Lycan and Shapiro (1986) constructed a system of possible worlds, out of our familiar if not unproblematic abstract entities, identifying "nonactual worlds" with certain sets of structured propositions--including singular propositions--and identifying "nonexistent" individual objects with certain complex properties.  In our formal system  Shapiro and I accommodated the transworld identity, in the strict sense, of actually existing individuals:  Since "worlds" are just sets of propositions, it is easily determined which individuals inhabit a given nonactual world; individual i is in world w iff w has as an element some proposition singularly about i (that is, some proposition of the form <i, ?x(Fx)>.)  Thus, actually existing individuals are transworld-identified directly with themselves, not with sets of properties or with mere counterparts; we embraced Haecceitism.

    We wanted to treat the transworld identity of "nonactual individuals" differently, since neither Shapiro nor I would countenance their existence in any Concretist sense.  To say that "the present king of France could exist" is to say only that it is possible for there to be a present king of France, which for us is in turn to say that some possible world has as a member the proposition "(Ex)(x is now king of France)."  Unlike actuals, we said, merely possible individuals do have qualitative essences--individuating properties that necessarily apply to at most one individual per world apiece--and "accessible" ones at that.  Let us call these qualitative essences characteristic properties, or "characteristics" for short.<1>  In the Lycan-Shapiro system, in fact, nonexistent possibles are identified with their characteristics.  Moreover, the characteristics at each possible world are stipulated.  A given individual’s characteristic serves to pick out that individual at other worlds.  Suppose, for example, that P is a characteristic at world w, and  thus that "(Ex)Px" is a member of w.  Suppose also that another world w’ contains "(Ex)(Px & P’x)" for some property P’; then we would say that the merely possible individual identified by P in w has P’ in w’.<2>  Thus Shapiro and I pejoratively distinguished the individuation of nonexistents from that of actuals.<3>  We embraced traditional essentialism regarding the former despite our Haecceitism regarding the latter.
    I now doubt the wisdom of that policy, having since found what I think is powerful motivation for Haecceitism regarding nonactuals.  I am not entirely convinced either way, so I shall continue to take the Lycan-Shapiro line seriously; call it the Conservative Position.  In this chapter I shall expound a defense of the Conservative Position, but go on to argue the opposing view as well and show how that view resists the Conservative arguments.  Here, then, comes the Conservative case.

    Contra Concretism, it would seem that nonactual individuals differ sharply from actual ones:  (i) Actual individuals are given; by definition, they are part of the world we live in.  They can be perceived and ostended.  (And I assume that the proper names of actual individuals neither abbreviate flaccid descriptions nor are equivalent to them in any other semantical way; see Chapter 7.)  By contrast, mere possibilia are not given or (pace Routley [1980]) encountered, but specified, stipulated, or constructed.  Moreover, (ii) the only way to specify a nonexistent is to proffer a description--"the golden mountain," "the present king of France," "the winged horse ridden by Bellerophon," "Jimmy Carter’s older brother."<4>  (It is these specifying descriptions that Shapiro and I took as the characteristics.)  And consequently (iii) it is arguable that nonexistents figure in no singular propositions, and that they are the objects of neither de re modalities nor de re propositional attitudes.<5>  Offhand one sees no clear sense in which one can have a particular nonactual object in mind as opposed to one which is qualitatively just like it (though later on I shall take a giant step toward providing such a sense).
    I pause to rebut an objection to the Conservative Position, put to me on separate occasions by Pavel Tichy and Penelope Mackie.  It goes roughly as follows:  Assume that (the real) Everest could have been uniquely made of gold.  Then there is a world w1 at which Everest = the golden mountain.  Everest at w1 is of course transworld-identical with Everest here at @, and on the Conservative doctrine, Everest at w1 is also transworld-identical with the golden mountain at a third world w2 even though the golden mountain at w2 is not Everest there and (let us say) Everest does not even exist at w2.  Contradiction:  By transitivity of identity, Everest here = the golden mountain at w2, but by hypothesis that is false.
    What is wrong with the argument is that transworld identity is not the same notion as contingent identity within a world, and so the appeal to transitivity is an equivocation.  There are at least three ways of seeing the point.
    (i)  Transworld identity is the relation being a value of the same world-line as.  That is, you at @ are transworld-identical with Jones at w  just in case you and Jones are manifestations of the same world-line at your respective worlds.  Now actually that should be, manifestations of a same or common world-line, because every individual at every world is a value of more than one world-line (different kinds of world-lines corresponding to different sortal kinds of item, á la Geach [1967] and Hintikka [1972]).  A human-being-shaped world-line is a different sort of world-line from a role-shaped or officey world-line.  Regarding Everest being the golden mountain at w1, that single object is the value at w1 of each of two distinct world-lines, a mountain-shaped one and a Russellian role- or niche-shaped one.  So Everest at w1 does share a world-line with Everest at @, and Everest at w1 does share a world-line with the golden mountain at w2; but it does not follow that Everest at @ shares a world-line with the golden mountain at w2, for the foregoing two shared world-lines are distinct world-lines.  Hence it does not follow that Everest at @ is transworld-identical with the golden mountain at w2.  (Doubtless on some crazy scheme there is a truly weird world line that has as values both Everest at @ and the golden mountain at w2, but that is irrelevant here.)
    (ii) In other perfectly ordinary cases of intra-world contingent or role-occupant "identity" the inference obviously fails.  At @ David Lewis is a philosopher and George Bush is President of the USA.  Let us say that at w1 David Lewis is President of the USA, and at w2 Irving Smedley is President of the USA, while David Lewis does not exist at w2.  The presidency is the same office from world to world, and Lewis and Smedley are the same official from w1 to w2, but as before no one would think of inferring that the real David Lewis is President at w2 just because he is President at w1.  Qualitative essences for nonactuals (or for anyone) make those "individuals" into roles or offices, occupied at a world by this or that individual in the more "genuine" sense of the term "individual."
    (iii) Whether one is a Lewisian Concretist-cum-counterpart-theorist or an Ersatzing wimp, the relation of transworld identity and intra-world identity are not the same relation.  (a) The latter, but not the former, can be contingent.  (b) If one is a Lewisian, there is no genuine transworld identity but only counterpartship, while there is genuine intra-world identity.  If one is an Ersatzer, transworld identity is literal numerical identity for actual individuals, but "identity" within a nonactual world consists just in an identity proposition’s being an element of that world.  (When the identity proposition is a contingent one, it is not even genuinely an identity proposition, being equivalent in Russell’s way to a subject-predicate form.)

