Russell's "tautologies" are theorems in extensional logic, specifically, first order propositional calculus or first order predicate calculus; but Marcus' tautologous statements are theorems in intensional logic, specifically, quantified modal logic (quantified S4). To see how she uses Russellian or "early analytic philosophy" terminology to express her novel ideas about modality, let us trace her use of the relevant words in her pivotal essay "Modalities in Intensional Languages" [1961]. Marcus also uses the word "analytic", sometimes interchangeably with "tautologous", much as A.J. Ayer did in his Language, Truth and Logic, but we shall see that Marcus' modal use of these words differs fundamentally from Ayer's and Russell's use of these words.

In her [1961], Marcus characterizes certain classes of statements, e.g., identity statements involving names, as "tautologies", "analytic truths", "logical truths", "logically necessary truths", "necessary truths" and "not contingent truths". But we would be mistaken if we interpreted her use of these words to express the nonmodal concepts that Ayer and Russell used them to express. Since her paper is about quantified modal logic and its application to natural or ordinary language, we know that by a "logical truth" or a "logically necessary truth" she is talking about "logic" in the sense relevant to quantified modal logic (e.g., her semantic construction on pp. 319ff. is in terms of quantified S4). Thus, she begins her discussion of identity statements by saying [1961: 307]:

In the light of the previous remarks I would like to turn specifically to the criticisms raised against extended modal systems in connection with identity and substitution. In particular, I will refer to my extension of Lewis' S4 which consisted of introducing quantification in the usual manner and the addition of the axiom (6) à ($x) A ® ($x) àA. I will call this system QS4.

She takes identity statements such as "Venus is Phosphorus" to be logical truths (in the sense pertinent to QS4). In her discussion of theorems of QS4 and its "logical truths" on pages 307-308, she uses "tautology" to mean a logical truth; e.g., she says the fact that "aIb" is a "tautology" "is precisely the import of my theorem (8)" [1961: 308], where theorem (8) is a theorem of QS4 stating:

"(8) (xIy) o(xIy) where 'o' is the modal symbol for logical necessity." [1961: 308]

Marcus also says "aIb" or "Venus is Hesperus" an "analytic" truth; she writes: "What I have been arguing in the past several minutes is, that to say of an identity (in the strongest sense of the word) that it is true, it must be tautologically or analytically true". Marcus is using "analytically true" in an hitherto unfamiliar sense of this phrase. The word "analytic" is standardly used in many different senses, sometimes in an epistemic sense to be knowable a priori, by reflection on meaning alone, sometimes in a linguistic sense to mean a sentence that is transformable into a logically true sentence by substitution of synonyms for synonyms, sometimes in a conceptual sense to mean a thought where one concept contains another concept as a part, sometimes in a definitional sense, to mean derivable from logic together with definitions, and sometimes in a logical sense to mean a logically necessary truth. In what sense does Marcus use it? The preceding quotations suggest that "analytic" means tautological and that tautological means (or is a species of) logically necessity (in the sense of QS4). Is there any further evidence that she means "analytic" in the sense of (modal) logical necessity?

In the subsequent discussion of her paper [Marcus, et al. 1962: 140, my emphases] she distinguishes between a narrow and broad sense of "analytic": "Necessary attributes would correspond to analytic functions in the broader sense of analytic. These may be thought of as a kind of essential attribute, although necessary attribute is better here. For these are attributes which belong necessarily to every object in the domain, whereas the usual meaning of essentialism is more restricted. Attributes like mathematician and cyclist do not correspond to analytic functions."

The broader sense of "analytic" is used in a modal logical sense, rather than a definitional, epistemic, linguistic or conceptual sense; an broadly analytic function corresponds to an attribute that belongs necessarily to every object in the domain. This broad sense of "analytic" has metaphysical import, for it is defined in terms of every object in the domain. "Domain" is used here in the sense of her "semantic construction" in her [1961], where all possible worlds are said to include only objects that belong to the actual world. She has a constant domain semantics: the "domain of objects" is all actually existing objects. A broadly analytic function corresponds to an attribute that belongs necessarily to every object in every possible world.

This use of traditional terminology to introduce new ideas shows why it is easy for superficial readers to misunderstand Marcus' theory, and may explain in part why her novel ideas were mistakenly attributed to later thinkers (at least up to the mid-1990s, when her originality became very widely recognized). But note that some, such as Kaplan [1986], Almog [1986] and Devitt [1989], correctly understood her [1961] before the widespread recognition came in the mid-1990s. (In terms of widespread recognition, I am talking about her contributions to the philosophy of language and metaphysics; she has had widespread recognition for her contributions to logic since the late 1940s and also widespread recognition for other contributions, e.g. to the literature on moral dilemmas.) In the case of "analytic", Marcus is trying to express a new idea by adding "broader" to a term that is traditionally used in other senses, just as Plantinga [1970] adds "broader" to the traditionally used "logically necessary" to express his conception of metaphysical necessity, in the sense pertinent to non-trivial essences, such as being prime, being a number, and world-indexed properties such as being snub-nosed in Kronos, where "Kronos" is a name for the actual world. Plantinga pointed out in an October 1969 talk at Cornell University, published as [1970] that the concept of a broadly logical necessity belongs to metaphysics, not epistemology, and argues that there are logical necessities that are known a posteriori. This idea, also evident in Marcus [1961; 1962], was popularized by Kripke [1971; 1972] and thereby entered the philosophical mainstream. (For an extensive discussion of this idea, see my essays #1, #3, #7 and #10 in [Humphreys and Fetzer, 1998].)

The modal-metaphysical import of Marcus' use of "analytic", "tautological" and "logically necessary" is also apparent from her definition of a logically true sentence. She defines logical truth in terms of her semantic construction, which she says "corresponds to the Leibnizian distinction between true in a possible world and true in all possible worlds". As some other modal logicians do at this time, she uses "possible world" interchangeably with "model", so that "every model" is synonymous with "every possible world". She defines a logically true sentence as follows: "A logically true sentence is one which would be true in every model." [1961: 319]. Notice that she does not say that a logically true sentence is one which is (a) true by virtue of its form alone, or (b) is self-evident or (c) is a priori or (d) is knowable by reflection on its meaning alone or (e) is a sentence whose denial is self-contradictory or (f) is true by linguistic convention, (g) is provable by truth tables, etc., etc.

She also talks about counterfactual situations in her discussions of possibility and necessity, which is another indication of her modal-metaphysical understanding of these notions. In the following, Marcus is talking about the criteria for deciding whether "evening star/morning star" and "Scott/author of Waverley" are being used as names or contingent descriptions: " . . . if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g. someone else might have written Waverley, the star first seen in the evening might have been different from the star first seen in the morning--then they are not identities." [1961: 311, my emphases].

