00:0 (2015) 114

NOUS
doi: 10.1111/nous.12126

Quantified Modality and Essentialism1

SAUL A. KRIPKE
The Saul Kripke Center and the Graduate Center of the City
University of New York
It is well known that the most thoroughgoing critique of modal logic has been
that of W.V. Quine. Quines position, though uniformly critical of quantified modal
systems, has nevertheless varied through the years from extreme and flat rejection
of modality to a more nearly moderate critique.
At times Quine urged that, for purely logico-semantical reasons based on the
problem of interpreting quantified modalities, a quantified modal logic is impossible; or, alternatively, that it is possible only on the basis of a queer ontology which
repudiates the individuals of the concrete world in favor of their ethereal intensions.
Quine has also urged that even if these objections have been answered, modal logic
would clearly commit itself to the metaphysical jungle of Aristotelian essentialism;
and this essentialism is held to be a notion far more obscure than the notion of
analyticity upon which modal logic was originally to be based.
If I have understood correctly the revised version of Quines From a Logical
Point of View (henceforth, LPV),2 only the last objection, the objection to modal
essentialism, is retained in this most recent statement. But the objectionable
features of essentialism still seem, in Quines opinion, to be sufficient reason to
declare, so much the worse for quantified modal logic.3 (LPV: 156)
The present paper attempts to analyze this objection; and my conclusion, indeed,
is that Quines fears of the notion, though by no means groundless, nevertheless do
not count as any objection to quantified modal logic, but only to certain applications thereof. I shall state these points with greater precision after a brief review of
some fundamental concepts.
Let us initially make a distinction between the pure and the applied semantics of
a formal system of modal logic, say the quantified modal logic that I have called
S5. In Kripke (1959a) I have given a semantical treatment of this system and have
shown it to be complete; in fact, the entire argument can be carried out within
Zermelo set theory. But the semantics in that paper was not intended merely to
be set-theoretic; it had an intuitive content. The members of the set K were to be
understood as so-called possible worlds;4 and a proposition (or sentence) was to
be necessary iff true in all possible worlds.
Here the notions of necessity and possible world may be interpreted in several
ways; perhaps the most common interpretation identified necessity with analyticity.
(I will explain below how one might interpret possible world.) This interpretation
of necessity as analyticity gives one an applied semantics for a system of modal

C

2015 Wiley Periodicals, Inc.

NOUS

logic; but this semantics is by no means the only applied semantics one might set
up.
As Quine himself has pointed out (1976: 169), necessity might be construed more
narrowly, as validity with respect to the logic of truth functions and quantifications
and perhaps classes. Or it might be construed more liberally, as say some sort of
physical necessity. But, if we are dealing with a single system of modal logic, all
these alternative interpretations, giving different types of applied semantics, will
nevertheless yield semantical notions having a common mathematical structure;
and this mathematical structuresay the one set up in Kripke (1959a)will be the
pure semantics of the theory.
The distinction between pure and applied semantics can be made clear by analogy
with the ordinary two-valued truth-functional propositional calculus. In Principias
classic presentation, the calculus was developed syntactically, using a specified set
of axioms and rules of inference. A semantical analysis of the system was discovered
later by Post and Wittgenstein.5 Formulae of sentential logic were to be assigned
truth-values T or F according to certain rules. A formula was to be tautologous if
it had the value T under all possible assignments; and the theorems of Principia
demonstrably coincided with the tautologies. This was a pure semantics for set
theory. T and F could be interpreted in whatever way we like; mathematically, we
need consider only mappings of formulae onto two distinct objects. Nor need we
interpret the Principia  as disjunction or anything else; it need only obey an
abstractly stated truth-table. This, then, is a pure semantics for two-valued logic;

and it can be extended, as Godel

showed, to a pure semantics for extensional
quantification theory.
But the pure semantics can be interpreted in several ways. One might interpret
the sentential variables as ranging over English sentences, and T and F as ascribing
truth and falsity, respectively, to the sentences. In this case, the truth-table for
the connective  would require that it be interpreted as disjunction, the English
inclusive truth-functional sense of or. One might just as well take propositional
variables as ranging over the same English sentences, but let T ascribe falsehood
and F truth. The truth-table for  then would require that this connective be
interpreted as conjunction, the English and. Principia would be interpreted as
asserting the self-contradictory character, not the truth, of its theorems.
These two interpretations of Principias sentential calculus are radically different,
yet they have a common mathematical structure revealed in the truth-table. The
two different applied semantical readings correspond to a single pure semantics;
and both the applied semantical systems are problematic in a way that the pure
semantics is not. For as Quines difficulties about indeterminacy show, the concept
of truth is not clearly well-defined for arbitrary sentences of an arbitrary natural
language; and it may not be translation invariant. Again, when sentences become
vague, the question of their truth becomes vague also; and in such cases as ethical
sentences, we often dont know whether to speak of their truth-value.6
The answers to these questions about truth are by no means clear, but this does
not mean that the pure semantics is unclear. Nor does it mean that there might
not be interesting applied interpretations of Principias sentential logic, rigorously