    Despite my initial remarks in its defense, the tendentious Conservative Position is entirely questionable.  One may be particularly dissatisfied with its assumptions that there are no singular propositions involving nonactuals and that (accordingly or vice versa) there are no attitudes or modalities de nonactuals.  Among others, Lewis (1983) and McMichael (1983a, 1983b) have maintained that an adequate modal semantics must accommodate all these features, contrary to our assumptions;<6> let us consider McMichael’s argument in particular.

    McMichael raises a question regarding modal properties of nonactuals.  It takes several different forms, but I think the most trenchant version of the objection is a simple modal one:  Given such-and-such a nonexistent object introduced and universally accepted as being "the F," could that object have failed to be F?
    Intuitions differ from case to case, and from person to person even within cases.  I distinguish three main types of case: (a) "nonexistents" generated solely by definite descriptions made up on the spot, such as "the golden mountain" or "the town just north of Mayberry in which Marie Curie discovered the secret of eternal life"; (b) fictional objects already well entrenched in literature and/or in culture;<7> and (c) nonactual items endowed with their own modal properties by initial stipulation.  Let us consider these seriatim.
    Could the golden mountain have been made of aluminum, or of marshmallow?  Could the town just north of Mayberry... never have been remarkable in any way at all, or have been located near Kinston?  I do not know what to make of these questions.  They seem senseless, or at least to have no answers, since the descriptions they involve have been introduced brutely, without context, and there is no background story or other information against which to measure modal stretch.  For nonexistents of type (a), descriptive or qualitative characteristics of the sort we have posited seem entirely appropriate.  However, types (b) and (c) are less straightforward.  Let us turn to the case of fictional individuals; but first a small digression is required.
    For clarity,<8> we must distinguish fictional characters properly so-called from fictional persons or other individuals.  A fictional character, in one common use of the phrase (cf. van Inwagen [1977]), is an actual though abstract literary entity, ontologically on a par with a novel or a story.  In one sense characters are proper consistuents of novels and stories,<9> and are open to description and evaluation of the same sorts as are their containing works themselves:  A character may be well-drawn or a two-dimensional cardboard pinup, vital or inessential to the plot, more ingeniously or less creatively conceived than Dorothy Sayers’ Mervyn Bunter, a faithful realization of the author’s original intention or an unexpectedly evolved departure, popular or unpopular with readers, etc.  Such things cannot (except metaphorically) be said of real flesh-and-blood people; no more can they be said of fictional flesh-and-blood people in the sense I intend.  A fictional person or other individual is a nonactual possibilium having, at its own home world or worlds, the properties described in some (presumptively but not always actual) fiction.<10>  Thus, fictional people are fat or thin, smart or stupid, candid or devious, and the like.
    Once the foregoing distinction is enforced, the question of whether there are fictional individuals at all  becomes controversial.  Some writers, such as van Inwagen (1977), decline to treat "fictional characters" as nonexistent possibilia or as denizens of possible worlds even if they grant that in some sense there are other kinds of nonexistent possibilia; other writers (including Kripke [1972/1980, 1972b], Kaplan [1973] and Fine [1984]) deny possibility to fictional individuals tout court.  I shall not join the labyrinthine issues involved here;<11> for purposes of this chapter I shall simply continue to assume that in whatever sense there are nonactual individuals, there are in particular fictional individuals, falling in with the view that actual fictions describe nonactual possibilia.  If positive argument is needed, I offer just three brief points:  (a) Fictional individuals such as Pegasus, Hamlet, Sherlock Holmes et al. have always been used as paradigmatic examples by partisans of nonexistent possibles (though I suppose such philosophers may have been hopelessly confused from the beginning).  (b) Novels and stories generally seem to say things that could have been true even though they are not.  If so, and if possible-worlds semantics constitutes our standard device for handling modal statements, then sentences occurring in fictions are true at worlds other than our own and so it seems that fictional individuals exist at those worlds.  (c) As we shall see, fictional individuals appear to have modal properties, and--again according to our general practice of explicating modality in terms of worlds--this seems to show that fictional individuals somehow inhabit worlds.  Let us now end our digression and return to this last issue.

