Why Realisms about Fiction Must (and Can) Accommodate Fictional Properties

: The topic of ﬁctional objects is a familiar one, the topic of ﬁctional properties less so. But it deserves its own place in the philosophy of ﬁction, if only because ﬁctional properties have such a prominent role to play in science ﬁction and fantasy. What, then, are ﬁctional properties and how does their apparent unreality relate to the unreality of ﬁctional objects? The present paper explores these questions in the light of familiar debates about the nature of ﬁctional objects.


Introduction: Fictional Predicates
There are predicates that originate in fiction and whose meaning seems to rule out their possible application to ordinary individuals.This is most evident in fantasy and science fiction, which are the genres of fiction we have primarily in mind in this paper.Tolkien's Bilbo Baggins is a hobbit, while in The Lord of the Rings, a palantír is an indestructible ball of crystal used for communication and to see events in other parts of Earth [1,2].There are no true predications of these predicates of ordinary objects.No ordinary object is a hobbit or a palantír.Indeed, no ordinary object could satisfy these predicates. 1 Being small in stature, living in holes in the ground, and having long hairy feet, for example, is not enough to be classed as a hobbit.'Hobbit' names a humanoid species or "race" introduced in the novels by J.R.R. Tolkien, and there is much more to a species than having certain superficial features.The biological features that make someone a hobbit, like the features that make something a palantír, exist only from the point of view of the relevant works of fiction, and these works are quiet about such kind-making features.In that sense, nothing could possibly be a palantír or a hobbit. 2  It is common to suppose that we understand predicates by grasping the properties they denote.If so, we should say that competent, engaged readers and viewers grasp such properties as being a hobbit.But how could this be?For the reasons given, they only exist as properties from the point of view of the relevant works of fiction.Outside of the works, there are no conditions X, not even vague conditions, that determine when something is or is not a palantír or a hobbit, so outside of the works there are no properties designated by the fictional predicates 'is a palantír' or 'is a hobbit'.That suggests that "properties" of this kind are, in an important sense, not real.If there really are such things, they are merely fictional properties, and just as a fictional detective is not a detective, so fictional properties are not properties.
But in that case, what are they? 3 If they aren't really properties outside of the stories themselves, we surely have a problem.How are we to understand our engagement with works of science fiction and fantasy in that case?And how can we account for the manifest truth of statements involving works of this kind, such as (i) 'In the Tolkien stories, Bilbo Baggins is a hobbit', (ii) 'Hobbits don't exist', (iii) 'Hobbits are merely fictional creatures', and so on?All of these statements are truths that feature the fictional property of being a hobbit, whatever that might be.
The "engagement" feature of the role of fictional properties is so important that it should serve as a constraint on accounts of their nature and role; call it the Engagement Constraint.No less important is the fact that the referents of fictional predicates must be sufficiently property-like (despite not existing as properties) to fulfil the semantic role of helping to account for the truth of sentences like (i)-(iii) above.Call this the Quasi-Property Constraint.
The present paper explores the nature of fictional properties in the light of the centrality of the Engagement and Quasi-Property Constraints, familiar debates about the nature of fictional objects, and recent work on the idea of fictional properties, focusing, in the main, on realist theories because of their prominence in the literature.The paper's organization is as follows.After a quick tour of realist theories of the nature of fictional objects, Section 2 describes some apparent problems that fictional properties pose for these theories, problems that have been taken to be fatal for such theories in an important recent paper by Sarah Sawyer [6].In Section 3, we show that these problems are illusory.Contrary to Sawyer, each of the theories (with one possible exception) has the resources to make sense of fictional properties, although we also argue that each theory thereby inherits a version of the problem that affects, or is commonly thought to affect, the theory's corresponding account of fictional objects.Section 4 measures the theories up against our two constraints and tentatively concludes that while all of these theories meet them more or less well, they do worse in one way, and better in another, than Sawyer's own preferred theory, which is an anti-realist pretence theory.

Fictional Objects
Since the 1970s, there has been an explosion of work on the ontology and metaphysics of fiction, with the focus of most of this work being fictional objects or entities (or ficta, as in [7]).Taking fictional realism, both about fictional objects and fictional properties, as our focus in this paper, we take realism about fictional objects as falling into two main camps (for a survey, see [8]): (1) Meinongian theories according to which fictional objects lack existence.These include the two-kinds-of-properties Meinongianism of Routley and Parsons (2KP Meinongianism, for short) and the more recent Modal Meinongianism of Priest and Berto [9][10][11][12]; and (2) Abstract object theories according to which fictional objects have full-blooded existence as abstract objects of some kind (this includes Zalta's metaphysically oriented Object Theory, often interpreted as a version of Meinongianism, and the Artefactualism of Kripke, Thomasson, and Voltolini (among many others), on which fictional objects are the product of a creative process of some kind [4,7,[13][14][15]. 4  As an aid to readers who may be more familiar with some of these theories than others, here is a slightly more detailed account.