It is obvious that Marcus does not have a linguistic or epistemic concept of "necessary"; for Marcus "necessary" does not mean "discoverable by reflection on linguistic meaning alone". If it did mean this, necessarily true identity statements involving names are identity statements whose truth can be determined simply by reflecting on their meaning. But this cannot be what Marcus meant; indeed, she denied that names have meaning, let alone have meanings that can be grasped by reflection alone. She writes: "This tag, a proper name, has no meaning. It simply tags." [Marcus, 1961: 310]. In order for the the sentence "Hesperus is Phosphorus" to be known to be true by reflection on its linguistic meaning alone, "Hesperus" and "Phosphorus" would have to express concepts or descriptive senses, or some other sort of meanings, grasped by reflection. But with Marcus' theory, we have no such concepts or "graspable meanings" and we have merely meaningless syntactical entities (the tags), the identity sign, and astronomical bodies millions of miles away--there is nothing here to be reflectively grasped by contemplating the sentence or the meaning of the names in the sentence. Here we see the first rejection of the Russellian theory of ordinary proper names that was still prevalent in the 1950s and (in many quarters) in the 1960s, namely, the ideas that names are disguised or abbreviated descriptions. Marcus rejects this idea and instead says that ordinary proper names are nondescriptive and directly referential. This is one of the decisive moves from the Russellianism of the first half of the 20th century to the Marcusianism that dominated the second half of the 20th century.

Marcus was the first person to make the distinctions mentioned, as is evinced by the fact that she distinguished necessary attributes (in the essentialist sense) from tautological attributes (which are narrowly logical modalities). John Burgess mistakenly writes that Marcus "uses the following expressions interchangeably in the quoted passages and through her talk:

( ) is necessary

( ) is tautologically true

( ) is analytically true

it is not an empirical fact that ( )". (See Humphreys and Fetzer, 1998: 109].)

In contradistinction to this claim, Marcus emphasizes that tautologies are but one species or type of necessities, the other type belong to essential attributes, such as being either a mathematician or not rational. She writes:

"If r is some abstract then we can define

(57) x Î o r = _df o (x Î r), _ o r = _df (x)((x Î o r)

and

(58) x Î à r = _df à (x Î r), _ à r = _df (x)((x Î à r)

It is clear that among these abstracts to which _ o may be validly be affixed, will be those corresponding to tautological functions, e.g., y(yIy), y(_x v _x), etc. It would be appropriate to call these necessary attributes, and the symbol 'o' is a derivative way of applying modalities to attributes." Similarity, all of the attributes of (50) through (54) could in the sense of (58) be called contingent, where 'à' is the derivative modality for contingency of attributes. However, if (50) is true, then the attribute of being either a mathematician or not rational could appropriately be called necessary, for

(59) (x) o (x Î y (Mx Ú ~ Rx))." (p. 318-319. My emphasis.) Item (50) is (where M = being a mathematical and R = being rational):

(50) (x) (Mx ® Rx) º (x) o (Mx É Rx) º (x) ~ à (Mx × ~ Rx).

Several comments need to be made to explain this crucial and rarely understood passage from Marcus [1961]. To begin with, the first definition in (57) translates as "x exemplifies the necessary attribute r = _df necessarily, x exemplifies the attribute r". This definition allows for non-trivial essential attributions as well as logically necessary attributions. For example, it allows "necessarily, Cantor exemplifies self-identity" and "necessarily, Cantor exemplifies being a mathematician". Self-identity is a logically necessary attribute and being a mathematician is a non-trivial essence of Cantor.

The second definition in (57) corresponds only to logically necessary attributes; it translates as "r is a logically necessary attribute r = _df for every x, x exemplifies the necessary attribute r". No non-trivial essential attribute is exemplified by everything, since non-trivial essential attributes sort individuals into kinds or distinguishes between individuals of the same kind. Using Marcus' example of a non-trivial essence (being a mathematician), "for every x, x exemplifies the necessary attribute being a mathematician" is not a true substitution instance of the second definition in (57).

With this translations of (57) fresh in our mind, we can understand the clause "It is clear that among these abstracts to which _ o may validly be affixed, will be those corresponding to tautological functions". Marcus say "among these abstracts" rather than "all these abstracts" since she is talking about the second definition in (57), rather than the first definition, and the second definition allows both logically necessary attributes (which correspond to tautological functions) and non-trivial essential attributes (which do not correspond to tautological functions). The genus, necessary attribute, contains two distinct species, those corresponding to tautological functions and those not corresponding to tautological functions. Pace Burgess and Soames, being necessary is not the same thing as being tautological or being logically necessary.

Regarding this passage, one of her points is that nontautological attributes, such as being a mathematician or not rational, are necessary in a nontautological sense if we add to the modal logic system QS4 a thesis that is not provable in QS4, namely, thesis (50) stated above. More exactly, and quoting Marcus, ". . . if we introduced constants like 'cyclist', 'mathematician', etc., and appropriate meaning postulates then the attribute of being either a non-mathematician or rational, would also be necessary. Necessary would correspond to analytic functions in the broader sense of analytic [not in the narrow sense of the analytic functions that modal logic "countenances as being necessary" [1961: 318]]. These may be thought of as a kind of essential attribute . . ."[Marcus, 1962: 140]

However, there are other sorts of essential attributes, such as being a mathematician, that are necessary but not in the modal-logical sense of "necessary" (in the modal-logical sense of "necessary", it would be a theorem of a modal system that the attribute necessarily belongs to an object). Marcus writes "I know of no interpreted modal system, even where extended to include predicate constants such as your example [item #50], where properties like being a mathematician would necessarily belong to any object. The kind of uses to which logical modalities are put have nothing to do with essential properties in the old ontological sense [which are merely ontological necessary, not logically necessary]. The introduction of physical modalities [such as physical necessity] would bring us closer to this sort of essentialism" [Marcus, 1962, pp. 140-141.

Thus, we have tautological or modal-logical necessities, such as being self-identical, and we have two sorts of nontautological necessities, the "broad analytic necessities', such as being either a non-mathematician or rational, and the "ontological necessities", such as Cantor's being a mathematician. Thus, not only does Marcus use "tautological" in a different sense than Russell, but she distinguishes two sorts of nontautological necessities that belong to modal metaphysics but not to Russellian metaphysics, namely, broadly analytic necessities and ontological necessities.

If we are to develop a semantic construction for modal logic, e.g., QS4, then we are dealing with logical necessity and "all possible worlds" means all logically possible worlds. Marcus's semantic construction in her 1961 paper is for QS4. If we wish to talk about the ontological necessity of non-trivial essences, then we talk about a sub-class of the class of all logically possible worlds, namely, the sub-class of all ontologically or metaphysically possible worlds. There has been much unclarity about this distinction in the literature. The main point that has not been clearly discussed in the literature on rigid designation is that if the class of all possible worlds that is relevant to rigid designators is the class of all ontologically possible worlds, then the class of possible worlds relevant to rigid designators is not the class that is relevant to the semantics of modal logic. For example, "Plato" does not (in natural language) refer to possible worlds in which Plato is a window pane, since this is a logically, but not ontologically, possible world. In all the ontologically possible worlds in which Plato exists, he possesses the non-trivial essence of being rational and being human, and the natural language name or tag "Plato" rigidly refers only to Plato as he exists in these ontologically possible worlds. The ontologically possible worlds are a proper subset of the set of all possible worlds countenanced in the semantics for a modal logic system such as QS4 or QS5. (For more on this, see [Humphreys and Fetzer, 1998: 257-271].)