    Consider Perry White, fictional editor of the Daily Planet.  Might he have gone into bricklaying, philosophy or nursing instead of journalism?<12>  I am pulled two ways here.  On the one hand, "Perry White" is only a fictional character, a creation of a human author; all there is to him, so to speak, is his two-dimensional role in the story as Clark Kent’s unsympathetic, cigar-chomping boss.  Had the author left out this "boss" character entirely and at the same time introduced a janitor, or an itinerant drunk who wanders in off the street, and happened to name it "Perry White," this new character would not be Perry White in the sense in which we now use that name; it would simply be a different character despite the accidental sameness of monicker.
    On the other hand, as is well known, fictional characters do not have only the properties explicitly ascribed to them in their native works, but also those which may fairly be extrapolated or assumed on the basis of the text and its setting.  Now, the original character Perry White is (in the story) a man.  Presumably he has toes, even though we never see them and even though there is no specific toe length that he has.  Similarly, like any man he presumably has a genetic code, even though there is no specific genetic code that he has.  And why not suppose that he has the everyday sorts of modal properties that ordinary men have as well?  Every actual man is such that he might have pursued a different occupation, so why not Perry White?  In one episode or another White might even be moved to remind Clark Kent of this possibility, saying something like, "I should have listened to my mother when she told me to stay out of journalism because of idiot milquetoast jerk reporters like you!"
    But if we grant (for this or any other reason) that White might have joined a different trade, it seems we are saying that at a world other than that described in the Superman stories, White does go into bricklaying or whatever instead of editing the Planet.  And this conflicts with our assumption about his qualitative essence.  If his editorship is not a reliable transworld identifying mark, the same can be said of any other feature attributed to him by his creator; and so it seems we must award him a haecceity, just as if he were actual.  Yet this would be extremely problematic, as our original argument for qualitative essences implies and as I shall further consider below.  (Can we imagine a story just like the Superman corpus except that in it Perry White’s and Clark Kent’s haecceities are switched??)
    It might help to distinguish two sorts of truth, in a thoroughly familiar way: truth simpliciter from truth-in-fiction.  I want to say it is not true (period), even though it is indisputably true-in-fiction, that there is a person (Clark Kent) who can leap tall buildings at a single bound.<13>  If this distinction is sound, it facilitates the accommodation of our occasional Haecceitist feelings about fictional characters.  It is indisputably true-in-the-comics that Perry White has toes, genes and modal properties.  It is not true in reality--or, less tendentiously, true simpliciter--that "he" does.  Whatever account we might be moved to give of truth-in-fiction, what concerns us in fashioning semantics for natural language is truth, period, and it is not true in this sense that there is someone called "Perry White" who is a newspaper editor but might have been a bricklayer.  What is true--at the outside--is that there is a character, Perry White, who is (designated as) Clark Kent’s editor, and that is all there is to it; he has modal properties only in-the-story.
    Unfortunately the matter cannot be settled so easily.  For I have granted at least that White has his modal properties in-fiction, and as in Chapter 3, the semantics of this "in-fiction" operator remains to be determined.  When provided, it may be found to cause trouble for my brand of actualism.  Indeed we may expect it will, for the only natural possible-worlds interpretation of "It is true-in-fiction that P" is something like, "It is true that P at every world compatible with the implications and presumptions of the (relevant) story,"<14> and this seems to establish one or more worlds at which White becomes a bricklayer.

    The objection is reinforced by an argument of McMichael’s that is not tied to the vexing vagaries of fiction but is based on plain iterated modalities of type (c).  Again (almost) indisputably, there might have been someone having such-and-such a property F that also had but might have lacked a further property G; for example, Rose Kennedy might have had a son, distinct from Joe Jr., Jack, Robert and Ted, who went into philosophy but who might instead have gone into bricklaying.  Thus on anyone’s possible-worlds semantics, fiction and its awfulness completely aside, it seems there is a world w containing the supernumerary philosopher Kennedy and a further world w’ also containing that very Kennedy but at which he went into bricklaying instead of philosophy--not because of any work of fiction but simply because the envisioned possibilities seem genuine.  Thus we are forced to consider transworld identity-conditions for McMichael-individuals.
    We might naturally try to assimilate such individuals to cases of type (a).  That is, we might insist that the imagined philosopher son of Rose Kennedy must remain a philosopher throughout all the worlds he inhabits just as the golden mountain must remain golden, since he has in effect been stipulated on the spot in the same way.  But this would be to deny the truth of our original iterated modal sentence, and that sentence still seems true (in reality, not in fiction of any sort).  We must grant that "Rose Kennedy’s fifth son" is a philosopher at some worlds and not at others.  We can get away with granting that, so long as we are allowed to split the original description "Rose Kennedy’s fifth son who is a philosopher" into essence and accident: let that individual’s essence be simply "x is Rose Kennedy’s fifth son," abbreviated "R(x)."  One world (=/ @) contains R(x) as an characteristic and "(Ex)(R(x) & x is a philosopher)" as an element; yet another world containing R(x) has "(Ex)(R(x) & x is a comedian)" as an element instead.
    But if the philosopher son is only contingently a philosopher, might he not also be only contingently a son?  It would seem that just as there might have been an F (distinct from every actual individual) who is but need not have been G, there might have been an F (ditto) who need not have been F.  Consider "There might have been someone in the doorway who need not have been in the doorway."  If context provides an extrinsic essence for the supposed person (say, "the person I am about to say hello to"), there is no difficulty, but if no such extrinsic essence is ready to hand, the person is stuck in the doorway at least pro tem.  If we are to respect these iterated singular modalities in the absence of "extrinsic" essences, it seems we must move to nonactual haecceities of some sort after all; we must grant that there is a property of being N,  where N is a proper name of our imaginary philosopher, that persists from world to world despite variation of all his ordinary features.<15>  If there is a nonqualitative property of being Perry White, for example, and a nonqualitative property of being Clark Kent, then there is after all a world just like that of the Superman stories in which everything is the same except that White has all of Kent’s qualitative features and vice versa.
    Numerous objections may be brought, and have been brought, against this idea of nonqualitative haecceities for nonexistents.  The Conservative Position still has plenty of fight in it.<16>  Yet I think that the objections can be circumvented, albeit with some considerable effort, and that there is more to be said in support of such haecceities than might at first appear; so in what follows I shall offer, to those theorists who do insist on taking the iterated modal properties seriously, a Haecceitist alternative, since I maintain that the idea is coherent and defensible even if it seems unattractive.<17>