(1a) For 2KP Meinongianism, fictional objects have both nuclear and extranuclear properties, where the former are properties that they have in the works in which they appear, and the latter are properties that also depend on their relationship to the real world.For example, Sherlock Holmes [17] has the nuclear properties of being a detective and living for much of his working life at 221B Baker St., etc., while also having the extranuclear properties of being a fictional character, being among the most famous literary characters ever created, and lacking existence.(Since he exists from the point of view of the work, he also has a weak nuclear property of existing, but this is not the property we have in mind when we deny his existence.An object O exists just when there exists an actual entity that has precisely the properties of O.An existent entity has either property P or non-P for any property P that can be intelligibly predicated of it, and since fictional objects are always incompletely characterized, it follows that none exist.)(1b) Modal Meinongians agree with 2KP Meinongians that fictional objects do not exist.But they deny any distinction among kinds of properties and deny that such objects exist in some weak sense.Instead, fictional objects are selected from the realm of nonexistents through the imaginative endeavours of authors when writing works of fiction.At the actual world, Holmes is a purely fictional character, among the most famous ever created and non-existent, while he has the properties of being a detective and a human being (and existing!) at (possible or impossible) worlds that realize the way Doyle represented the world to be when writing the Holmes stories.
(2a) Unlike Meinongian theories, abstract object theories claim that fictional objects actually exist, but as abstract objects rather than concrete objects.According to Zalta, who has the most developed such theory, Holmes is literally a human being and he is literally admired by millions of readers of the Holmes stories-but not in the same sense.For Zalta, there are two ways of having properties: an object like Holmes exemplifies being admired by millions of readers of the Holmes stories, while Holmes encodes the properties that he has according to the stories in which he appears, such as being a detective and a human being.(For this reason, Zalta's theory is sometimes classed as a two-modes-of-predication form of Meinongianism.) (2b) By contrast, Artefactualism holds that fictional objects are abstract objects generated by means of the creative process that an author goes through when writing a work of fiction.Thus understood, fictional objects literally have properties, like being an existent abstract artefact and being one of the most famous fictional objects ever created, while it is only in works of fiction that they have such properties as being a detective and a human being.
It is important to stress the role that properties play in all these accounts, and in particular, the role played by properties that fictional objects have within works of fiction.All agree, for example, that what makes it eminently worth engaging with Holmes and what explains Holmes's fame among readers past and present are the properties he has in the Holmes stories: the fact that he is a Victorian detective who solved numerous fiendishly difficult crimes, lived for a time at 221B Baker St, London, with his friend Dr Watson, was cocaine-addicted, and so on.Furthermore, in one way or another, these various accounts all accept that either the very genesis or, at least, the identity of fictional objects involves the properties they have in works of fiction.For Modal Meinongianism and Artefactualism, the genesis of fictional objects requires them to be either selected or generated through the imaginative activities of authors as they write their stories-there is no bare imagining of an object; imagining an object involves imaginatively entertaining a scenario in which there is something or someone doing certain things and standing in certain relationships to other things (it may be quite unclear how much of the story is needed for this task). 5 Thus, for all these accounts, there seems to be some dependence of fictional objects on properties.One variety of such dependence is a kind of genetic dependence of fictional objects on properties, although maybe only a narrow subset of the properties with which they are finally credited.Once such an object has been selected or generated, Modal Meinongians and Artefactualists allow the object and the stories in which it features to ground the truth of appropriate metafictional claims involving the same or other properties: for example, 'According to the Holmes stories, Holmes was a cocaine-addicted private detective' or 'Holmes was a cocaine-addicted private detective in worlds that realize the Holmes stories'.So for Modal Meinongianism and Artefactualism, there is a semantic dependence of in-the-fiction property attributions on fictional objects and the way they are represented in the fictions in which they feature.
