Itmight come as a surprise that the majority of those who work on the logic and metaphysics of fictional objects and names—objects like Sherlock Holmes and names like “Sherlock Holmes”— acknowledge the existence, in some sense, of those objects and the (existential) referential character of those names.1placeholder “In some sense” because, of course, it is also the case that Sherlock Holmes does not exist and “Sherlock Holmes” refers to nothing existent. But then, one asks, how is it possible to account for sentences like
(1) Sherlock Holmes plays the violin
or, for that matter,
(2) Sherlock Holmes is a fictional object?
(1) and (2) appear to be informative, indeed true, and how can one explain it? With (1), one might argue that it is informative and true in, or of, a possible world distinct from the actual one: the world described in the novels and short stories by Arthur Conan Doyle. But (2) is unmistakably spoken in, or of, the actual world, where the extension of the predicate fictional object is empty, as there is no such (existent) thing, and there is no such thing as Sherlock Holmes; so, for all intents and purposes, saying (2) would appear to be tantamount to saying
(3) Awth is ghgh,
where, equally, there is (in the actual world) no such thing as Awth and nothing belongs to the extension of the predicate ghgh. Despite this superficial (and paradoxical) analogy, however, (2) continues to be informative and true in the actual world whereas (3) is not, hence there must be a way in which the object Sherlock Holmes and the predicate fictional object (but not the object Awth and the predicate ghgh) contribute to the actual informational content, and truth, of (2); and what could that be if not that Sherlock Holmes, somehow, belongs to the actual world and fictional object picks out in it objects like him and not, say, like Napoleon?
Such has been the state of the art for the last several decades, when no significant progress has been made through countless papers and books and admirable displays of ingenuity. The reason for this dispiriting situation is to be found in the fact that the whole debate is not conducted at the appropriate level of generality. What is at stake here, in one of many incarnations of the fundamental (logical and metaphysical) issue of how thought relates to being (or, in a linguistic jargon, how being relates to language), is the choice between transcendental realism and transcendental idealism (from now on, respectively, TR and TI). I will say more about these two stances in what follows; for the moment, let me preface my discussion with a simple diagnosis. The whole debate I hinted at is located within TR and, as Kant says in the first Critique, “It is really this transcendental realist who afterwards plays the empirical idealist.”2placeholder An empirical realist is someone who lives her ordinary experience in a world of ordinary existent objects—tables, chairs, trees, and (existent) detectives. An empirical idealist is someone who lives that experience in a world of ideal objects: Berkeleyan ideas or fictional detectives. And, says Kant, one is forced into such an absurd conclusion by taking the wrong transcendental stance. Being committed to TI, I am an empirical realist: I believe that there is no Sherlock Holmes, that there are no fictional objects, and that (1) and (2) are true.
I have articulated my understanding of TR and TI in great detail elsewhere;3placeholder here I will provide as much as is needed for present purposes. TR and TI have nothing to say about the world, actual or otherwise: their subject-matter is the conceptual, or logical, space that is used in thinking about the world—or anything else. TR is, essentially, Aristotle’s position: in its structuring of logical space the basic concept is that of a substance (or object, or thing), and every other concept is either dependent on it (as, say, quality and quantity are—they are quality and quantity of a substance) or defined starting with it (as, say, human is defined as rational self-moving animate material substance). Objects (from now on I will use this term for them, as being more popular nowadays) are also what exists primarily: if an object ceases to exist, its qualities automatically cease to exist, too—they are ontologically, as well as logically, dependent on it. Hence a nonexistent object is a contradiction in terms and, in a famous quote by Bertrand Russell, so would be a name that purported to refer to it. Speaking of the (alleged) name “Romulus,” Russell says:
“If it were really a name, the question of existence could not arise, because a name has got to name something or it is not a name, and if there is no such person as Romulus there cannot be a name for that person who is not there.”4placeholder
Where people went from this statement, in the half-century or so during which Russell’s view on the topic was the dominating one, was to agree with him that “Romulus” (or even “Bertrand Russell”) was not a name at all, indeed not even an independent component of the sentences to which it seemed to belong, though not many were prepared to follow him in the more radical claim that Romulus (in case there existed such a person, or Bertrand Russell) was not an object but a logical construction out of sense-data (or sensibilia). For this view, at any rate, sentences like (1) or (2) were not such a big problem, as they could be tamed by rephrasing them in terms of Russell’s definite-description theory and thus exempted from any mention of objects—predicates were all that was required.