    Let us state more fully the main objections to Haecceitism for nonactuals.  One is that with which I began: that nonexistents can be introduced only by description.<18>  A second (closely related) is that even if we can coherently think of a world just like this one save for the switching of Adam and Noah, we simply cannot distinguish two worlds which differ only in the switching of the alleged haecceities of Rose Kennedy’s philosopher son who might have gone into bricklaying and her seventh, bricklayer son who might instead have gone into philosophy, or (cf. Adams [1981]) two worlds which differ only in the switching of two nonactual electrons.<19>  Third, if nonactuals have haecceities and can differ numerically without differing qualitatively, then it ought to be possible for us to have a particular object in mind, or to have a propositional attitude toward that object, without having in mind a qualitative twin; yet this does not seem possible.<20>
    Fourth, as Adams observes (1981, pp. 11-12), an ordinary haecceity (assuming real people have them) bears a special relation to its owner: it could not exist without its owner’s (that very person’s) existing also.  But if actualism is correct no nonactual "individual" can enter primitively into any relation, and in particular there can be no state of affairs which consists simply in Perry White’s exemplifying Perry-Whiteity.  To put the point slightly differently:  Predicates which express the haecceities of actual individuals are syntactically and semantically complex, consisting of the identity sign concatenated with a rigid designator whose reference has been fixed by a process involving causal contact with the referent itself; but a "nonactual individual" is not genuinely an individual on our view, and cannot stand as a term in a genuine identity predication.  Fifth, it is not easy to see the difference between an "unexemplified haecceity"<21> and a good old Meinongian nonexistent possible, once there is no visible residue of familiar qualitative properties to go proxy for the noxious Meinongian object; McMichael (1983a, p. 61) asserts that the introduction of haecceities for nonactuals "seems tantamount to acceptance of possibilism."  Sixth,<22> consider a single possible world containing two planets qualitatively identical to each other--say, each planet contains a replica of Conan Doyle’s Victorian England and in particular a Sherlock Holmes figure.  What conceivable ground could there be for choosing one of the two Holmes figures and deciding that it is the real Holmes while the other is only its qualitative duplicate?  To attach a Holmesian haecceity to either figure would be entirely arbitrary.
    All six of these objections have force;<23> that is why Shapiro and I fashioned our official semantics without recourse to unexemplified haecceities.<24>  But for those who want to take the putative modal properties of nonexistents more seriously, I now offer an approach that will take at least some of the sting out of our objections to haecceities for nonactuals.

    Most of the objections stem from the fact that nonactual individuals stand in no causal relation to us and are known only by description.  But it is possible to frame a Causal-Historical theory of reference, even for empty singular terms, that affords a finer-grained individuation scheme.  Consider fictional individuals again.  At least part of what makes Perry White the person he is are the circumstances of his character’s creation.  We might say that a fictional person qualifies as being (identical with) Perry White if and only if the relevant use of that person’s name is connected in the right historical way with Jerry Siegel’s original act of writing (in the real world).<25>  Several sorts of cases fall nicely into line with this suggestion.  (i) If some writer or cartoonist in Apex, North Carolina, who has miraculously never heard of the Superman comics, just as miraculously happens to invent a character also named "Perry White," that character is not the same character as Mr. Siegel’s Perry White, even if he or she is similar in remarkable respects to Clark Kent’s editor.<26>  However, (ii) if someone undertakes to write a spinoff strip about the original Perry White’s journalistic triumphs and tribulations or about his private life, with the original firmly in mind and with the intention of grafting the new stories smoothly onto the old--as Ruth Plumly Thompson continued L. Frank Baum’s Oz books with Baum’s explicit imprimatur, and as George Macdonald Fraser has brilliantly written about Thomas Hughes’ Flashman--I am willing to count the new White as being the same character as the old.<27>  (iii) Mr. Siegel could have written a new Superman story in which Superman (or another character) discovered that the Planet editor was an imposter, and that the real Perry White had years ago decided against journalism and gone into philosophy instead (but changed his name for reasons of euphony to "Hilary Putnam").  This would stick, it seems to me; after the new story had been published it would be true-in-fiction that Perry White was really a philosopher and had only erroneously been thought to be Clark Kent’s editor.<28>
    If some such Causal-Historical criterion could be made to work, it would provide a means of distinguishing qualitatively identical characters and so lessen the pressure of our objections to unexemplified haecceities.  The initial problem was that nonactuals are identified only by description because they are causally unconnected to us; but as we have seen, their names, their dubbings, and subsequent acts of referring or "referring" to them are not so  unconnected.  Potentially they can be used to distinguish descriptively identical characters.  If this conjecture is sound, our first objection fails.<29>  So does our second; for transworld identities in the "nonactual" analogue of Chisholm’s "Adam"/"Noah" example can be established similarly, when pegged to real-world fictive acts and authors’ intentions.  In particular, we can extrapolate Kripke’s emphasis on stipulation to the case of nonactuals.  What distinguishes the White-and-Kent-switched world from its progenitor is simply stipulation backed by the right historical connection and the right intentions, just as for existents.<30>  The same can be said even for McMichael-individuals: Rose Kennedy’s philosopher son is himself at another world, even if he has shed his philosophizing for bricklaying and so on, if the utterer of his original supposition so stipulates under the right conditions.  Our third objection goes wide as well.  Odd as it may seem, I can want Pegasus rather than a descriptively identical winged horse, if I know that the latter nonexistent’s name (homonymously "Pegasus") has a different ancestry from that which figures in the classical myth, e.g., if it was made up last week by my highly and coincidentally imaginative plumber.<31>
    Our fourth, fifth and sixth objections are harder to turn aside, even with the aid of our Causal-Historical idea, but we can say at least a bit in response to each of them, in turn:  Regarding the fourth, Adams is bothered by the fact that an ordinary (exemplified) haecceity is identified by reference to its owner and could not exist in the absence of singular states of affairs involving that owner, while no actualist ontology can provide a genuinely singular state of affairs involving a nonexistent.  But the Causal-Historical account can help at least with the identification problem; the haecceity of a nonexistent could now be identified by reference to a fictive act or other quasi-dubbing.  Moreover, we do not have to suppose that a haecceity exists in the absence of its owner, for we can say that at the relevant nonactual world, its owner does exist--the presence of the haecceity and the existence of its "owner" are one and the same state of affairs there.  (Neither the robust Haecceitist nor the Lewisian Concretist will be satisfied by this account, of course; the unexemplified haecceities and "nonexistent individuals" it provides are pale shadows or feeble imitations of real Haecceities and individuals.  That is as it should be, for an actualist.  All I claim is that the account is itself formally coherent and affords a tenable identifying criterion for unexemplified haecceities.)
     Regarding the fifth objection, the question of what advantage unexemplified haecceities have over bare Meinongian possibilia, I would point to the one key difference:  I have argued that the real trouble with Meinongian possibilia is that (a) they force us to disambiguate the existential quantifier as between actual existents and nonexistents, but that (b) at the same time we are allowed no expressible means of doing so; the Meinongian merely asserts that the quantifier is ambiguous and arrogantly declines to discuss the matter further.  By contrast, an unexemplified haecceity is an actual item, a property, which does not differ in ontological status from any other property despite being "blank"--it is part of the transmundane framework.  (What marks it as exemplified at a world where the "corresponding individual" exists is just the relevant existential proposition’s being an element of that world; thus I sidestep Adams’ objection.  Of course, these "haecceities" for nonactuals are not like those of real individuals, since they are not composed of the identity relation concatenated with the individuals themselves; they are watered down.<32>  That is to be expected from Ersatzers.)
    Regarding the sixth and final objection:  Here again we rely on Kripke’s seemingly justified penchant for stipulation of the identities of individuals at other worlds.  Which "Holmes" really is Holmes depends on our stipulation in the real world, provided that our stipulative act is connected in the right historical way with Conan Doyle’s writing of the Holmes stories.<33>  If our stipulative act is not connected in that way with Conan Doyle’s fictive act, then we cannot attach Holmes’ haecceity to either of the "Holmes"’s, and we are not describing a world containing the authentic Sherlock Holmes at all.  (Incidentally, the double-"Holmes"ian world is actually an embarrassment to qualitative essentialism regarding fictional characters, more so than to an Haecceitist view.  For essentialism implies that qualitatively identical individuals are one and the same individual.  Thus all there would be to distinguish two "Holmes"’s in the double-Doylean world would be their relations to each other’s planets, provided that the planets were themselves distinguished by containing actual individuals as well as the nonexistents proper to them.<34>  This may be coherent, but is not very attractive; and there is no way that I can see to decide which "Holmes" is Holmes--yet on the qualitative-essentialist view, at least one of them would have to count as Holmes, unless Conan Doyle had resourcefully avoided future counterexample by stating explicitly that there was no other planet containing a Holmes-replica.)
    Thus a Haecceitism for nonactuals can be sustained, if not loved.