Other theories insist on the dependence of the identity of fictional objects on the properties that they have within a fiction.Such dependence is a core thesis of both 2KP Meinongianism and Object Theory.On these accounts, properties that a fictional object has in a work of fiction provide criteria of identity or individuation, either wholly or in part.Terence Parsons, for example, claims that "the Φ of story s = the object x which has exactly those nuclear properties that the Φ has in s" ([9], p. 55; the view is broadly shared by other versions of 2KP Meinongianism).Zalta and his co-authors similarly insist on providing conditions that identify unique denotations for fictional names.According to OT's identity criterion for abstract objects, abstract objects a and b are identical iff they encode exactly the same properties; and since a fictional object encodes just the properties that it has in the work in which it appears, fictional objects a and b are identical iff they have the same properties in the works in which they appear.A partial reliance on such properties is also found in Voltolini's version of Artefactualism.For Voltolini, this set of properties provides a necessary condition on the identity of fictional objects; that is, fictional objects a and b are identical only if a and b are ascribed the same properties in the works in which they appear ( [7]).
So, these various approaches give us three different kinds of dependence theses: a thesis of the genetic dependence of fictional objects on properties that they have in stories, a thesis of the constitutive dependence of fictional objects on properties that they have in stories, and a thesis of the semantic dependence of statements concerning the properties that fictional objects have in stories on the objects and their stories.In the next section we will rehearse an argument due to Sarah Sawyer that suggests that such dependencies spell trouble for nearly all of the realist accounts of fictional objects here described.

How the Question of Fictional Properties Creates Trouble for Fictional Realism-Or Seems To
In 'The Importance of Fictional Properties', Sawyer suggests that fictional predicates spell trouble for both familiar realist and familiar antirealist accounts of fictional characters or objects: 6   [T]he. ..question of whether a fictional predicate refers to a fictional property is, as far as I know, rarely discussed.And yet fictional predicates are as important a part of fiction as fictional names, and, prima facie, give rise to similar semantic and metaphysical questions.Thus, if fictional predicates do not refer to fictional properties, then semantic questions arise about how to make sense of the apparent phenomena of meaning, reference, and truth. . ..If, on the other hand, fictional predicates do refer to fictional properties, then metaphysical questions arise about the nature and scope of those properties.. . . . ..I argue that the question of whether a fictional name refers to a fictional character is inherently bound up with the question of whether a fictional predicate refers to a fictional property.Consequently, the former, more discussed question (about fictional names and characters) cannot be answered independently of the latter, generally neglected question (about fictional predicates and properties).Crucially, a number of semantic theories of fictional names and metaphysical theories of fictional characters. ..presuppose unquestioningly that fictional predicates are guaranteed a referent.I argue that this presupposition is inconsistent with antirealist theories of fictional characters and that it cannot be taken for granted by realist theories of fictional characters either.( [6], pp.208-209) Sawyer adds that the considerations she advances "favour. ..a theory which is consistently antirealist about both fictional characters and fictional properties".We will leave comment on antirealist theories of fictional objects until the final section when we raise some questions about Sawyer's own preferred pretense theory.But we think Sawyer is right to highlight the importance of fictional properties and right to throw out her challenge.We can put the challenge in terms of the dependencies highlighted above: according to realist theories, fictional objects depend on properties that they have in stories.In fantasy and science fiction stories (but not just these), many of these "properties" (being a hobbit, say) are fictional properties.Because fictional objects depend on these fictional properties, it is incumbent on realist theories to show that fictional properties do indeed exist: without them, the dependence claims fail.
At first glance, the prospect for such theories does not look promising.Consider 2KP Meinongianism.Its proponents tend simply to assume that a fictional predicate like 'is a hobbit' stands for a perfectly ordinary property rather than (as we have been assuming) a fictional property, something that does not exist as a property.Parsons, for example, classifies being a unicorn as a paradigm nuclear property, much like being made of gold and being a mountain ( [9], p. 25).The reason is likely to be that he takes 'unicorn' as short-hand for something like 'white horse-like animal with one horn'.Proponents of Artefactualism and Modal Meinongianism tend to take a similar line. 7But pace Parsons and others, there is an immense gulf between a property like being a white horse-like animal with one horn and being a unicorn.Familiar thought experiments show that being a white horse-like animal with one horn is not sufficient for being a unicorn (nor, perhaps, that it is necessary, but put that aside).Similarly, being a humanoid creature with large furry feet is not sufficient for being a hobbit.What more it takes, however, is left indeterminate in the works of fiction or myths that feature the terms, which is why the predicate 'is a hobbit' does not stand for a property but, at best, a fictional property.That there is such an entity, however, is what needs to be shown.