But those were the old days, b. K. (before Kripke); then we were instructed (by the appropriate supernatural revelation) that proper names directly refer to their objects, in fact directly refer to one and the same object in all possible worlds. And, with this new view, the problem was out there, and was a serious one. For “Sherlock Holmes” is a name, and according to Kripkean gospel it must directly refer to the (same) object Sherlock Holmes in a variety of worlds, including the one(s) in which Conan Doyle’s fiction is true. It might, of course, refer to nothing in the actual world, but that brings us back to the beginning of this essay: how then can we account for the actual truth of (2)? Perhaps by distinguishing various senses of being: a stronger existence and something weaker? Perhaps Meinong was right after all?
Here is where I leave TR, which is ultimately no concern of mine. I find it a worthless stance; but I also find it worthless to write “philosophy” by pointing out obvious blunders of what is characterized as one’s opposition. For the remainder of what I am to say, I am going to turn to a constructive task: to showing how the problem is resolved in TI.
The basic concept in TI’s logical space is that of a representation (German Vorstellung, which would be better translated as presentation, as no recurring or repetition is implied). A representation could be thought of, to begin with, as what in other (phenomenological) lines of thought is called an intentional state, in which something intends something. Examples would be:
(4) I see a brown table
(5) I hear the note A
(6) I think of Pegasus
(7) I read of Sherlock Holmes in A Study in Scarlet.
What is intended is the object of the representation, and what intends is its subject; consistently with Kant’s teaching, the subject here is always designated as “I”—the pronoun “I,” for him, must be able to accompany all representations. But the use of this pronoun must not be taken to entail reference to any entity, let alone to me, nor should we think of the uses of “brown table” and the like as having similar entailments. Both the intending I and the intended objects are, at this stage, only parameters of a representation, much like a quality of a substance is but a parameter of it in the Aristotelian framework: no substance, no quality there; no representation, no intended subject or object here. Otherwise put, both “I” and “brown table” occur here in a highly opaque context, from which they cannot (initially) be detached: just as looking for a dog, in Quine’s notional sense, is a kind of looking and makes no suggestion about the existence of a dog being looked for, here seeing a brown table is a kind of representation and makes no suggestion about the existence of a brown table being seen, or of anyone seeing it.
Representations aggregate in clusters: for example,
(8) I see a brown table as the clock shows 1pm
(9) I see a brown table as the clock shows 1:01pm
(10) I see a brown table as the clock shows 1:02pm.
And clusters of representations (not individual representations) are tested by way of categories, that is, conceptual criteria of objectivity. To the extent that a cluster of representations is (to cite the most uncontroversial such criteria) consistent, causally connected, and spatiotemporally determined, its members will be objective and the objects represented—which, before the test, have no independent status—will be objects, period, ready to be detached from their representations in what would thus become transparent contexts.
I said “to the extent that” for good reason. Ideally, one would want (reason would want) this operation to go all the way and establish the definitive objectivity of some represented objects. But such an outcome is impossible. The Antinomies show that establishing the definitive objectivity of even a single represented object would require extending our analysis to the entire world; and, when we apply categorial criteria to a system of representations fit to represent the entire world, we run into contradictions. So we have, in Kant’s celebrated word of caution, “to deny knowledge”5placeholder and recognize that objectivity can only ever be established to a point, that all the represented objects to which we provisionally apply this qualification run the risk of collapsing against the next hurdle, that thoroughly objective things (“things in themselves”) are noumena (matter of pure thought), and that what we deal with in ordinary experience are phenomena—fragile, defective presences for which no warranty can be issued that tomorrow, or the next minute, they will not prove inconsistent or otherwise categorially inadequate. We will still patiently make the effort of detaching them from their contexts and will operate with them with as much independence as we can grant them, but knowing that those contexts could always suddenly swallow them into utter opaqueness.