Quantified Modality and Essentialism

formulated, yet closely resembling in motivation the intuitive interpretation of T as

applicable to true English sentences. Indeed, Tarski has given us a method, applicable to a large class of formal systems, of defining the concepts of truth and falsity;
and the T and F may thus be interpreted as referring to truth and falsity, respectively, in a suitably formalized language. This applied semantics makes rigorous the
notions of truth and falsity which were more vaguely used in interpreting English.
Quines criticisms of modal logic have, as we have pointed out, taken several
distinct forms. Some of Quines published criticisms have seemed to indicate an
expectation that any semantical treatment of modal logic, pure or applied, would
fail, or else collapse modality. Such, for example, would seem to be the upshot of
the discussion of modal logic in Quine (1960: 198). For although Quine asserts
that his remarks are predicated on the interpretation of modality as analyticity
(an applied semantics), the actual argument in that section, leading to a postulate
that annihilates modal distinctions, would apply to any interpretation of necessity
satisfying certain simple formal conditions. (Indeed the same applies to the so-called
Morning-Star Paradox; it too can be stated formally on the basis of certain axioms
governing descriptions.)
Sometimes, on the other hand, Quine has maintained that quantified modal
logic, though perhaps admissible, requires an intensional ontology which rules out
individuals and classes; so Quine argues in the first edition of LPV.
Now, one of my primary purposes in Kripke (1959a) was implicitly to refute both
of these claims: I presented a rigorous pure semantics for modality, with no special
assumption about the character of the values of the variables. In the revised edition
of LPV, Quine indeed explicitly acknowledges that limiting the values of ones
variables to intensional entities is neither necessary nor sufficient for quantified
modal logic. There is no explicit acknowledgement that a pure semantics for a
quantified modal logic is indeed possible, but we shall interpret Quines silence
(in the revised LPV) on the point as a tacit acknowledgement of this fact also. (In
order that the remaining issues be clearly defined, an explicit acknowledgment from
Quine of this fact would be highly desirable.)
Although we interpret Quine as now at least tacitly acknowledging the possibility
of a pure semantics for modal logic, we can suppose that he fears that the process
of transforming it into an applied semantics may lead to special difficulties (at least
in the intended interpretation), largely related to essentialism. The number nine, for
example, is determinable by two conditions; it is the unique natural number x such
that,
1. x + x + x = x  x
where only natural numbers are in the equation, as well as the unique natural
number x such that,
2. There are exactly x planets.
Now (1), but not (2), must be taken as a necessary property of the number 9.7 For
if we wish to assert that,

NOUS

3. (x) N(8 < x < 10)

where x ranges over natural numbers and N is necessity, it is clear that if 9 is
that number it must be specified by a property like (1), not a property like (2).
For it is by no means necessary that the number of planets is between 8 and
10. Further, it appears that any quantified modal logic must accept a distinction
between the necessary and the contingent attributes of an object. Self-identity, for
example, is surely a necessary property of every object: if we are permitted at all
to apply necessity to an open sentence, surely (x) N(x = x) is true. But if p is any
contingent statement, x = x & p will be true, but contingently true; and hence x has
accidental, as well as essential properties.8 In Three Grades, Quine expresses
Aristotelian essentialism as saying that you can have open sentences Fx and Gx
such that,
4. (x)(NFx & Gx & NGx). (1976: 174. Notation slightly revised.)
So far, one cannot quarrel with Quines assertions; but in one way, they are
strangely formulated. Surely the objections to essentialism are not objections to the
truth of a sentence of the form (4), but rather to the very meaningfulness of essential
predication. Granted the intelligibility of this notion, it is not at all surprising that
self-identity is an essential attribute of every object, or that x = x & p is contingent
for contingent p. What is at issue is the question whether essential predication makes
sense; i.e. whether,
5. NFx
ought to be well-formed as an open sentence. For (5) asserts that F is an essential
attribute of the object x; and if this assertion is meaningful, it is by no means
surprising that it is sometimes true. Thus I find it odd that Quine writes, in LPV,
new edition, that Miss Barcans system hints at essentialism through her theorem
(x, y) (x = y  N(x = y))9 (LPV: 156). No theorems but simply the formation
rules are needed to assure us of the trivial fact that all quantified modal logics are
essentialist.
But this is a minor criticism. Let us return to the beginning. Quine, in both
editions of LPV, differentiates between essentialism and analyticity thus:
The Aristotelian notion of essence was the forerunner . . . of the modern notion of
intension or meaning. For Aristotle it was essential in men to be rational, accidental
to be two-legged. But there is an important difference between this attitude and the
doctrine of meaning. From the latter point of view it may indeed be conceded (if only
for the sake of argument) that rationality is involved in the meaning of the word man
while two-leggedness is not; but two-leggedness may at the same time be viewed as
involved in the meaning of biped while rationality is not. Thus from the point of
view of the doctrine of meaning it makes no sense to say of the actual individual, who
is at once a man and a biped, that his rationality is essential and his two-leggedness
accidental or vice-versa. Things had essences, for Aristotle, but only linguistic forms
have meanings. (LPV: 22)