If we examine Russell's extensional notion of necessity, we will see an even more dramatic difference between Russell's theory of necessity and Marcus' intensional, modal notions of necessity. By "is necessarily true". Russell means the extensional, temporal notion, is always true; for Russell, "necessity" means what is always true in the actual world, not the modal notion of what is true in all possible worlds (e.g., see Russell, 1918]. But Marcus does not use "necessary truth" in an extensional, temporal sense but in the metaphysical-modal sense to mean truth in all possible worlds. Russell says that "one may call a propositional function necessary, when it is always true" [1918: 231, my emphasis], but Marcus rejects this temporal definition of necessity and accepts a possible world definition. As I indicated above, a necessarily true sentence is one that is "true in all possible worlds" [1961: 320]; she uses "model" as a synonym for "world" and writes about logically possible worlds (as distinct from ontologically possible worlds): "A logically true sentence is one which would be true in every model". [1961: 319]

But Marcus did not claim that all necessities are de dicto. Indeed, she was the first to introduce into 20th century analytic metaphysics the idea that there are de re necessities or necessarily possessed properties. Marcus defines an individual's logically necessary possession of a property, not in terms of it being true a priori or always true that the individual possesses the property, but in terms of it being true that the individual possesses the property in every possible world (model). "To say of a thing a that it necessarily has a property _ (o(_a)) is to say that _a is true in every model. Self-identity would be such a property." [Marcus et. al., 1962: 133]. Of course, a merely essential or ontologically necessary property is not possessed in every logical model, but only in all the models that are ontologically possible.

Another distinction that Marcus originally introduced into the post-Russellian or Marcusian philosophy that dominated the second half of the 20th century is the distinction between necessary identifications and non-trivial essential attributions. "aIb" (e.g., "Tully is identical with Cicero") is an identification, whereas, a Î o r may make a non-trivial essential attribution (in which case it attributes to the individual a the ontologically necessary property r.).

This concludes my account of the different senses that Russell and Marcus attached to the words "tautology", "analytic" and "necessary".

2b. Identity and Names in Russell and Marcus

Let us proceed now to a further main difference between Russell's and Marcus' analysis of sentences such as "Scott is the author of Waverley" and "Scott is Scott". This concerns the very different senses in which they used the word "identity". Marcus uses "identity" to express a relation different than strict equivalence, material equivalence, and other weaker sorts of equivalence relations, but Russell did not use "identity" to express this strong relation. For Russell, the identity sign can flanked by a name and a definite description, but a central idea in Marcus' philosophy is that the identity sign can be flanked only by names. Russell used the word "identity" to express what Marcus calls an equivalence relation weaker than identity. Furthermore, for Russell, identities can be contingent, but for Marcus, identities are always necessary. Russell writes: "That illustrates how 'the author of Waverley' is quite a different thing from a name. You can prove this point very clearly by formal arguments. In 'Scott is the author of Waverley, the 'is' of course expresses identity, i.e., the entity whose name is Scott is identical with the author of Waverley" [1918: 245]. Also see Russell [1918: 247]. Here Russell is using "Scott" as an artificial example of a name, since in ordinary usage (according to Russell) it is an abbreviated description and in reality names are demonstratives such as "this" and "that". But Marcus denied that the "is" in "Scott is the author of Waverley" expresses the identity relation and she denies that identity statements can be contingently true.

Let us now turn from metaphysics to the philosophy of language and note a major distinction between Russellian and Marcusian philosophy of language. One crucial difference was briefly mentioned above; Russellian logically proper names are names in a private language for my fleeting sense data with which I have an immediate acquaintance. "The only words one does use as names in the logical sense are words like 'this' or 'that'. One can use 'this' as a name to stand for a particular with which one is acquainted at the moment. . . . [A proper name] seldom means the same thing two moments running and does not mean the same thing to the speaker and to the hearer. " [1918: 201]. Indeed, logically proper names are necessarily private; in a logically proper language "all the names that it would use would be private to the speaker and could not enter in the language of another speaker" [1918: 198]. Russell's names (i) are in a private language, (ii) refer only to a person's introspected sense data, (iii) rarely mean the same thing in two successive moments, (iv) are demonstrative words, such as "this" and "that", (v) are directly referential in the sense that the speaker is immediately acquainted their referents and (vi) are not intersubstitutable in metaphysical modal contexts, since for Russell there are no such contexts (for Russell, contexts involving "necessarily", "possibly" and the like are temporal contexts). But Marcus' names (i) are in a public language, (ii) refer to ordinary things in the universe, such as Venus or Walter Scott, (iii) usually have the same reference across long periods of time, (iv) are not demonstratives such as "this" and "that", but are what are ordinarily called "names", e.g., "Scott", "Venus", etc., (v) are directly referential regardless of whether or not the speaker is immediately acquainted with their referents and (vi) are intersubstitutable in (metaphysical) modal contexts.

In summary, Russell's theory that "a true identity statement involving names is necessarily true" is the theory that the true private language statement about my sense data, "this is this", is always true and expresses a kind of identity relation that can obtain contingently and (in the case of names) obtains between two items with which I am immediately acquainted. Marcus' theory that "a true identity statement involving names is necessarily true" is the theory that a true public language statement about publically observable things in the universe, such as "Venus is Hesperus", is true in every possible world, where "is" expresses her strong relation I of identity (which can never obtain contingently), and which often obtains between items with which I am not immediately acquainted, and where the statement is a "tautology" in the sense of being a theorem of quantified S4 (rather than of the extensional logic of Whitehead's and Russell's Principia Mathematica).

3. The Transition from Russellianism to Marcusism

An astonishing feature of Marcus' [1961] is its remarkable originality and novelty; her theory appears almost ex nihilo, rather than as a development of a growing tradition. Whereas Russell was decisively influenced by G. E. Moore (from 1903 to 1912) and later by Wittgenstein (1913-1919), there is no such influence on Marcus. The nearest thing to her [1961] is her [1960] and one can view her [1960] as a development of a line of thought that began with C.I. Lewis, extended through Quine, Carnap, Smullyan and Fitch. But her [1961] is a radical advance from her [1960] and does not belong to the developing line of thought from C.I. Lewis through Smullyan and Fitch. Let us consider this historical background in some detail.

Marcus' main influence in the philosophy of language is her development of the theory of nondescriptivist, rigid reference, or what is sometimes called "The New Theory of Reference". This theory is about natural language, but an account of its origins requires us to discuss theories of artificial, logical languages, specifically, the languages used in interpretations of quantified modal logic.

A syntactics for propositional modal logic first appeared in C.I. Lewis's [1918] and a syntactics for quantified modal logic first appeared in 1946 with Marcus's [1946]. An interpretation or semantics for modal logic first appeared in Carnap's [1946]. (But note that Carnap was not the first philosopher to see that the semantics for modal logic could be developed in terms of Leibnizian possible worlds. C.I. Lewis recognized this in his 1943 article on "The Modes of Meaning"; for example, he writes "An analytic proposition is one which would apply to or be true of every possible world" [1943-44: 236-249]. But Lewis did not develop the semantics; Carnap did this in his [1946] and [1947].) The key idea of Carnap's semantics is that the interpretation of "o A" was based on the Leibnizian idea that necessary truth is truth in all possible worlds. Contrary to what I (mistakenly) said in chapter 10 of [Humphreys and Fetzer, 1998], Carnap's possible worlds are not state descriptions. Rather they are possible states of affairs represented by state descriptions. Thus, Carnap writes: "A state-description is a class of sentences which represents a possible specific state of affairs . .[1946: 50]. For Carnap, "o A" means A obtains in all possible worlds (or that the sentence "A" is true in all state descriptions). A state description is a maximally consistent set of atomic sentences. It is a logically consistent set W of sentences, such that for each atomic sentence S, either S is a member of the set W or the negation of S is a member of the set W.