This is as far as I need to go here into the general structure of TI. Let us now draw some consequences from it relative to our current concerns. First, there is no difference in principle between Bertrand Russell (or the table now in front of me) and Sherlock Holmes. Bertrand Russell surfaces in representations like
(11) I read of Bertrand Russell in The Philosophy of Logical Atomism
(and the brown table now in front of me in representations like (4), (8)-(10)), and Sherlock Holmes surfaces in representations like (7); then I start applying categorial criteria to them and I realize that (11) coalesces promisingly with a cluster as much as (7) coalesces promisingly with another one. One of these clusters may be favored over the other (more about this later); but neither of them will ever reach a definitive stamp of approval—both are bound to remain in the provisional, uncertain status proper to phenomena at large. Still, a number of what are called in the relevant literature metafictional sentences can be read, and judged true (with all due circumspection that pertains to this framework), no differently than sentences dealing with more garden-variety lore. Based on an analysis of the clusters where they belong,
(12) There are three characters in English literature who kill their mothers
(13) There are characters in 18th-century novels who explore unknown territories and characters in 20th-century novels who face extraordinary circumstances close to home
offer no greater complications to being read and evaluated than do, based on an analysis of the clusters in which they belong,
(14) There are three people where I grew up who were orphans before the age of ten
(15) There are people where I grew up who ended up living abroad and people where my wife grew up who are still living in their family home.
And notice: what I just said requires no appeal to possible worlds distinct from the actual one: to the world, say (or worlds), of 18th-century novels. In the world in which I live, the one and only actual world (which is, in this framework, never fully defined), I deal with Robinson Crusoe and Lemuel Gulliver, with Leopold Bloom and Philip Marlowe, as much as with Louis XVI of France and George Washington, with Charles Lindbergh and Charles De Gaulle: all of them have surfaced in my experience through reading books, and for all of them I have tried (more or less successfully) to weave a network of connections that could lead me to attributing to them some definite predicates and building some definite expectations about them—to moving them from utterly opaque to (partially and provisionally) transparent contexts.
The second point to be made is that some of the predicates I attribute to objects in the manner I just described are what I could call local ones: attributing them depends on the consideration of a limited amount of the representations in which they surface. One need not go very far into the study of an object and its context to decide that it is an orphan, or went to America, or is the tallest member of its family. (1) attributes a local quality to Sherlock Holmes, and
(16) Sherlock Holmes is a friend of Dr. Watson
attributes a local relation to him;
(17) Pierluigi had dark hair
(18) Pierluigi was Ermanno’s best friend at the age of fourteen
attribute a local quality and a local relation, respectively, to Pierluigi. Other predicates I call global ones: attributing them depends at least on consideration of the whole context of representations in which the relevant objects surface, and may depend on consideration of the whole context of available representations. Such would be the case for sentences like
(19) Pierluigi is a totally undefined person past the age of 20
(20) Of all people, Pierluigi is the one with the starkest contrast between his definiteness up to a certain age and his total indefiniteness past that age.
Cases of this sort that would belong to fiction (but remember: I have not yet drawn any distinction between what is fiction and what is not) would be sentences like
(21) Godot (in Waiting for Godot) is a totally undefined character
(22) Of all characters in fiction, Godot is the one that is at the same time most significant and most undefined.
One might believe that in moving from local to global predicates we are shifting to a different language level (say, a metalinguistic one); but that would be a mistake. Any definiteness we assign to an object (and I said that it is necessarily going to be partial) will depend on drawing conclusions from what information is provided by clusters of representations; it is just that sometimes it will be enough to consider a limited cluster while other times we will need to bring out all the information we have—about that object or altogether. And any result of our analysis will always be couched in perfectly ordinary, object-language sentences like (1), (16)-(22).