Quantified Modality and Essentialism

As Quine points out in the revised LPV, the founders of modern modal logic
(C. I. Lewis, Carnap) intended to base their system on a notion of analyticity,
but instead they have based it on the much less clear notion of essentialism.10
Quine argues that although in the doctrine of meaning it makes sense to say that
qua the square of three nine is necessarily greater than seven, qua the number
of planets its relation to 7 appears to be accidental. Qua mathematician one is
necessarily rational but contingently two-legged; qua cyclist, the reverse. From the
point of view of the doctrine of meaning, there is no distinction possible qua object,
whether the man who is both cyclist and mathematician is essentially rational or
not. Thus essentialism is bafflingmore so than the modalities themselves. (Quine
1960: 109)
The necessary distinctions are made out carefully and judiciously in Quines
Three Grades of Modal Involvement. This paper, though written some years
ago, comes remarkably close to Quines present position; further, it is one of his
most careful statements of the issue. He distinguishes three different degrees of
acceptance of the notion of necessity. The first, or least degree of acceptance, is
this: necessity is a semantical predicate applicable to names of statements. Thus we
may say, abbreviating this predicate as Nec,
6. Nec (9 > 7)
7. Nec (Fermats Last Theorem)
If necessity is analyticity, (6) says that 9 > 7 is analytic, while (7) says that
Fermats last theorem is analytic. But we might take necessity as a statement
operatorthe second degree of modal involvement. To do so we prefix necessity, now symbolized by N, to statements, not their names, obtaining, instead of
(6) and (7), respectively:
8. N(9 >7)
9. N(x, y, z, n) (x > 0 & y > 0 & n > 2 & xn + yn = zn )
This permits, of course, iterated modal operators, since once we have attached a
modal operator to a statement we can attach it again:
10. NN(9 > 7)
11. NN(9 > 7)
But still, necessity operators are inapplicable to open sentences.
The third grade, as we might guess, applies necessity to sentences, open and
closed alike; thus it clearly involves one in essentialism. Now although in most
publications Quine has interpreted necessity as analyticity, and predicated his remarks on this interpretation, it is clear that the problem of essentialism can be
approached in a more general context, namely: what philosophical or logical differences are there among the various grades of modal involvement? What new
assumptions are required for the transition from the first grade to the third?

NOUS

Surprisingly enough, our answer will be: none. Anyone who accepts the first
grade must accept the third; hence, in particular, anyone who accepts analyticity
accepts essentialism also.11 The latter cannot be more untenable than the former.
And the founders of modal logic did not deceive themselves when they thought it
could be based on analyticity alone. The core of this paper will be devoted to the
detailed substantiation of this claim. The substantiation will not be philosophical,
but will be undertaken with mathematical rigor; so anyone who denies this denies a
mathematical result. In particular, to set up a sharp opposition between the theory
of essences and the theory of meaning is, in a sense, to deny a mathematical result.
The claim is thus very strong; that my argument is as convincing as, say, the proof
of the Cauchy integral theorem. However, after the argument has been presented,
there will be some qualification, showing that there is some room for discussion
left; and these remarks will be more nearly tentative.
To proceed with the argument: Suppose we have a predicate is necessary, applying to statements (of a language L already containing truth-functions) and hence
attaching to names of statements, such that
12. Every necessary statement is true.
13. If p and p  q are necessary, so is q.
14. Every truth-functional tautology is necessary.
If analyticity is definable at all, it satisfies these conditions. But notice that a
wide variety of semantical predicates satisfy them: logical truth, quantificational or
set-theoretic validity, theoremhood in a particular formal system, necessity relative
to certain premises, or even truth.
Now let us define a maximal consistent set of statements of L in the following
manner: It is to be a set H of statements of L, such that,
15. For every closed sentence p of L, either p or p is in H.
16. If p is necessary, p is not in H.
17. If p and p  q are in H, so is q.
This notion, except for the reference to necessity, is a familiar device of modern
extensional logic. Its modal connections have been insufficiently noticed; it corresponds to the modal notion of possible world.12 We now define an equivalence
relation between sentences (closed) of L. Two statements, p and q, are equivalent
if the biconditional p  q is necessary. We say two such sentences determine
the same proposition; we can then define a proposition as an equivalence class of
statements. Propositions, thus defined, are not intensional enough to be objects
of belief,13 since analytically equivalent sentences determine the same proposition;
but they are intensional enough for modal logic. Purists who object, reserving
propositions as entities intensional enough to be believed, may replace my usage of
proposition by modal value of a proposition. Now if p is a member of the set H,
intuitively, we will say that p is true in the possible world H. A possible world, then,
because of condition (15), must assign at least one truth-value to every statement;
because of conditions (13), (16) and (17), it assigns at most one. A possible world