    There are two further possible ways of accommodating modal properties of nonactuals.  Both involve denying that nonactuals can be literally identical across worlds.  One is to make all nonactuals worldbound, and import a counterpart relation to replace transworld "identification."  This is of course Lewis’ (1983) choice, and in effect Morton’s (1973) and McMichael’s (1983b);<35> each author suggests some sparse structural axioms governing the posited relation.  This goes against my constructivist grain; I could not in good conscience take something so crucial as primitive.  (And I have no idea how the relation might be defined without thereby giving rise to a qualitative essence; remember again my contention (fn 16) that iterated modalities pose just as nasty a problem for Lewis as they do for actualists.)
    The final option is to bite a very obdurate bullet and deny flatly that there is any way at all to identify nonactuals across worlds<36>--our characteristics would serve to identify individuals within worlds but never across worlds.  This would make our iterated modal predications essentially false: necessarily, Perry White could not have been a philosopher, nor Rose Kennedy’s fifth son a bricklayer, nor the person in the doorway somewhere else.  Period.  But I see no advantage of this option over the Conservative Position.

    As I said above, if we are to eschew unexemplified haecceities and unexplicated counterpart relations alike, we need to explain away the contrary Haecceitist intuition. What I think the Conservative theorist must say is that (some) people’s imaginative faculties are prone to supply definiteness and determinacy where there is none.  If the name "Smedley" is introduced as being that of a possible fat man in the doorway, one’s imaginative faculty may ignore the fact that no determinate genetic code nor any other true individuator has been supplied along with the bare description, and suppose that an individuator has been fixed even though we do not know it.  On this false supposition, we may then accordingly (seem to) attribute modal properties to the "individual."  But since there is in fact no individuator, the modal properties are unreal.

    It may be and should be wondered what my account has to say about Geach’s (1967) problem of "intentional identity."  Crudely, the problem is that a nonexistent individual may be the topic of each of two different propositional or other psychological attitudes--attitudes of the sort we would usually think of as de re, even though there is no actual res toward which the attitudes are directed.  Geach’s example is

            (1)  Hob thinks a witch has blighted Bob’s mare, and Nob thinks that she (the same witch) has lamed Cob’s sow.

Another example might be

            (2)  Gonzo fears there is a burglar at the door and hopes that that very burglar has not stepped on his turtle.