The route from logical languages to natural language followed a path started by Quine in 1941, in a footnote to an article he contributed to Schlipp's The Philosophy of Whitehead. Here Quine first stated his famous "paradox" about quantified modal logic. It is not widely known that Quine first stated this paradox in 1941, four years before modal logic was first quantified by Marcus. This paradox is often traced back to a later writing of Quine, his [1947] or sometimes his [1943]. For example, Soames makes this frequent mistake in historical accounts of analytic philosophy: Soames says his "historical survey begins with Quine's 1943 paper, 'Notes on Existence and Necessity'. It was there that Quine first [sic] set out his argument that ordinary objectual quantification into any opaque construction is unintelligible, and hence illegitimate." [Humphreys and Fetzer, 1998: 14]. In fact, the paradox as it first appears, in Quine's [1941: 142:, n. 26], comprises two sentences of footnote 26:

"C.I. Lewis and C.H. Langford (Symbolic Logic, New York, 1932), e.g., use a non-truth-functional operator 'à ' to express logical possibility. Thus the statements:

à (the number of planets in the solar system < 7)

à (9 < 7)

would be judged as true and false respectively, despite the fact that they are interconvertible by interchanging the terms '9' and 'the number of planets in the solar system', both of which designate the same object."

If it were not for this footnote (which led to Quine's later restatements of the paradox) and the responses to this paradox, it is doubtful that the path leading to the theory of nondescriptive, rigid names and the associated Marcusian ideas would ever have been developed in its extant form.

The traditional form of the paradox in virtually all later discussions is taken from Quine's [1947: 47] formulation, where "C" means congruence (which may be the relation of identity, but Quine does not wish to prejudge that):

"Morning Star C Evening Star and o (Morning Star C Morning Star).

Therefore, according to (ii) ['An existential quantification holds if there is a constant whose substitution for the variable of quantification would render the matrix true']:

(1) ($x) (x C Evening Star and o (x C Morning Star)).

But also:

Evening Star C Evening Star and ~o (Evening Star C Morning Star),

so that, by (ii),

(2) ($x) (x C Evening Star and ~o (x C Morning Star))."

Since (1) and (2) are mutual contraries, we have a paradox. Before we consider the first relevant response, Smullyan's in 1947, one other idea needs to be introduced which proved crucial for the Marcusian philosophy that emerged in 1961 and became predominant later. In his [1943] Quine discusses but rejects a principle of universal intersubstitutivity: "Given a true statement of identity, one of its two terms may be substituted for the other in any true statement and the result will be true." [1943: 113]. This statement may appear to be natural or plausible, given the indiscernability of identicals. If two individuals x and y are identical, they are indiscernible; this may suggest the idea that singular terms for x and y should be intersubstitutable in any context, including modal, temporal, deontic, epistemic and other intensional contexts. But Quine, under the influence of Russellian or preMarcusian philosophy, argues that singular terms are not intersubstitutable in modal and epistemic contexts.

The idea that singular terms, most notably names, are always "purely designative" was associated with this universal intersubstitutivity thesis. Quine also rejects this Marcusian theory of names: "Failure of substitutivity [in modal and epistemic contexts] reveals merely that the occurrence to be supplanted is not purely designative, and that the statement depends not only upon the object but on the form of the name" [1943: 114]. Quine holds the Russellian idea that ordinary names (as distinct from logically proper names) have a meaning ("criterion of application") in addition to their designative feature, and that the meaning, not just the designation, is relevant to modal and epistemic contexts. By rejecting the Marcusian universal intersubstitutivity thesis and the associated Marcusian thesis that names do not have a meaning but instead are purely designative, Quine has to mention these theses and make an issue about whether they are plausible, and his discussion of these theses was a motivating factor in the future discussions that led to their full flowering in Marcus' [1961]. (And in regards to the universal substitutivity thesis, it was more fully developed in Marcus [1975], where she showed Quine's and others criticisms of it are invalid.)

But we ought to recall at this juncture the formulation or conception of a universal intersubstitutivity thesis is not in any sense "original" with Quine. This thesis has been regularly propounded or discussed throughout the history of philosophy, and can be traced at least as far back as Aristotle's Topics vii, I. The point is that Quine drew attention to it in a way that proved influential in the development of post-Russellian philosophy.

The philosophy of language Quine rejected (or rather, a theory relevantly similar to it) is also mentioned and rejected four years later, in Carnap's 1947 book, Meaning and Necessity. Carnap calls the theory "the name-relation method". Carnap writes [1947: 98] that the name-relation method is based on three principles (the numbers in brackets are my insertions) and is meant to apply to an artificial, logical language:

[1] Every expression used as a name (in a certain context) is the name of exactly one entity; we call it the nominatum of the expression. . . . [2] A sentence is about (deals with, includes in its subject matter) the nominata of the names occurring in it. [3] The principle of interchangeability (or substitutivity). This principle occurs in either of two forms: [3a] If two expressions name the same entity . . . the two expressions are interchangeable (everywhere). [3b] If an identity sentence ['... is identical with--- '] is true, then the two argument expressions '... ' and '---' are interchangeable (everywhere).

Carnap considers this to be a method for discussing an artificial or ideal language and he believes this method is "more customary" [1947:96] than his own method of extension and intension. Carnap's undertakes an extensive discussion of this name-relation method and rejects this method, mainly for the reason that he believed substitutions in modal contexts did not preserve truth value, a claim that would be clearly refuted by Marcus in her [1961]. Further, Carnap had no sympathy for directly referential names and believed names referred to their intensions in modal contexts, and thus were not rigid designators (in that their reference changes from extensional to modal contexts). Marcus [1961] would also decisively undermine this Carnapian view.

The second relevant event in 1947 is Smullyan's innovative use of the "purely designative" and "substitutivity" theses in modal contexts to respond to Quine's paradox. Smullyan is responding to Quine's [1947], in which Quine's point is that "when modal logic is extended (as by Miss Barcan) to include quantification theory. . . serious obstacles to interpretation are encountered." [Quine, 1947: 43-48]. Smullyan suggests two ways for interpreting quantified modal logic, each of which is able to resolve Quine's paradox; but only the first way is pertinent to the nondescriptivist and rigidity theory of Marcusian philosophies. This solution appears in the third sentence in this passage:

It is possible that by 'constant' is meant what is commonly understood by 'proper name'. Under this interpretation it appears evident to this reviewer that the principle of existential generalization is true. However, we observe that if 'Evening Star' and 'Morning Star' proper-name the same individual they are synonymous and therefore B is false. [Smullyan, 1947: 140]

Statement B is the premise: "Evening Star is congruent with Evening Star and ~o Evening Star is congruent with Morning Star." In the above-quoted phrase, "what is commonly understood by 'proper name'", Smullyan is not referring to ordinary proper names, but to what philosophers of logic of that time commonly understand as proper names, i.e., proper names in Russell's sense (Russell's logically proper names, "this" and "that", which directly refer to my sense data). Smullyan characterizes his response to Quine's charge of contradiction as a response that employs Russell's theory of descriptions and proper names: "We have seen that in terms of Russell's theory of descriptions and proper names this contradiction can be avoided." [1947: 141; my italics]. Smullyan's employment of Russell's theory of logically proper names (as distinct from Mill's theory of ordinary proper names) in his first response to Quine was later noted by Prior [1967: 10-11] and others.