Quantified Modality and Essentialism

is thus the totality of facts in that world. Thus equivalent statements, being true
in precisely the same worlds, determine the same class K of worlds H; namely, p
determines the set of worlds H such that p is true in H, and if p  q is necessary, q
is true in the same worlds as p. Conversely, if p and q express distinct propositions,
then p  q is not necessary, and a little mathematical detail, familiar to logicians
and too long to be expounded for others, suffices to show that in this case there is a
world H containing p but not q, or a world containing q but not p. So propositions
determine in a one-one fashion classes of worlds; but it by no means follows that
every class of worlds corresponds to a proposition, in the sense of proposition defined abovealthough, if it does correspond to a proposition, that proposition is
uniquely determined. So we could have defined proposition in a wider sense, thus:
A proposition is a class of worlds; or, equivalently, it is a function assigning to each
world a truth-value T or F. The propositions in the old sense will, in general, determine a subclass of the propositions in the new sense. In the new sense, whenever
we are given a determination of truth-conditions, we have a proposition; or more
exactly, when we are given a class K of worlds H, the class K can be interpreted
as a proposition asserting that one of the worlds H K is the real world. (N. B.:
The set G of all the true statements satisfies the conditions (15)(17); it can be
called the real world.) Classes of worlds which correspond to no equivalence class
of sentences might be said to be propositions inexpressible in the language L,
but propositions nevertheless; for it is fully determined in what situations (worlds)
they are true and in what situations they are false. In what follows, as long as
possible, we will try to show how our concepts could be defined using either notion
of proposition, the narrow one of an equivalence class of statements or the wider
one of a set of worlds.
Now it is clear that truth-functional statement composition determines corresponding truth-functional operations on propositions as well. Let us use capital
letters for propositions: P, Q, R, . . . Now if P and Q are any two propositions
(in the wider sense), let them correspond to classes K1 and K2 of worlds; then P
 Q is defined as the set-theoretic union K1
K2 ; that is to say, P  Q is to
be true in a world H if and only if either P or Q is true in H. If we look at a
proposition in the narrower sense as an equivalence class of sentences, so that p is
a sentence corresponding to the proposition P, and the same for q and Q, then we
can stipulate that P  Q is the equivalence class determined by p  q . We need
only verify that if p is equivalent to p , and q to q , then p  q is equivalent to p
 q ; and this verification is immediate. So in this manner all the truth-functions
can be defined.
Now let P be any proposition in the wider sense. We define N(P) to be the class K
of all worlds if P is the class of all worlds; otherwise N(P) is the empty set of worlds.
This definition can be rephrased, more intuitively, thus: If P is true in all possible
worlds, so is N(P); otherwise N(P) is false in all possible worlds. Alternatively, if
we were using the narrow sense of proposition, so that P is the equivalence class
determined by the sentence p, we could say: N(P) is the equivalence class of p
 p if p is necessary; otherwise N(P) is the equivalence class of p & p . It
is easy to verify that the definition is independent of the choice of statement p.

NOUS

(Further, it should be verified that the two definitions, the one for the narrow sense
and the one for the wider sense of proposition, coincide on the narrow sense; and
this should also be verified for disjunction above.) In either definition of N, N is
a unary function mapping propositions into propositions; hence it can be iterated
ad libitum, and combined with truth-functions at will.
Statements can be interpreted as denoting or meaning the propositions they
determine; and thus the necessity operator can be attached to statements: If p
determines the proposition P, we can use N(Np) to denote the equivalence class
N(NP).
Thus the transition from the first to the second grade of modal involvement
has been accomplished. In his paper on these grades, Quine remarks that it is
significant that in modal logic there has been some question as to just what might
most suitably be postulated regarding such iteration (169; note omitted). Here
Quine understates his case; this question has long been one of the most perplexing
and vaguest problems of modal logic. I have shown, in work announced in my
abstract, Semantical Analysis of Modal Logic (1959b), that the step from the
first to the second grade of modal involvement can be made in various ways,
leading to alternative systems of iterated modalities. In fact, the major systems
proposed in the literature can be obtained by some method at this stage.14 For our
present purposes it suffices to consider only one of these methods, the one used
above; I mention that it leads to S5.
We have a modal propositional logic; how can we pass to a modal system with
quantifiers? It is this step which Quine fears is fraught with danger and lurking
metaphysical assumptions. Let us be very careful, then, that we do not introduce
any new assumptions here; our assurance lies in the fact that, in this discussion we
shall continue doing pure mathematics, and will not make philosophical remarks
of any sort. The problem as stated above, is how to make sense of the locution,
5. NFx
Now, suppose we have a domain D of individuals, the values of the variable x of
(5). Before we attempt to interpret (5), let us first attempt to interpret the simpler
open sentence
18. Fx.
x is supposed to be a free variable ranging over the domain D.
What sense, however, can we make of the letter F? Classically, in the writings of
Russell, F was supposed to be a propositional function, but this terminology has
gone out of fashion. The term propositional function was ambiguous; sometimes
it meant attribute, sometimes open sentence or statement matrix.15 But the uses
of this notion in Principia make the proper interpretation clear; a propositional
function must be interpreted as an attribute, not a statement matrix; and clearly
attributes, whatever they may be, are within the province of modal logic. The F,
then, in Fx, ascribes an attribute to the individual x. But what is an attribute?16
The term propositional function gives us a clue long neglected in the literature. An
attribute is simply a mathematical function assigning a proposition to each element