Each of these sentences is open to any number of interpretations, some of which are obviously unproblematic from anyone’s point of view.  But according to Geach there is a further sort of reading which resists being displayed  by any standard logical means: that on which both of the attitudes in question are directed toward nonspecific individuals, yet in some sense their foci are the same.  E.g., Gonzo fears that there is a burglar at the door (that is all he fears--there is neither any actual individual nor any particular pre-designated nonactual such as Raffles that he suspects of burgling); yet he hopes that that burglar has not stepped on his turtle, not just that there is no burglar who is at the door and has stepped on it, or that there is no burglar who he fears is at the door and has stepped on it, or the like.
    Geach’s problem seems to me cognate with Lewis’ and McMichaels’, in that it turns on the transworld identity of nonactuals who are not presumed to share their most salient properties.  It is another case of a nonactual individual introduced by one modality, which is said to have further modal properties (here, nonalethic de re attitudinal properties) in its own right:  (1) says that Hob thinks there to be a witch who did so-and-so, which witch is also thought by Nob to have yet another attribute; (2) says that Gonzo fears there to be a burglar, which burglar is also hoped by him to be a non-chelonicide.  The natural possible-worlds semantics for each sentence takes us to one or more worlds in which there is an individual of such-and-such a description (=/ any existing individual) and then to another world in which that same individual has some further property yet (perhaps) may not satisfy the previous description, thus forcing our earlier question of transworld identity for nonactuals.
    If Geach’s problem is cognate with Lewis’ and McMichael’s problem in this way, that would predict a corresponding intuitive reaction on our part.  In response to Lewis and McMichael the Conservative theorist doubted that the putative iterated modalities really had sense, on the grounds that (roughly) the initial hypotheses determine only generic individuals that are not nonqualitatively distinguishable; the individuals do not, except possibly by virtue of further imaginative and stipulative activity on our part, have any qualitative or modal reach beyond what is provided by the initial hypotheses.  The same is true of Geach-individuals, and so if we slam ourselves back into Conservative mode, we should expect to find ourselves with the same suspicion of senselessness.
    Prediction confirmed: I do find myself with that suspicion.  Once the various unproblematic readings have been set aside, I do not hear any natural or intuitive and coherent interpretation of (1) or (2).  (The same is argued by Dennett [1968].)  Indeed, I have more trouble hearing such an interpretation for examples of Geach’s sort than I do for McMichael-individuals.  And when we consider the matter in terms of possible worlds, an explanation of this difference in tractability readily presents itself:  McMichael’s examples begin with statements of possibility, which semantically are associated with existential quantifiers--at some world there is some individual which has such-and-such a further modal property.  But Geach’s begin with propositional-attitude verbs, which call for universal quantification over worlds (cf. Hintikka [1962])--at every world compatible with what Hob thinks, there is some individual which has the further attitudinal property.  In Geach’s case we are explicitly forced to envision a multiplicity of worlds, each inhabited by a witch, the various witches being quite different and presumably distinct from each other across the various worlds.  This makes it even harder intuitively to attach sense to the suggestion that Nob’s thought picks out any one of these different witches to the exclusion of the others, and so harder to understand how Nob could in any sense have in mind "the same" witch as Hob has.  But if nonactuals have haecceities, we should expect that Geachian sentences might have genuinely de re interpretations, the haecceities being there all along to serve as the attitudes’ rei.
    Thus an equivocal conflict between the Conservative Position and Haecceitism over Geach’s issue:  If one accepts the strong readings that Geach demands for his sentences, (a) the Conservative view is embarrassed by its lack of any obvious way of providing those readings, and (b) Haecceitism is strongly suggested in the way just noted.  But if for Hintikkan reasons one rejects the strong readings, then correspondingly Haecceitism is discouraged and Conservatism is confirmed.  And at the same time, pre-existing allegiance to the one policy or the other powerfully affects one’s pro- or antiGeach intuitions.
    But with only modest ingenuity, sense can be supplied to (1) and (2) on at least one more subtle Geachian understanding than the unproblematic ones.  For if Hob’s witch can be pegged to his own original mental act, then Nob’s witch may be pegged in turn to Hob’s witch in virtue of the (presumed) historical connection between his (Nob’s) mental act and Hob’s.  The trouble will be in getting Hob’s "witch" pegged in the first place, because of our semantical mandate that Hob has a multiplicity of equally appropriate witches, but a sufficiently clever and detailed Causal-Historical criterion might be able to surmount it and deliver a Haecceitist account even here.  At least, the strictly Conservative account can be avoided and Hob’s witch allowed her cross-modal properties.
    Yet even on a Causal-Historical view of things, full-bore Haecceitism is blunted in Geach’s cases by Hintikka’s point.  Look just at the first conjunct:  Hob thinks some witch or other has blighted Bob’s mare.  That indefinite NP does not create a haecceity for Nob’s thought then to latch onto, even if it does introduce a role-shaped world-line that (1)’s pronoun latches onto, nor does anything else in the conjunct create one; there is no singular reference in the clause, not even to a nonactual.  Perhaps, then, Geach’s examples do not militate either way as between Conservatism and Haecceitism.

 Edelberg (1986, forthcoming) has carried the "intentional identity" literature an important step further.  Consider his ingenious example:

...Arsky and Barsky jointly investigate the apparent (but not actual) murder of Smith, and they jointly conclude that he has been murdered by a single person, though they have no suspect in mind.  A few days later, they jointly investigate the apparent (but again not actual) murder of Jones, and together they conclude that Jones was also murdered by a single person.  At this point, however, a disagreement ensues.  Arsky thinks that the two murders are completely unrelated (and that the man who murdered Smith, but not the man who murdered Jones, is still in Chicago).  Barsky, however, thinks that one and the same person murdered both Smith and Jones.  (forthcoming, p. 26)

            (3)  Arsky thinks someone murdered Smith, and Barsky thinks he [the same person] murdered Jones.

But (4) seems false:

            (4)  Barsky thinks someone murdered Jones, and Arsky thinks he [the same person] murdered Smith.

Thus intentional identity (if we grant it is real) is asymmetric!
    Naturally, the Conservative Position cannot handle this case any more than it could Geach’s original one.  But as before, if one has qualitative-essentialist intuitions one will insist that (3) is true only on some of its deflationary interpretations and that the intended interpretation is incoherent.  And even if one does not have qualitative-essentialist intuitions, there is the problem of multiple murderers, different ones at Arsky’s various doxastic alternatives.  But suppose one is a Haecceitist and also finds (3) and (4) perfectly intelligible.
    For Haecceitists in Geach’s case, I suggested that a Causal-Historical theorist might peg Hob’s witch to Hob’s original mental act, and then (somehow) peg Nob’s witch in turn to Hob’s, even though that would not seem to create a full-fledged haecceity.  But Edelberg’s example is not so simple, for it adds the extra problem of asymmetry.  No single haecceity would do.  Even if we were able to latch Arsky’s belief onto a particular haecceity H, and then latch Barsky’s first belief also onto H (in order to make (3) true), Barsky’s second belief (reported in (2)) would presumably also latch onto H, and (4) would come out true rather than false.
    The only way to avoid this collapse would be to suppose that either Arsky or Barsky or both were doxastically related to each of two haecceities rather than just one.  For Barsky that suggestion is implausible, though I can imagine a theory according to which Barsky’s beliefs really were de distinct (nonactual) people even though with believers’ customary underprivileged access his internal representations of the two people were indistinguishable.  A two-haecceity hypothesis is initially more plausible for Arsky, since Arsky does think that two murderers are in the field.  Barsky’s nonactual suspect would then be identified with Arsky’s first murderer rather than Arsky’s second.  But then things go wrong:  If Barsky is related to just the one haecceity, then sentence (4) comes out true again.
    Notice that no Edelberg asymmetry can arise when the detectives’ beliefs are de real people.  Suppose Carsky believes de re of Giancana that he murdered Smith and Darsky believes similarly that Giancana murdered Jones.  Then sentence (2’) comes out true, not false.

            (2’)  Darsky thinks someone murdered Jones, and Carsky thinks he [the same person] murdered Smith.