Smullyan's second and preferred way of responding to Quine, in terms of Russell's theory of descriptions, is that the necessity of identity thesis for individuals is false, which is an implication of the following sentence, the crux of his second response: "For if it is not necessary that the morning star exists then it is not necessary that the morning star is self-congruent" [1947: 140]. This is a rejection of one of the premises of Quine's paradox, viz., "o (Morning Star is congruent with Morning Star)". Note that Smullyan's argument is manifestly invalid, since the possible nonexistence of the morning star does not imply it is not necessary that the morning star is self-congruent or self-identical. To say that it is necessary that the morning star is self-congruent is to say that in each possible world in which the morning star exists, it is self-identical. This is quite consistent with saying that there are some possible worlds in which the morning star does not exist.

Smullyan is talking about artificial (formal or logical) languages (specifically, about artificial languages suitable for interpreting quantified modal logic), rather than natural or ordinary language. For this reason, Russell's argument that in ordinary language, expressions such as "Evening Star" or "Hesperus" are not proper names but abbreviated descriptions is not pertinent to the discussion. In his artificial, logical language, Smullyan is not required to use words in the way they are ordinarily used. But Smullyan does remark that the treatment of "Evening Star" or "Hesperus" as a logically proper name in modal contexts does not accord with ordinary language and therefore that this particular solution to Quine's paradox is disadvantaged relative to his second solution. He prefers the solution in terms of Russell's theory of the primary and secondary occurrences of definite descriptions, which (Smullyan believes) accords more with ordinary language. Thus, Smullyan concludes his review by saying:

In the judgement of the reviewer, the complications to which Carnap and Church resort underscore certain advantages obtained by retaining Russell's treatment of descriptions with its associated doctrine concerning the primary and secondary occurrence of descriptive phases. This doctrine, as Carnap admits [1947: 140], has the advantage of being in close accord with ordinary usage. It also permits a logical theory which unifies the theory of quantification with that of modality in a manner which Quine believes impossible. However, in defense of Quine's skepticism, in the reviewer's opinion, there is not a scintilla of evidence. [1947: 141].

This point needs to be emphasized, since some historians of philosophers do not correctly understand the relation between Smullyan's ideas and Marcus' ground-breaking ideas. For example, Michael Devitt [1989: 220], after indicating that Marcus originated the concept of rigid designation in her [1961], has a footnote implying that Fitch and Smullyan antedated her. Devitt makes no attempt to present textual evidence to support his claim, so it is hard to evaluate the basis of his claim. But Scott Soames does quote from Smullyan and thereby makes the source of his misunderstanding of Smullyan easy to identify. Soames thinks Smullyan is both talking about names in ordinary language and is advocating the theses that ordinary names are directly referential and obey the necessity of identity. Soames says that, according to Smullyan there is a distinction between "definite descriptions and 'what is ordinarily understood by 'proper name'" [Humphreys and Fetzer, 1998: 68] and Soames believes that this quote from Smullyan is referring to ordinary proper names. But as I have already indicated from my quotes from Smullyan's review, what he means is what is ordinarily understood, by philosophers of logic, to be a proper name, and he says exactly what these proper names are: "We have already seen that in terms of Russell's theory of descriptions and proper names this contradiction can be resolved." [1947:141, my italics] Second, Soames is mistaken is claiming in Smullyan's [1947], "the necessity of identity for names, but not descriptions, is implicitly endorsed" [Humphreys and Fetzer, 1998: 68]. However, as I have already quoted from Smullyan's piece, he explicitly rejects the necessity of identity for names; Smullyan writes that the necessity of identity thesis is false since it is not necessary that the morning star exists: "For if it is not necessary that the morning star exists then it is not necessary that the morning star is self-congruent" [1947: 140].

Furthermore, contrary to Soames' claim that Smullyan is talking about ordinary language rather than artificial language, Smullyan makes it clear throughout his article that he is talking about artificial language. This is the point of his mentioning that the correspondence of his theory of artificial language to "ordinary language" is an "added" (but unnecessary) advantage of his preferred, artificial language theory that names are not directly referential but are abbreviated descriptions. Thus, Smullyan says that one of his ways of resolving Quine's paradox in an artificial language has, as an added advantage to its logical validity, the feature of "being in close accord with ordinary use." Being in accordance with ordinary language is not a necessary condition of (but merely an added advantage for) a solution to Quine's paradox, since Smullyan takes the discussion to be about a logical language used in an interpretation of quantified modal logic, not about ordinary language. Smullyan is also saying that this added advantage belongs to the resolution in terms of primary and secondary occurrences of descriptive phases, not to the resolution in terms of logically proper names, which reflects Smullyan's belief that names in ordinary languages are not directly referential, but truncated descriptions. (Smullyan reaffirms his preference for a descriptional theory in his [1948].) Here we are still a long way from the Millian theory of ordinary proper names, the necessity of identity for Millian names, the intersubstitutivity of Millian names in modal contexts, the rigidity of Millian names, etc., that first appear in Marcus' [1961].

Soames concludes about Smullyan [1947], Fitch [1949; 1950] and Marcus [1961] that "this response to Quine, first given by Smullyan, and later repeated by Fitch, and then Marcus, was a straightforward application of Russell . . ' [Humphreys and Fetzer, 1998: 72, my emphasis].. It should be evident from the texts I quoted that it is false that Marcus repeated Smullyan's ideas, and that it is false that Marcus' ideas are a "straightforward application of Russell".

At least three steps need to made to get from Smullyan's papers (and Fitch's relevantly similar papers) to the contemporary nondescriptivism and rigidity in the philosophy of language and essentialism in metaphysics. First, somebody needs to endorse, and not merely mention as one of many possible solutions to Quine's paradox, the thesis that names are directly referential and intersubstitutable in modal contexts. Second, somebody must start talking about natural language and ordinary names, and not merely about names in an artificial language used for purposes of interpreting quantified modal logic. Third, the names need to be conceived as Millian names (in the broad and familiar contemporary sense) rather than as Russellian logically proper names.

The first step (endorsing the direct reference and substitutivity thesis) is taken by Marcus in her "Extensionality" [1960]. In this article, she continues in the tradition of talking about artificial or logical languages that are suitable for interpreting quantified modal logic. She sums up her paper as follows: "I have tried in this brief paper, to characterize the theory of extensionality, and to show that logical systems are more or less extensional" [1960: 62]. In a footnote [1960: 61, n. 2] Marcus represents herself as continuing in the Fitch (and thus Smullyan) tradition of stating that one of the ways to resolve Quine's paradox is to treat the relevant expressions as proper names, in which case there is intersubstitutivity. She writes about the sentence "It is necessary that the evening star is the evening star" and says that if it "involves proper names of individuals then 'the evening star' may replace 'the morning star' without paradox . . ." [1960: 61].