Quantified Modality and Essentialism

of the domain D. Assume we have a particular propositional function F, and a

particular individual x; then by definition F assigns a proposition P to x. If P is
necessary (i.e., holds in all possible worlds), then we call F an essential attribute
of x; if P is necessary, F is excluded by x; otherwise F is a contingent attribute of
x. If Fx assigns P to x, NFx assigns NP; so, having made sense of (18), we have
also made sense of (5).
One modification, to be sure, forces itself upon us when we introduce the quantifiers. We have defined Fx; how are we to define (x)Fx? From the present point
of view (x)Fx can be defined as the (infinite) disjunction of all the propositions
Fa, Fb, Fc, . . . , where a, b, etc. are the elements of D; i.e., if we consider all the
propositions assigned to the elements of D by the propositional function F, (x)Fx
is to be their disjunction. If the domain is finite, say a1 , . . . , an , the notion of the
disjunction of all the propositions Fa1  . . . Fan is clear no matter which notion
of proposition we adopt; for finite disjunctions have already been defined for either
type of proposition. If D is infinite, however, we need an infinite disjunction; and
here is where the trouble begins. If we construe propositions in the wider sense,
as classes of possible worlds, no difficulty ensues; for then the infinite disjunction
of any set of propositions is simply the set-theoretic union of all the corresponding classes of worlds. The narrower version of proposition, however, according to
which propositions are just equivalence classes of statements, yields no corresponding procedure without special assumptions on the nature of the language L. We
might make these assumptions, but there is no need to do so; and we do not wish to
complicate the present paper any further. So henceforth, by proposition we shall
mean proposition in the wider sense; i.e., class of possible worlds.17
Existential quantification, then, comes to be explained as follows (if F is an
attribute or propositional function, i.e., a function assigning a class of worlds to
each member of D). Then, if D is the domain of this function, let X be its range; X
is a set of classes of possible worlds. The union of all the classes in this set is what
we mean by (x)Fx. Finally, in terms of this, let us explain the notation (x)NFx:
If F assigns a proposition P to an object a, then NF assigns NP to the object a.
Thus given an attribute F, we have defined an attribute G = NF expressing the
necessity of the attribute F. To say (x)NFx is to say (x)Gx; and the latter has
been defined above.18
So we have arrived at the third grade of modal involvement without recourse to
any philosophical notion beyond that of a semantical predicate satisfying (12)(14).
Just as the step from the first to the second grade is by no means unique, so the
step to the third grade is non-unique. But for the purposes of the present paper it
suffices simply to show that there is at least one way of making this step.
Let us give some examples, to make the intuitive content of the preceding definitions clear: Consider the question how to make sense of the statement:
3. (x) N(8 < x < 10)
This statement asserts that the attribute of being between 8 and 10 is essential to
some natural number. Now, how are we to define this attribute in terms of our
formalism? Let D be the domain of non-negative natural numbers. To say 8 < x

10

NOUS

< 10 is to say that 8 < x and x < 10. Let us analyze both of these. 8 < x, or Fx,
can be defined as the following proposition: The numbers 1, 2, . . . , 8 are to be
mapped into the contradictory proposition (the empty set of worlds); the numbers
9, 10, . . . are to be mapped into the necessary proposition (the set of all possible
worlds). Similarly we define x < 10 as necessary of 1, . . . , 9 and impossible of all
other natural numbers. To adopt a different definition is simply to define a different
attribute. Indeed, both of the attributes 8 < x and x < 10 are special cases of the
binary attribute x < y, which is necessary of certain number pairs and impossible
of others. When we assign the value 10 to y, we get the monadic attribute x < 10;
and similarly for 8 < x. If, in fact L already contains a notation for number theory,
the attribute < can be defined in a more intuitive way: < is to assign to the pair
of numbers m and n, the proposition determined by the sentence 0(m) < 0(n) , where
0(m) means 0 followed by m strokes. But it is noteworthy that the attribute < does
not depend on the presence in L of any particular notation at all.
I do not mean to claim that the method I have used to introduce quantified
modality is the only conceivable method of doing so, or the best, or that other
methods may not lead to peculiar philosophical problems. I only claim that I
have given one method of setting up quantified modal logic; and that this method
requires only the existence of a semantical predicate satisfying (12)(13). Does
this show that essentialism is after all not involved in quantified modal logic, that
essentialism can be defined in terms of analyticity, or that analyticity involves as bad
a metaphysics as essentialism? The option is unreal.19 In this paper, we have shown
the following simple fact: If analyticity is well defined, so is quantified modal logic.
And even if analyticity is not well defined, there are other necessity predicates,
such as theoremhood in a specified logical system. The modal systems built on
these predicates are interesting even if the classical applied semantics, in terms of
analyticity, is not.
Having made these remarks, I qualify them as follows: The procedure adopted,
though satisfactory for all purposes of quantified modal logic, is unsatisfactory if
one wishes to make a further demand. The language L may itself contain quantifiers,
ranging over a universe U; what is the relation between these quantifiers and the
quantifiers introduced above, over a domain D? The answer: there need be none.
It may even be necessary in L that U contains at most three elements; this by no
means prevents us, in our previous construction, from using a domain D with four
or more elements. The question then, arises: Can one introduce quantified modality
in such a way that the quantifiers introduced have a natural connection with the
quantifiers in L? A discussion of this (vaguely formulated) problem would prolong
this paper to an inordinate extent. Let me simply assert that in the analysis of this
problem, difficulties of essentialism very similar to those suggested by Quine do
indeed arise. But these difficulties are at most objections to one way of introducing
quantified modalities; they do not apply to the method sketched above.
Let us conclude with certain remarks on the intuitive problem of essentialism.
Basically the problem is as follows: What attributes must an individual, say Jones,
have in order to qualify as Jones? Which of his attributes can vary from possible