When the actual objects of de re belief are identical, the identity is perforce symmetric.  What this seems to show is that neither Edelberg’s example nor probably Geach’s before it  is a straightforward case of de re belief, in part because they do not involve haecceities; the resumptive pronouns that figure in such cases should receive some less ambitious treatment.<37>  But I shall have to leave the matter at that.  So far as I can tell, Edelberg’s example is at worst neutral on our question of Haecceitism for nonactuals, which leaves that Haecceitism unharmed.
 
 

1 Shapiro and I originally called these individuating properties simply "essences," but in view of our earlier distinctions that usage might cause confusion here.

2 N.b., it does not follow that P is a characteristic in w’.

3 Rescher (1975, Ch. 3) elaborates and defends a similar idea of "supernumerary" individuals.

4 For similar and more extended defenses of an inegalitarian attitude towards the individuation of nonexistents, see Kaplan (1973, 1975), Rescher (1975, Chs. III and IV), Kripke (1972), and Plantinga (1974); Robert Howell questions this argument, albeit programmatically and somewhat obscurely, in (1979, sec. 7).  Somewhat different defenses may be found in Skyrms (1981) and in Adams (1981, sec. 2).

5 Pace Howell (1979), Pollock (1980), Routley (1980), Lewis (1983), and others; see the reply to objection 5.5 in section 5 of Lycan and Shapiro (1986).

6 Their concern is glancingly anticipated by Morton (1973).

7 Actually the inclusion of fictional characters as nonexistents and as inhabitants of possible worlds is neither unproblematic nor uncontroversial; see below.

8 I am grateful to Dick Smyth, Kate Elgin and Ken Taylor for insistent criticism on the present point.

9 In another sense, obviously, the constitutents of novels and stories are sentences, words, letters or the like.

10 By "fictional" here I mean merely fictional; for simplicity I ignore individuals who are mentioned in fiction but are also actual, and I even more emphatically ignore the hybrid people found in romans à clef.

11 See also Castañeda (1979), Howell (1979), Parsons (1980), Routley (1980), Fine (1982), Bertolet (1984a, 1984b), and all the multifarious works further cited in these.

12 This is not intended as a question about what Mr. Jerry Siegel (the author of Superman) might have chosen to do with the character, as when we ask whether he might have made White a health nut.  Cf. my distinction between fictional characters and fictional people.

13 This observation strikes me as intuitively and strictly correct, but I admit I have encountered some hardy bibliophiles who maintain that Kent truly and literally works at the Daily Planet and doubles as Superman and so on.

14 For key refinements, see Lewis (1978).

15 McMichael offers two more subtle arguments against pinning nonexistents of his sort to "qualitative essences":  that given a suitably bifurcated and symmetric world, any alleged qualitative essence might fail to discriminate an inhabitant from its doppelgänger, and that two nonactual worlds might be connected by a chain of individually small changes that gradually switch two characters’ properties (cf. Chisholm’s (1967) original "Adam"/"Noah" case as discussed in the previous chapter, though McMichael’s own argument relies on less aggressive assumptions).  I think both arguments can be resisted, by enforcement of the alternative view of nonactual-individuation that I shall sketch below, so I shall not pursue them here.
    N.b., it is also open to a granite-jawed Actualist to deny after all the truth of McMichael’s iterated modal sentences on the grounds that they would make good sense only on a strong version of realism about possibilia.  One might imaginatively stipulate that Rose Kennedy had a philosopher son in addition to the ones we know, and count it true-in-the-stipulated-scenario that the son might have gone into bricklaying, but refuse to give a standard possible-worlds semantics in turn for truth-in-a-scenario and refuse to grant it true simpliciter that the imaginary son might (literally) have gone into bricklaying.

16 Incidentally, McMichael presents his troublesome argument as a problem for Actualism specifically, as if the Meinongian/Lewisian "possibilist" were not to be troubled by it.  Initially we can see why:  it is the Ersatzer or other Actualist like me who denies that nonactual individuals are in propria persona constituents of other worlds, and so has trouble accounting for the transworld identity of nonactuals.  But I do not see why McMichael’s problem does not upon examination infect "possibilist" semantics as well.  Consider Lewis’ Mature Concretism.  Superficially that view offers a straightforward account of singular iterated modalities:  Perry White might have gone into nursing iff he has a counterpart at some other world (still distinct from @) who does go into nursing; the counterpart is a flesh-and-blood individual similar to White, just as robust a constituent of his or her own world as White is of his, and so there is no embarrassing ontological asymmetry of the Conservative sort.  But, I contend, McMichael’s dilemma rears its head when we ask what counterpart or similarity relation is operative.  Not an effable, qualitative relation, since to set up the dilemma at all we need the assumption that White could have lacked any of the features conventionally associated with him by the stories.  But presumably not a nonqualitative, sui generis property either, for this would be (as Lewis himself agrees on p. 229 of [1986]) to leave a mystery as the sole ground of modalities de nonactuals.  Thus I do not see that as regards McMichael’s problem Lewis is any better off than I am.

17 Shapiro and I (1986) provided a harmless corresponding variation on our original formalism.
    I am of course neglecting an option:  inaccessible qualitative essences for nonactuals.  Someone might argue that if actual individuals have inaccessible essences, then by analogy nonactuals do as well; e.g., a nonactual human being might have its genetic code as its essence.  But this would create a nasty problem for the case of fictional individuals at least (anticipated by Kripke [1972/1980, 1972b]):  There are countless worlds containing "Sherlock Holmes" figures, each of whose "Holmes" is realized by a genetically different individual.  But according to inaccessible-essentialism, at most one of those individuals would be the real Holmes, to whom we refer in using the name, for given that one is the real Holmes, the others must be distinct though very similar people who are called by the same name.  Yet the choice of one real Holmes from among the multitude would be arbitrary; it does not seem that any of the realizers would be the real Holmes to the exclusion of the others.

18 Robert Howell (1979, sec. 7) protests that specification by description is not the only alternative to introduction by ostension; but his positive hint of a tertium quid is so far very obscure.  (It seems structurally similar to Lewis’ [1983] method of perceiver-relative counterparts.)