In Marcus' [1960], we do not see the reservations expressed by Smullyan (and Fitch) about the response to Quine that involves treating "evening star" as a proper name, and we do not find here any sympathy with Smullyan's and Fitch's suggestion that a solution in terms of definite descriptions may be preferable because of greater conformity with ordinary usage. Marcus views the "proper name" solution as provably valid, given her theorems about identity, indiscernability, weaker equivalences, and about the intersubstitutivity of expressions; she argues that the "proper name" solution follows from her theorems. But Marcus in this article has her attention only on the interpretation of logical systems, and is not concerned to make any claims about ordinary usage. Further, we have as yet no indication that she has in mind Millian names as distinct from the Russellian names that Fitch and Smullyan used, although the reference to Fitch suggests that she does not at this time see herself as departing from the theory of names employed in Fitch's and Smullyan's articles. The "Big Leap" into the New Theory of Reference is made next year, in her "Modalities and Intensional Languages" [1961]. This is where the Russellianism that dominated most of analytic philosophy of language and metaphysics for the first half or longer of the 20th century received its first decisive and fatal blow.

4. The Leap to Natural Language and a Detailed Theory of Modal Stable Expressions in Marcus' [1961]

The striking feature about Marcus' [1961] is that she restates the "proper name" solution to Quine's modal paradox and then proceeds to show its conformity to ordinary or natural language and that the relevant "proper names" are ordinary names in the Millian sense and exhibit necessity of identity. She further argues that Millian names are stable in modal contexts and are modally stable signifiers. She later in the article provides a possible world semantics for quantified S4, which includes modally stable individual constants, and thereby provides a formal backing to her theory of rigid names in natural language. The so-called "new theory of reference" makes its first appearance in this article, and appears in a full and explicit form. In addition to the ideas mention, other ideas belonging to the "new theory of reference also first appear in her [1961], which I have discussed in sections one and two. One idea in the 'new theory of reference' that does not appear here is the causal or historical chain theory of names, originated by Peter Geach in [1969] but almost universally misattributed to Kripke's 1970 talk ("Naming and Necessity") and Donnellan's [1970] paper. Marcus's sympathetically mentions and even seems to endorse some version of this Geach-Donnellan-Kripke theory in her [1975], but there seems to be at least a typo in that her reference to Geach (one of the extremely few references to Geach in this connection in the literature) lists Geach's [1969] article as having a 1970 date, thereby obscuring Geach's historical priority to Donnellan and Kripke.

There is a historical and logical connection between the theses discussed by Quine and Carnap in the 1940s and Marcus' first statement of the modally oriented, direct reference theory of ordinary names. Marcus applies to natural language something analogous to Quine's "universal intersubstitutivity thesis" and the idea that names are always "purely designative", and also something relevantly analogous to Carnap's "name-relation method". Quine and Carnap rejected this position for both natural and artificial languages, but Marcus adopts a version of this position for both logical and natural languages (and defends it at length in her [1975]). We have seen how these theses were discussed by Quine, Carnap, Smullyan and Marcus [1960] in their interpretation of quantified modal logic; also see Follesdal [1961]. Marcus' [1961] stands out by virtue of her arguments that these theses apply to ordinary language.

4a. Ordinary Language

In the first three sections of this paper, I analyzed the meaning of Marcus' [1961]. However, here I need to quote enough of this article to show that "the Big Leap" to a theory nondescriptivist, rigid names in natural language is made. A lengthy quotation is needed, since many philosophers are under the misimpression that Marcus does not discuss natural language in this article and thus that her [1961] is not importantly different from her [1960] article on artificial language. For example, both Soames and Burgess have put forth unsubstantiated statements in [Humphreys and Fetzer, 1998] that it is doubtful that Marcus is talking about natural language at all in her [1961]. But a reading of the following passages will show that nothing could be farther from the truth than the claim that Marcus never discusses natural language in her [1961] or that it is doubtful that she meant her theory to apply to ordinary language.

At some points of her 1961 paper, Marcus is talking about artificial languages (e.g., in the entire section called "Semantic Constructions"), and at other points she is talking about natural languages, such as the following passage, which is a continuous text in Marcus' [1961: 309-10]:

In fact it often happens, in a growing, changing language, that a descriptive phrase comes to be used as a proper name--an identifying tag--and the descriptive meaning is lost or ignored. Sometimes we use certain devices such as capitalization and dropping the definite article, to indicate change in use. "The evening star' becomes 'Evening Star', and they may come to be used as names for the same thing. Singular descriptions such as 'the little corporal', 'the Prince of Denmark', 'the sage of Concord', or 'the great dissenter' are as we know often used as alternative proper names of Napoleon, Hamlet, Thoreau and Oliver Wendell Holmes. One might even devise a criterion as to when a descriptive phase is being used as a proper name. Suppose through some astronomical cataclysm, Venus was no longer the first star of the evening. If we continued to call it alternatively 'Evening Star' or 'the evening star' then this would be a measure of the conversion of the descriptive phase into a proper name. If, however, we would then regard (10) [The evening star eq the morning star] as false, this would indicate that 'the evening star' was not used as an alternative proper name of Venus. We might mention in passing that although the conversion of descriptions into proper names appears to be asymmetric, we do find proper names used in singular descriptions of something other than the thing named, as in the statement 'Mao Tse-tung is the Stalin of China,' where one intends to assert a similarity between the entities named.

That any language must countenance some entities as things would appear to be a precondition for language. But this is not to say that experience is given to us as a collection of things, for it would appear that there are cultural variations and accompanying linguistic variation as to what sorts of entities are so singled out. It would also appear to be a precondition of language that the singling out of an entity as a thing is accompanied by many--and perhaps an indefinite or infinite number--of unique descriptions, for otherwise how would it be singled out? But to give a thing a proper name is different than giving a unique description. For suppose we took an inventory of all the entities countenanced as things by some particular culture through its own language, with its own set of names and equatable singular descriptions, and suppose that number were finite (this assumption is for the sake of simplifying the exposition). And suppose we randomized as many whole numbers as we needed for a one-to-one correspondence, and thereby tagged each thing. This identifying tag is a proper name of the thing. In taking our inventory we discovered that many of the entities countenanced as things by that language-culture complex already had proper names, although in many cases a singular description may have been used. This tag, a proper name, has no meaning. It simply tags. It is not strongly equatable with any of the singular descriptions of the thing . . .

A few comments about this passage are in order. Obviously, if any philosophy of language is about natural language, this passage exhibits such a philosophy. Obviously, the example of randomized whole numbers being used as tags is used to illustrate the idea that tags or proper names in natural languages have no meaning at all and differ in this respect from descriptions. Thirdly, we can see in this passage that Marcus, not Keith Donnellan, was the first philosopher to note that definite descriptions have a purely referential use, and are not always used attributively. Indeed, the first seven sentences in this quoted passage explain her distinction between referentially used definite descriptions and attributively used definite descriptions. (For a further analysis of this comparison, see my essay #7 in [Humphreys and Fetzer, 1998].)