Move to the third step. Sometimes a cluster of representations fails categorial testing. One of them, for example, might include
(23) I saw a pink elephant in a corner of my study at 1pm
(24) I saw no pink elephant in a corner of my study at 1:01pm,
where it would be causally impossible to move an elephant out of my room in one minute—or to let it in at all. Another one might include
(25) I spoke with my father a little while ago
(26) I am aware that my father has been dead for over twenty years.
In the Transcendental Deduction of the first Critique (a more vastly misunderstood section of that work than most) Kant argues that a categorially adequate, hence also unitary, world goes hand in hand with a unitary experience, hence also with the unity of the subject of that experience—a subject that would then no longer be just a parameter of representations but some kind of entity patienty detached from them by categorial work, just like objects are. For reasons already voiced, no definitive unity will ever be gained for any of this: it will always be of the “to the extent that” type. Leaving that point aside, a failure of categorial testing clearly goes in the opposite direction, and presses for an explosion of the world into chaos and of the subject into schizophrenia.
If this devastating upshot is not forthcoming as often as it might threaten, it is because there are devices available for saving adequacy. Simplistic devices for saving superficial adequacy; but most of the time they will do. We could, for example, blame (23) on a hallucination and (25) on a dream, which means: we could isolate clusters to which (23) and (25) belong and, while we might enjoy continued exploration of them, take them to be constitutionally powerless to challenge those other clusters to which (24) and (26) belong. Predicates like existent and nonexistent are supposed to track the application of such devices. Within a limited cluster, a represented object could satisfy categorial testing, and be a credible candidate for objectivity; but then the cluster itself could become suspicious, and be rejected; then objectivity would be denied for it, and the object(s) represented in it labeled nonexistent. Therefore existent and nonexistent are global predicates, whose attribution depends on considering the whole cluster of representations currently available. There is no way that this whole cluster can get successfully through categorial testing—(23)-(26) are good evidence of that—so what is needed is the judicious work of isolating offending subclusters in order to leave a core that one can hope will pass muster: existent and nonexistent are the tags we attach to objects as we perform such work.6placeholder
I have had anecdotal exposure, in literature and film, to cultures that don’t do that, and take dreams and hallucinations at face value, but I have no idea of what their subjects look and feel like: I don’t know enough about them to judge. What I know is that, in the culture(s) I am familiar with, the trick I described is one that most people become conversant with quite early, though perhaps not immediately: I can imagine that infants and small children do not manage it as easily. Making full use of the trick, I can reach conclusions like the following:
(27) The pink elephant I saw in a corner of my study at 1pm was a hallucination, hence nonexistent
(28) The father of mine I spoke with a while ago was part of a dream, hence nonexistent.
We are finally ready for the entrance of the quality fictional—and you can figure how, in the absence of the hard conceptual work described above, handling that quality philosophically would be like doing handiwork without thumbs, or with thumbs only. Fictional (which, according to its Latinate etymology, means made up) is one more global quality, like hallucinatory or part of a dream, that is used to isolate portions of experience in order to prevent it from infecting the rest with contradictions—by making what surfaces in those portions nonexistent. Many fans of the notorious detective, for example, might have been led to assign to him the local predicate occurring in the sentence
(29) Sherlock Holmes lived at 221B Baker Street,
only to find out, when they visited that area of London, that there had never been (before 1990) that street number in that street; so, eventually, if they wanted to maintain the categorial adequacy of the larger portion of their experience, they had to isolate the cluster relevant to Conan Doyle’s novels and short stories and accept the truth of (2). It may be the case that small children are not yet very skilled at performing this operation (I have personal memories suggesting that); but surely most adults are able, based on it, to enjoy their fiction (even more than their hallucinations or dreams) without suffering any serious schism in their personality.