Quantified Modality and Essentialism

11

world to possible world, and which must remain? The former are contingent, the
latter necessary. Presumably, for example, he would still be Jones were he not a
mathematician; he would be Jones had he been born a day earlier; so both his
mathematical profession and his birthday are contingent attributes. But could we
imagine of Jones that he was not a human being but a planet? If not, being a
non-planet is essential to Jones, and we would count this attribute as essential.
The intuitive question of essentialism boils down to this: How does an individual
preserve his identity from one possible world to another? In the case of human
beings, we have only vague criteria of identity and identification, so the question is
vague (though not, I think, intuitively meaningless). In the case of natural numbers,
on the other hand, we use precise criteria of identity: the identity of a natural
number depends on its position in the natural number series, and nothing else.
Hence (1), but not (2), is intuitively an essential attribute of the number 9; for it
is (1), but not (2) that follows from 9s position in the series. That this account
of essentialism agrees with our intuitive notions is readily confirmed by the fact
that no one, intuitively, could ever take the reverse line of declaring (2), and not
(1), an essential attribute. If the world changed, the number of planets would be
different; but we would say, not that a particular natural number suddenly changed
its position in the series, but that there were more or fewer planets.20
In view of these facts, it is somewhat odd that Quine counts the distinction
between a fleeting and an enduring attribute as clearer than that of a necessary or
contingent one (Quine 1960: 199). For, as Prior has pointed out (and made this
point the subject of an entire book in Prior 1957), if we try to make a tense logic
for temporal (not eternal) sentences, the instants of time play a role formally quite
analogous to the possible worlds. A sentence which is always true counts as necessary. Again a problem of essentialism arises: how does an object preserve its identity
from one instant to another? Imagine, e.g., a building built with 10,000 bricks, and
then torn down brick by brick. When one brick is torn down, the building remains;
when 8000 are lost, it does not. For what n does possession of at least n bricks
become an enduring attribute of the building, one which remains a property of the
building as long as the building itself endures? Such questions can be set aside by
dealing only with objects defined independently of time, e.g., the four-dimensional
process-things, abstract entities, and the like; and similarly, one can set aside
questions of essentialism by introducing the domain of objects D independently
of the possible worlds. Essentially it was this expedient which was followed above,
when possible worlds were certain classes of statements of a language L. But the
doctrine that an essentialism of eternal attributes is intrinsically clearer than the
essentialism of modality, is one for which I can find little justification.21
I am well aware of the expository inadequacy of these final remarks, and of their
incompleteness, forced on us for reasons of space. Let us then recall the central
thesis of this paper: that quantified modal logic can be developed on the slender
basis of a single semantical necessity predicate. The development of this thesis,
prompted by Quines criticisms, may (so it is hoped), clarify the foundations of
modal logic. Thus in the same spirit that one might say that Berkeley, who by his

12

NOUS

searching criticism of the calculus prompted men more sympathetic to it than he to

clarify its foundations, made a substantial contribution to the calculus, so I think
it may be said that Quine is one of the principal contributors to modern modal
logic.22

Notes
1 There

is a long story behind this paper. It was written for a seminar given by Quine himself in
the academic year 1961/2, and discussed in class over a period of several weeks. Some years later, I was

surprised to hear from Hector-Neri Castaneda

that it had received wider circulation among philosophers.
I think I didnt even have a copy of the paper at the time, having handed my own to Quine for grading.