4b. The Detailed Concept of a Modally Stable Signifier

We also see in this article the modal argument for direct reference and the formal semantical notion of a modally stable and directly referential name, or the notion of a "rigid designator" in general. Marcus does not use the expression "rigid designator" for understandable reasons, namely, that this is an expression coined by Kripke and Kripke [1971; 1972] popularized Marcus' [1961] theory of modally stable names and modally unstable definite descriptions in a way that is too imprecise for Marcus' liking and that is also potentially misleading. (At the place in [1972] where Kripke introduced the notion of proper names as rigid designators, he made no mention of the distinction between rigid proper names and rigid descriptions). Thus, I shall instead call Marcus' proper names or tags "modally stable signifiers", meaning by this expression only the modal properties that Marcus ascribed to tags in her [1961].

Neither Marcus nor later philosophers talked about (what I shall call) the detailed concept of a modally stable signifier, so here I am making something more explicit than what is present in Marcus' or others' texts. texts.

The detailed concept of modally stable signifier contains an ordinary language part and a logical language part. The ordinary language part belongs to the philosophy of natural language and is an intuitive explication of the difference between modally stable and unstable signifiers in ordinary language. The logical part is a definition of modally stable signifier in terms of a semantics for modal logic, and embodies the semantical, logical notion of the referent of a singular term at a model or possible world. Somebody has a detailed concept of a modally stable signifier if and only if she has both the ordinary language and logical language parts.

The ordinary language part of the concept does not require a model-theoretic semantics, the notion of a possible world, or the concept of a singular proposition. The concept of a modally stable signifier can be stated in ordinary language terms by using only modal prefixes or adverbs ("possibly", "necessarily") and subjunctives ("might have been", etc.). The modally stable/unstable distinction between names and contingent definite descriptions is present in the following passage in Marcus [1961]:

Let us return now to (10) and (15). [(10) is "The evening star eq the morning star" and (15) is "Scott is the author of Waverley"]. If they express a true identity, then "Scott" ought to be anywhere intersubstitutable for "the author of Waverley", and similarly for "the morning star" and "the evening star". If they are not so universally intersubstitutable--that is, if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g. someone else might have written Waverley, the star first seen in the evening might have been different from the star first seen in the morning--then they are not identities. [1961: 311]

I shall now use the conceptual materials in Marcus' [1961] to formulate the logical part of the concept of a modally stable signifier. The logical explication is based on the semantics for modal logic and includes notions of individual constants or variables, modal operators, predicates, models (worlds) and related ideas. The logical explication uses these notions to define modal stability in terms of the referent of a singular term at a model or possible world.

Let us see how this explication is based on Marcus' 1961 model-theoretic construction that "corresponds to the Leibnizian distinction between true in a possible world and true in all possible worlds" [1961: 320). Her construction embodies the simplifying assumption that there is only one two-place predicate (R). Marcus presents:

a language (L), with truth functional connectives, a modal operator (à), a finite number of individual constants, an infinite number of individual variables, one two-place predicate (R), quantification and the usual criteria for being well-formed. A domain (D) of individuals is then considered which are named by the constants of L. A model of L is defined as a class of ordered couples (possibly empty) of D. The members of a model are exactly those pairs between which R holds. To say therefore that the atomic sentence R(a₁a₂) of L holds or is true in M, is to say that the ordered couple (b₁,b₂) is a member of M, where a₁ and a₂ are the names in L of b₁ and b₂. If a sentence A of L is of the form ~B, A is true in M if and only if B is not true in M. If A is of the form B₁× B₂ then A is true in M if and only if both B₁ and B₂ are true in M. If A is of the form ($x)B, then A is true in M if and only if at least one substitution instance of B is true (holds) in M. If A is àB then A is true in M if and only if B is true in some model M₁.

We see that a true sentence of L is defined relative to a model and a domain of individuals. A logically true sentence is one which would be true in every model. [1961: 319].

Here the models M, M₁, etc. are possible worlds, the individuals in the domain D exist in each world, and these individuals are referents of the individual constants a₁, a₂, etc. In a later article, Marcus said of her 1961 theory that "in my own sketch of a semantics for modal logic, the domains of individuals assigned to alternative worlds were coextensive. Given that one of the worlds is the actual world, no entities are spawned that are not in this world and no entities of this world are absent in others" [Marcus, 1993: 195].

Marcus uses the phrase "member of a model" or "member of a world" to refer to any pair of individuals between which R holds; thus the empty world contains no "members" in the sense that it contains no pair of individuals between which R holds.

If one prefers, one could say that the ordered couples are members of a model whereas the individuals in the domain D are present or exist in a model (the "present/absent" terminology conforms to Marcus' later phraseology in [1993: 195]. The "present in" or "exists in" terminology may be explained in part in terms of the fact that the quantifiers in each model range over the domain D, which is required for her derivation of the Barcan formula. Since the domain D is fixed, Marcus' worlds differ from one another in that her relation R holds between different individuals in different worlds (or does not hold between any individuals in her empty world). For example, Alice and Ed exist in one world M as related by the relationship of friendship and exist in another world M₁, but not as related by this relation.

This semantics for modal logic contains the notion of a referent of a singular term at a possible world and a dyadic rule of interpretation that connects an individual constant a to an individual b in the domain D. This may be expressed as the rule V(a)=b, where a is the individual constant and b the individual in the domain. This dyadic rule of interpretation entails the triadic rule of interpretation V(a, M)=b, which assigns to a the referent b in the model M. This entailment holds since Marcus' domain of individuals is constant across her models.

The triadic rule of interpretation is the formal semantical notion of a referent of an individual constant at a possible world. This formal semantic notion of modally stable signifiers, a "rigid" individual constant or name is also present in early work by Follesdal [1961] and Hintikka [1961; 1963], but Marcus was the first to present and publish this notion. It does not seem that this idea can be traced back earlier than 1961. Kripke's [1959] included neither individual constants nor the principle of the necessity of identity, which are both required for formal anologues of modally stable names. Carnap's writings [1947; 1947] do not advocate the formal or informal explication of a modally stable signifier. Carnap's proper names are not directly referential but have intensions (individual concepts) as well as extensions and in modal contexts Carnap's names do not refer to their extensions (the relevant individuals in the various possible worlds) but to their intensions. A necessary condition of a name being a modally stable signifier is that it has the same referent in both extensional and modal contexts, and Carnap's names, unlike Marcus', do not meet this condition. In fact, Carnap explicitly rejects the theory of what he calls "the name-relation" [1947: 98], which implies certain theses associated with the theory of modally stable signifiers, most notably, the thesis that co-referring names are intersubstitutable in modal contexts.

Accordingly, Marcus' [1961] was the first work to contain the detailed concept of a modally stable signifier, since it included both the ordinary language explanation the formal semantical or logical concept of a referent of a singular term at a possible world. Her [1961] contained the materials for two definitions of modally stable signifiers, one which applied only to directly referential expressions and one which applied to modally signifiers of all types, including descriptively stable descriptions. The direct reference definition would (formally put) read as follows:

(Direct): A token of a singular expression a is a modally stable signifier of an individual b if and only if (i) a directly refers to b in the actual world M, (ii) a directly refers to b at each possible world M₁, M₂, etc., in which b exists.