Along these lines, all the other metafictional sentences that seem to create difficulties in professional dealings of such matters (because of how committed those dealings are to TR) can be straightforwardly understood and evaluated. A sentence like
(30) There are characters in some 19th-century novels who are presented with a greater wealth of physical detail than is any character in any 18th-century novel
brings out a global relation, which requires considering (not the whole cluster of available representations, but) the whole cluster of representations relevant to 18th-century novels. A sentence like
(31) None of the towns and villages described in The Lord of the Rings trilogy actually exists
is no more troubling (or interesting), if not for the number of objects involved, than (2). And a sentence like
(32) Some characters in War and Peace (like Napoleon) are existent ones and some others (like Pierre Bezukhov) are not
does not go much further, though an intriguing epicycle of it would be if we discovered that Tolstoy’s Napoleon differed in some important respects from the Napoleon of history books—then, to maintain consistency, we would have to break Napoleon in two and let Tolstoy’s Napoleon go the way of fiction.
The reason why the problem here, after the hard conceptual work that was described, is not so much of a problem should be clear by now: in TI all represented objects you start with are to be treated uniformly, independently of whether or not, based on the evidence available, we decide that, in order to avoid trouble, some of them need to be cordoned off (that is, judged nonexistent—no two ways about it; no fancy distinctions of kinds of existence, or being; I am, I said, an empirical, or commonsensical, realist, for whom there is simply no such thing as Sherlock Holmes). In TR, on the other hand, you start with a stock of objects which are “what they are and not another thing” (to quote Bishop Butler) and, the moment you start looking into candidates for objectivity that do not present such inflexibly solid credentials, you don’t quite know what to say about them. In closing, I will mention another side of this coin, which might vex some though I find it exhilarating: philosophical temperaments are, undeniably, heavily disparate.
Because of the provisional nature that objectivity can only ever have for “objects” in TI, it is always to be taken as possible that the predicates existent and fictional (or hallucinatory, or part of a dream) we judiciously apportion around in the unending construction of the phenomenal world might have to be reversed at some point in the future (that what counts as common sense might be reversed), and War and Peace (or Waiting for Godot) might become a more veridical description of how things are than anything that has received that eminent stature today. Not likely? Maybe. But philosophy does not deal with what is likely: possibility is its range, and there is no denying possibility here. Which may be one way of reading the recurrent claim that there is some truth to art, even more than to newspaper reports.
I thank Louis Rouillé for bringing up this issue with me in extensive and engaging discussions. Among the authors and texts alluded to here, to whom/which also he directed my attention, I will mention Catharine Abell’s Fiction: A Philosophical Analysis (Oxford: Oxford University Press, 2020), Gregory Curry’s The Nature of Fiction (Cambridge: Cambridge University Press, 1990), Nathan Salmon’s “Nonexistence” (Noûs 32, 1998, pp. 277-319), Amie Thomasson’s Fiction and Metaphysics (Cambridge: Cambridge University Press, 1998), and Peter van Inwagen’s “Creatures of Fiction” (American Philosophical Quarterly 14, 1977, 299-308)
A369. Translated by Paul Guyer and Allen W. Wood (Cambridge: Cambridge University Press, 1998), p. 426.
Most explicitly in Kant’s Copernican Revolution (New York: Oxford University Press, 1987). Chapter 2 of my subsequent Ethics Vindicated: Kant’s Transcendental Legitimation of Moral Discourse (New York: Oxford University Press, 2007) provides a brief summary of that account.
The Philosophy of Logical Atomism (London: Routledge, 2010), p. 79.
Critique of Pure Reason Bxxx, p. 117 of Guyer and Wood’s translation.
This is how I read the often-quoted, but not well-grasped, claim in the first Critique that “Being is obviously not a real predicate, i.e., a concept of something that could add to the concept of a thing” (A598/B626, p. 567 of Guyer and Wood’s translation). The object of a representation is something like a brown table or a winged horse, identified through concepts like brown, table, winged, and horse because intuitions without concepts are blind (without concepts, nothing gets represented). Existent (or being) does not belong there: it is only after properly reviewing the entire cluster of available representations that I can decide (always provisionally, of course, as the cluster can expand and contradict my previous conclusions) that I saw an existent brown table but a nonexistent winged horse. In Kant’s words, being “is merely the positing of a thing or of certain determinations in themselves” (ibid.). He does not deny that certain objects exist or do not exist, hence that, logically speaking, being is a predicate, but wants to be clear about what kind of predicate it is.