Castaneda
suggested I submit the paper to Nous and located the original versionwhich contains some
markings in the margins, presumably due to Quineat the Harvard Philosophy Department library. I
agreed to submit it to Nous, though I mentioned in a letter that it needed some corrections. The paper
was accepted for publication in 1966 and the idea was to include a reply by Quine as well, but this
never came about and the final version was not submitted (until now). Recently, some people at the
Saul Kripke Center, and also Graham Priest and Philip Bricker, read it and thought it should still be
published. So I resubmitted it to Nous and was told by the editor that it would be accepted if it hadnt
already been accepted!
I would put things somewhat differently now, though the technical result is, as I say in the paper,
impeccable. Some footnotes will reflect this (all footnotes, except footnote 18, are new). Unfortunately,
I dont have a copy of the letter mentioning the corrections I thought were needed then. But by 1966 I
had developed many of the philosophical views in Naming and Necessity, and would not think of this
as the primary way of interpreting essential predication. Maybe that is what I had in mind.
2 Presumably I meant the 1961 edition, current when I wrote this paper.
3 He adds: By implication, so much the worse for unquantified modal logic as well . . . I myself
assumed that modal logic is a proper subject only if quantified modal logic makes some sense. This isnt
really the atmosphere today where much writing emphasizes only the propositional systems, which have
application to computer science. See Goldblatts history of modal logic (2006: 2), which explicitly says
that he includes little about quantified modal logic.
4 I have described the construction in Kripke (1959a) in the same way elsewhere, but strictly speaking,
such a description is not correct. Possible worlds as such, do not appear as objects in the paper. But
each possible world determines a model of a given formula, in which it is true or false. Many possible
worlds may determine the same model of a given formula. The set K is the set of all models of a formula
corresponding to some possible worlds. (And, of course, need not include all models, contrary to a
misunderstanding I have seen.) A formula including modal operators is then defined as valid, if it is
true in all elements of any non-empty set K. (Actually, in the 1959a paper, I designate one element of
K as G, the one corresponding to the real world. Then I define validity as truth in any such G. This
doesnt matter for the purposes of the 1959a paper, since any G could be designated, but it is of some
importance elsewhere.)
5 Principia Mathematica already knew this semantics. See the interpretation of the calculus in
terms of 0 and 1 in the first paragraph of section *4 (p. 115). They appear, however, not to regard this
as the intended interpretation. (They presumably got this interpretation from the workers in Boolean
algebra.) They also recognized, following Frege, that their connectives are truth-functional.
The usual modern interest in the notion of truth-functional tautology (though really implicit in
the section of Principia Mathematica mentioned) stems from the work of Wittgenstein (1922) and Post
(1921). Post in particular proves that the theorems of the Principia propositional logic are precisely the
truth-functional tautologies, thus proving both the completeness and the decidability of the system.
6 I didnt really wish to commit myself here to Quines views about the indeterminacy of translation.
Also, as to vagueness, one possible view is that the truth predicate is vague as well. For ethical and other
statements of value, the problem of whether they have truth-values is well known and I did not make
any commitment about it.

Quantified Modality and Essentialism

13

7 As elsewhere, I have not changed Quines example of 9 as the number of planets, though astronomers may reject it today. (It seems to me to be of clear philosophical significance that the astronomers could change their minds about the number of planets, but mathematicians cannot change
their mind about what 9 is; though this has no significance for the present paper.)
8 Actually, an interpretation of quantified modal logic is in fact possible that would restrict necessary
properties to properties like self-identity that are obviously necessary independently of how an object is
described, and are therefore unsusceptible to Quines objections. This is not, of course, the interpretation
I have come to prefer. See Fine (1978), section II.
9 Actually, Barcan follows Principia Mathematica in defining identity as a second-order notion. In
Principia it is (F) (Fx  Fy), and she also considers a variant notion with the material conditional
replaced by a strict conditional. These notions are shown to be coextensive, or even strictly equivalent
given S4. To get the necessity of identity out of this, one needs to use in addition what has been called
the Barcan axiom.
I attribute the usual proof of the necessity of identity, taking identity as primitive, and which is much
simpler, to Quine himself in Kripke (1959a).
10 Or, at least, they need to do so if they intend modality to interact with quantification in the
usual way. I am not certain why I included Carnap with C. I. Lewis. Lewis introduced modal logic (or
really, originally, strict implication) into modern logic. I dont think he mentioned the combination with
quantification. (Barcan, Carnap, and Church published the first quantified systems.)
11 By essentialism I think I simply meant that modal operators can apply to open sentences, and
that quantifying in to such modally-operated open sentences is acceptable. No deeper philosophical
doctrine was involved.
12 However, Carnaps notion in Meaning and Necessity (see p. 9) of a state-description, though
not exactly the same as the one given here, is closely related. I should have noticed this fact.
13 In literature later than this paper, this assertion has not been universally accepted. There are
those who may try to get out of the problem, or mitigate it. See, for example, D. K. Lewis (1979: section
II, pp. 5145and ignoring here his later modification for de se attitudes). Here even propositions in
the broad sense of sets of worlds can be objects of arbitrary propositional attitudes, although Lewis
acknowledges that this leads to problems.
14 I am not sure what I had in mind here. My modeling of the systems of modal propositional logic
in terms of properties of a relation R between worlds has now been published (Kripke 1963 and 1965),
but I am not sure what I thought it had to do with the present construction.
15 In reading Russell this way, I was following Quine. Perhaps some others have also read propositional function as meaning open sentence. Today, I would not read Russell and Principia this way. A
Russellian proposition was a structured object (see Kripke 2005: 2378). A propositional function was a
potential proposition with letters (blanks) to be filled by objects of the appropriate types. When filled in,
it yielded a proposition. So it was a function of a special sort yielding propositions as values, but did not
mean an arbitrary function from objects to propositions. It did not mean an open sentence (though I
recall somewhere a loose passage in which it was called an expression, where taken literally it was not
a linguistic object at all). I do think that the concept of replacing objects in Russellian propositions was
probably somewhat obscure, regarded non-linguistically.
(A complication is that later Russell came to repudiate propositions in this sense, and accepted only
facts; but most of the interpretation of Principia goes in the old way.)
For the way I have been using the term propositional function, see also Quantified Modal Logic
and Quines Critique: Some Further Observations, fn 1.
16 Attribute here is another piece of Quinean terminology. I dont know if it is the best way of
describing propositional functions (see the previous note and the discussion of what Quines skepticism
about quantifying in might amount to if the theory is thought to be a theory of propositional
functions).
17 This usage of proposition, not just in the present construction, has become widespread.
18 The definition can clearly extend to polyadic predicates.
19 Here I was no doubt imitating Word and Object, p. 265 (see the remarks on physicalism and
solidity). Actually, in the technical sense that I was using essentialism, of course no special metaphysics
is involved.