This covers not only Marcus' tags (proper names), but also her referentially used definite descriptions. Kaplan [1989] later argued it also covered indexicals, such as "I" and "here".

A second definition applies to descriptively stable expressions as well:

(Direct or Descriptive): A token a of an expression is a modally stable signifier of an individual b if and only if (i) a refers to an individual b in the actual world M, (ii) a refers to b at each world M₁, M₂, etc. in which b exists and (iii) a does not refer to anything at worlds M₃, M₄, etc. in which b does not exist.

Condition (iii) is meaningless in Marcus' [1961] semantics, since she held a constant domain theory; there are no worlds at which b does not exist. When Marcus is discussing a varying domain semantics in her [1971], she advocates adding the condition that in respect of the worlds in which b does not exist, a does not designate anything. In "Essential Attribution", Marcus writes that "individual names don't alter their reference, except to the extent that in some worlds they may not refer at all" [1971: 194]. Although some philosophers now mistakenly think the modally stable signification is equivalent to direct reference, and thus use only the first definition (Direct), it is to Kripke's credit that he used a layperson's version of the second definition in his [1971] and [1972] in which he popularized Marcusian philosophy and made it accessible in intuitive terms to the general philosophical public.

5. Conclusion

We have seen some of the respects in which Marcus' [1961] constituted the decisive and most influential turning point from the extensional Russellian philosophy of language and metaphysics that dominated most of the 20th century, and turned philosophers towards the modally oriented philosophy of language and metaphysics of the last forty years of the century. Many of her other contributions have been mentioned in essays by others, and I have presented further of her influential ideas in my essays in [Humphreys and Fetzer, 1998], and I also explained the influence of her ideas on metaphysics, metaethics and ethics in Ethical and Religious Thought in Analytic Philosophy of Religion. The only surprising fact is that it took until the mid-1990s for it to become widely recognized that Marcus' [1961] constituted the decisive turning point from the Russellian extensionalism of the first half of the 20th century to the modally oriented philosophy of language and metaphysics of the last forty years of the century.[1]

BIBLIOGRAPHY

Almog, Joseph. (1986). "Naming without Necessity", JOURNAL OF PHILOSOPHY 83: 210-42.

Burgess, John. (1998a). "Marcus, Kripke, and Names", in Paul W. Humphreys and James F. Fetzer (eds.) THE NEW THEORY OF REFERENCE: KRIPKE, MARCUS, AND ITS ORIGINS. Dordrecht. Kluwer Academic. 1998.

--------------. (1998b). "How Not to Write History of Philosophy: A Case Study" in Paul W.

Humphreys and James F. Fetzer (eds.) THE NEW THEORY OF REFERENCE: KRIPKE, MARCUS, AND ITS ORIGINS. Dordrecht. Kluwer Academic. 1998.

Carnap, Rudolph. (1947). MEANING AND NECESSITY. Chicago. University of Chicago Press.

----------------. (1946). "Modalities and Quantification", JOURNAL OF SYMBOLIC LOGIC 11: 33-64.

Devitt, Michael.(1989). "Against Direct Reference", in MIDWEST STUDIES IN PHILOSOPHY, VOLUME XIV: CONTEMPORARY PERSPECTIVES IN THE PHILOSOPHY OF LANGUAGE II (eds. P. French et. al.): Notre Dame: University of Notre Dame Press, pp. 206-240.

Donnellan, Keith. (1970). "Proper Names and Identifying Descriptions", SYNTHESE 21 (nos. 3-4): 335-358.

________________. (1966). "Reference and Definite Descriptions", THE PHILOSOPHICAL REVIEW 75: 281-304.

Fit, Kit. (1995). "Senses of Essence" in Sinnot-Armstrong, Walter; Raffman, Diana; Asher, Nicholas, (eds.) MODALITY, MORALITY, AND BELIEF: ESSAYS IN HONOR OF RUTH BARCAN MARCUS. Cambridge University Press, 1995.

Fitch, Frederick. (1950). "Attribute and Class", in Marvin Farber (ed.), PHILOSOPHICAL THOUGHT IN FRANCE AND THE UNITED STATES. Buffalo: University of Buffalo Publications in Philosophy.

-----------. (1949). "The Problem of the Morning Star and the Evening Star", PHILOSOPHY OF SCIENCE 16: 137-141.

Follesdall, Dagfinn. (1961). REFERENTIAL OPACITY AND MODAL LOGIC. Harvard University dissertation, unpublished.

--------------. (1997). ETHICAL AND RELIGIOUS THOUGHT IN ANALYTIC PHILOSOPHY OF LANGUAGE. New Haven: Yale University Press.

_______________. (1996). "Causation and the Logical Impossibility of a Divine Cause", PHILOSOPHICAL TOPICS 24: 169-91.

Smith, Quentin and Oaklander, L. Nathan. (1995). TIME, CHANGE AND FREEDOM. London. Routledge.

Quentin Smith. (1994) "Temporal Indexicals" in (L. Nathan Oaklander and Quentin Smith, eds.) THE NEW THEORY OF TIME. New Haven. Yale University Press.

Smith, Quentin and Craig, William Lane. (1993). THEISM, ATHEISM AND BIG BANG COSMOLOGY. Oxford: Clarendon Press.

-------------. (1991). "The New Theory of Reference Entails Absolute Time and Space", PHILOSOPHY OF SCIENCE 58: 411-16.

__________ (1988). "Tensed States of Affairs and Possible Worlds", GRAZER PHILOSOPHISCHE STUDIEN 31: 225-35.

__________ (1986). THE FELT MEANINGS OF THE WORLD: A METAPHYSICS OF FEELING. West Lafeyette: Purdue University Press.

Smullyan, Arthur. (1948). "Modality and Description", THE JOURNAL OF SYMBOLIC LOGIC 13: 31-37.

________________. (1947). Review of Quine's "The Problem of Interpreting Modal Logic", THE JOURNAL OF SYMBOLIC LOGIC 12: 139-141.

Soames, Scott (1998a). "Revisionism About Reference: A Reply to Smith", in Paul W. Humphreys and James F. Fetzer (eds.) THE NEW THEORY OF REFERENCE KRIPKE, MARCUS, AND ITS ORIGINS. Dordrecht. Kluwer Academic. 1998.

----------------. (1998b). "More Revisions about Reference, in Paul W. Humphreys and James F. Fetzer (eds.) THE NEW THEORY OF REFERENCE: KRIPKE, MARCUS, AND ITS ORIGINS. Dordrecht. Kluwer Academic. 1998.

[1] More material on the topic of this paper is contained in Paul Humphreys' and James Fetzer's (eds.) The New Theory of Reference: Kripke, Marcus, and Its Origins and in my Ethical and Religious Thought in Analytic Philosophy of Language. In addition, Humphreys' and Fetzer's book contains an excellent chapter by Sten Lindstrom (on Kanger's, Hintikka's and Kripke's modal logic) and a significant except from Dagfinn Follesdal' historically important 1961 dissertation on modally stable expressions. Their book also contains four essays by Burgess and Soames.

THE RIGID DESIGNATION THEORY

BIBLIOGRAPHY