19 Currie (1986) makes the point in terms of Ramsey sentences, claiming that grasping a Ramsification of a story is "all there is to understanding" the story.
    Coburn (1986) asks why, if Haecceitism holds for actuals, we cannot imagine a world qualitatively indistinguishable from @ but containing entirely different individuals and in addition a second qualitatively indistinguishable world containing still a new, third set of distinct individuals (which seems absurd).  The answer is that if the Conservative Position is correct, this absurdity is just what we should expect, but if Haecceitism regarding nonactuals should turn out to be true, Coburn’s hypothesis is not absurd after all.

20 Kraut (1979, p. 213) notes the peculiarity of someone’s (truly) claiming to want Pegasus but denying that he would be satisfied by just any winged horse that was ridden by Bellerophon etc. etc.

21 Note the ambiguity of "unexemplified" here as between "not instantiated at @," "not instantiated at such-and-such a world in which it exists," and "not instantiated at any world."  In the absence of qualification I shall always mean the first of these.

22 This final objection was put to me by Greg Currie.

23 Lewis (1983, pp. 22-23; 1986, sec. 4.2) poses a seventh:  that he can make no sense of the idea of an object’s existing in its own person at each of two different worlds and nonetheless having properties at one that it does not have at the other.  (He rejects the analogy of a thing’s having a property at one time but not at another.)  This intuition of bad craziness is of course the same as that which has always driven Lewis to counterpart theory as opposed to literal transworld identity; and it enables him to return at least some of the "incredulous stares" he is accustomed to receiving from complacent actualists.  For my part, as we saw in Chapters 2 and 4, I have no difficulty at all with the notion of having properties relative to worlds; I find the time analogy entirely supportive here, even if in other respects it is imperfect.

24 For McMichael’s own semantics see again (1983b).

25 A similar suggestion was made by Carney (1977).  Forbes (1986) takes this idea seriously, but does not end up a Haecceitist regarding nonactuals; he favors an analogue of the Searle-Wiggins-Armstrong view I criticized in Chapter 5.  I think this is in part because his conception of a fictional being hovers somewhere between van Inwagen’s notion of a fictional character and what I have called a fictional individual or person as opposed to a character.

26 Castañeda (1979) discusses an interesting actual case of this type.

27 Obviously there are vexing borderline cases here.  I am inclined not to treat pastiches in the way described; the character called "B*nd" in Christopher Cerf’s and Michael Frith’s Ian Fleming parody, Alligator (Harvard Lampoon, 1962), is not James Bond even though he is directly inspired by Fleming’s creation.  For that matter I do not countenance the protagonist of John Gardner’s recent, comparatively respectful "Bond" books, for the reason that Gardner’s hero is very distinctly Hollywood Bond--just good fun--as opposed to Fleming’s own very serious character.  (Mr. Fleming presciently made this distinction himself, in The Man with the Golden Gun, when Bond, amnesic, is being vetted for authenticity by members of his own Service:  his suspiciously new clothing and his choice of the Ritz hotel, both in fact dictated by his KGB brainwashers, are described by the Chief of Staff as "sort of stage Bond.")  I am more amenable to the Bond of Kingsley Amis (in Colonel Sun, written under the pseudonym "Robert Markham"), who seems totally genuine save for the interjection of one or two unmistakable and presumably irrepressible Amisisms.  The difference in identity between Gardner’s and Amis’/Fleming’s character is due, I surmise, to differences in the authors’ respective metalinguistic intentions, but the latter would be very hard to spell out.

28 It may seem that my own case (presented earlier) of the absence of the editor figure and the presence of an entirely different character who happens to be named "Perry White" is a counterexample, since the distinctness of the two seems to consist in the difference between their descriptive roles in the story.  This again depends on Mr. Siegel’s intentions.  If he had never created the editor figure in the first place, his alternative invention of the janitor or itinerant bum would have been simply a different fictive act, and so would have had a different character as its issue in any case.  If, on the other hand, Siegel had written in the editor figure to begin with but (for some peculiar reason) then decided that White would not have been like that but would have pursued a very different sort of life-plan eventuating in a janitor’s job or in vagrancy, the new White would count as the same character as the old despite his drastic transformation.

29 As McMichael has pointed out in conversation, this requires that we conceive of fictional characters and worlds as being created by authors rather than as having existed from time immemorial.  That is all right with me (especially since I have no great stake in the Haecceitist view I am now adumbrating); the view takes fictional characters to be something like fictive acts with qualitative stereotypes appended.

30 Fine (1982) relies heavily on the author’s "sayso" in the individuation of fictional characters.

31 I am not so sure about Robert Coburn’s (1986) question of whether God could have had a choice as between implementing Alvin Plantinga’s haecceity and a putatively distinct haecceity ("?x[x = Plantinga*]") that is not qualitatively distinguished from Plantinga’s.  Nasty theological questions and questions of Divine reference-fixing mingle here.  I suspect, on behalf of my developing Causal-Historical theory, that our current use of the name "Plantinga*" fails to connect in the right way with any bonafide dubbing on God’s part; but the matter needs considerable investigation.

32 All one can say is:  They are properties; they are nonqualitative; they are unexemplified; like the haecceities of actuals they are what track individuals through radical change from world to world; and their epistemology too goes by causal-historical chains.  Do they indeed exist?  Well, at a world distinct from ours, something has the property of being Holmes, and nothing has that property here.

33 It is not clear whether we should count qualitatively identical double-Doyle worlds as distinct if in one "Holmes" #1 is (stipulated to be) the real Holmes while in the other "Holmes" #2 is.  That depends on how we have antecedently distinguished #1 from #2, independently of the question of Homes’ haecceity; I suspect there is no ultimately coherent way of doing so.

34 That is not strictly right, as Currie has also observed to me; if one of the two planets contains actual individuals, such as Baker Street or Afghanistan, then that planet would be the one described by Conan Doyle.  To secure the point we would have to take as our example an entirely fictional, fantastical work that made no reference to any actual item.

35 McMichael takes as primitive an accessibility relation among maximally consistent "roles."

36 This is Adams’ (1981) view.

37 Edelberg himself uses the example to motivate working with theory-immanent notions of truth and reference rather than the standard "realist" word-world notions.