necewity as semantieal predicate refeets s non-Aristotelian view of necessity: necessity resides in the way in which we ssy ‘things, and not in the things we talk about, Necessity as statement operator is capable, we saw, of being reeonstrued in terme of necesity as a semantical predeste, but has, noverthe- leas, ite specisl dangers; it makes for an’ excessive and idle elaboration of laws of iterated modality, and it tempts one £0 4 final plunge into quantified modality. Thie lst complicates the logic of singular terms; wore, it lends us back into the meta- physical jungle of Aristotelian essentials.