14

NOUS

20 In this paragraph we can see how my views have changed in Naming and Necessity. In 1961/2,
when this paper was written, I had at least some of the intuitions about essential and accidental
properties of objects that I came to have later, though not all. However, I thought that what properties
were essential depended on criteria of identity [of individuals] across possible worlds. The idea that
such criteria are needed is of course, emphatically repudiated in the later work. I also, as the present
paper reads, did not already have in mind any fundamental distinction between a priori and necessary
truths, whereas this distinction is basic in Naming and Necessity. Nor did I distinguish these notions
from analyticity.
21 I would not view these matters the same way today either, though maybe the example of the
building may remain.
22 I would like to thank Gary Ostertag, Rosemary Twomey, and especially Romina Padro
for their
help in producing the present version. This paper has been completed with support from the Saul A.
Kripke Center at The City University of New York, Graduate Center.

References
Carnap, R. (1947). Meaning and Necessity. Chicago: University of Chicago Press.
Fine, K. (1978). Model Theory for Modal LogicPart II. The Elimination of De Re Modality. Journal
of Philosophical Logic, vol. 7: 125156.
Goldblatt, R. (2006). Mathematical Modal Logic: A View of Its Evolution. In Handbook of the History
of Logic Volume 7, Gabbey, D. and Woods, J (eds.), Amsterdam: North Holland: 198. (Check)
Kripke, S. (1959a). A Completeness Theorem in Modal Logic. Journal of Symbolic Logic 24: 114.
Kripke, S. (1959b). Semantical Analysis of Modal Logic. (Abstract) Journal of Symbolic Logic 24:
323324.
Kripke, S. (1963). Semantical Analysis of Modal Logic I. Normal Propositional Calculi. Zeitschrift fur
mathematische Logik und Grundlagen der Mathematik 9: 6796.
Kripke, S. (1965). Semantical Analysis of Modal Logic II. Non-Normal Modal Propositional Calculi.
In The Theory of Models, Addison, J. W., et al. (eds.), Amsterdam: North Holland: 206220.
Kripke, S. (2005). Russells Notion of Scope. Mind 114: 10051037. Reprinted in Kripke (2011): 225253.
Kripke, S. (2011). Philosophical Troubles. Collected Papers Volume I. New York: Oxford University Press.
Lewis, D. (1979). Attitudes De Dicto and De Se. Philosophical Review 88: 513543. Reprinted in Lewis
(1983).
Lewis, D. (1983). Philosophical Papers, Volume I. New York: Oxford University Press.
Post, E. L. (1921). Introduction to a General Theory of Elementary Propositions. American Journal of
Mathematics 43: 163180. Reprinted in van Heijenoort (1967): 264283.
Prior, A. (1957). Time and Modality. Oxford: Oxford University Press.
Quine, W.V.O. (1953). From a Logical Point of View. Cambridge: Harvard University Press. Second
Edition, 1961.
Quine, W.V.O. (1960). Word and Object. Cambridge: MIT Press.
Quine, W.V.O. (1976). Three Grades of Modal Involvement. In The Ways of Paradox, Second Edition.
Cambridge: Harvard University Press, pp. 158176. Originally published in Proceedings of the
XIth International Congress of Philosophy, vol. 1, 1953.
van Heijenoort, J. (ed.) (1967). From Frege to Godel: A Source Book in Mathematical Logic, 18791931.
Cambridge: Harvard University Press.
Whitehead, A. N., and B. Russell. (1910, 1912, 1913). Principia Mathematica, 3 Vols, Cambridge:
Cambridge University Press. Second edition, 1925 (Vol. 1), 1927 (Vols. 2, 3).
Wittgenstein, L. (1922). Tractatus Logico-Philosophicus, C.K. Ogden (trans.). London: Kegan Paul.