Arguably, then, none of these theories take the apparent non-existence of fictional properties seriously enough.It is therefore not surprising to find that Sawyer rejects realist theories of fictional objects on the grounds that (a) the various dependency claims require there to be fictional properties, and (b) these theories are unable to establish that there are any such properties.(The theories Sawyer canvasses are 2KP Meinongianism, Thomasson's version of Artefactualism, and Zalta's Object Theory; she doesn't consider Modal Meinongianism.) We think, however, that Sawyer overstates the problem for realism, and in the next section, we show how realist theories can answer the objections we have canvassed.

The Realist Fight-Back
Fictional properties, if there are any, are not properties.What could they be?In our view, most, perhaps all, of the realist theories under discussion have the resources to answer this question, although, as we later point out, these very resources are often regarded as the weak point of the various theories.In what follows, we describe these resources and show how they enable the theories to satisfy both the Engagement and Unreality Constraints.(We take it for granted that theorists for whom fictional predicates stand for genuine properties have no trouble meeting the Engagement Constraint.) Note that our claim is the strong claim that most, if not all, familiar realist theories have such resources.The weaker claim that some realist theories of fictional properties have such resources is easier to show.Consider Artefactualism.In claiming that Artefactualism lacks such resources, Sawyer considers a would-be attempt to show that a fictional predicate like 'is a hobbit' stands for a property, an argument based on Thomasson's pleonastic approach to ontology: 8 (F1) Bilbo Baggins is a hobbit (F2) Bilbo Baggins has the property of being a hobbit So, (F3) there is a property of being a hobbit that Bilbo Baggins has.
Sawyer rightly points out that this argument is unsound if Artefactualism is true since artefactualists deny (F1): it is only in the stories that Bilbo Baggins-an abstract artefact-is a hobbit.But modifying (F1) by adding the prefix 'In the stories . ..' is, of course, of no help.Doing so can at best lead to the conclusion that, in the stories, there is a property of being a hobbit that Bilbo Baggins has.
But while arguments that attempt to prove that being a hobbit is a genuine property indeed look hopeless (indeed, this should not surprise us, given the sort of Kripkean considerations adduced earlier), Artefactualism can bypass such arguments by construing being a hobbit as an artefactual entity in its own right.That, in fact, is the approach taken by David Braun in a chapter appearing in the very volume that contains Sawyer's paper: We do not often use terms that refer to fictional and mythical attributes, and so we do not often find such phrases in discussions of fiction and myth.But we are inclined to say 'There are no (real) hobbits, but there are fictional hobbits, such as Frodo Baggins', and sentences such as these seem true and are difficult to paraphrase away.Further, it is easy to see how fictional and mythical attributes could be abstract artefacts of roughly the same ontological type as fictional and mythical objects.We have some grip on the conditions under which fictional and mythical properties exist.( [20], p. 101).
Presumably, one reason why Braun thinks this is that the creation of fictional and mythical properties involves story-telling, much like the creation of fictional and mythical objects.As Braun rightly points out, however, we usually do not talk directly about fictional and mythical properties: we talk about them indirectly, by talking about hobbits and unicorns rather than the properties of being a hobbit or a unicorn.And there is much that is said about these in stories and myths.We know, for example, that hobbits like communal living, have furry feet, are of a small stature, are of a race descended from elves and humans, and much, much more.But these are properties of hobbits.What is not immediately clear is how this gives rise to the kind of genetic dependence of the artefactual fictional property of hobbithood on properties that hobbits have in the stories, to parallel the way Holmes, for example, is genetically dependent on the properties ascribed to Holmes in the Conan Doyle stories.
The answer becomes clear from an analogy that will become increasingly useful as we go on: the strategy of Ramsification for dealing with the meaning and reference of theoretical terms (see especially [21]).Suppose that F is a theoretically defined term and that T(F) is the conjunction of theoretical assumptions containing F. On Lewis's formulation, F's meaning can be specified as 'the unique entity x such that T(x)', where T(x) is the open sentence-T's realization formula-that results from T(F) when F is replaced throughout by the variable x. 9 If x is a predicate-variable, then T provides a second-order specification of the property referred to by F. We cannot, of course, use such a strategy to define 'is a hobbit', since being a hobbit is a fictional property (an artefact, if Artefactualism is correct), rather than a property.But what we can say is that the Hobbit stories describe the genesis of the fictional property of being a hobbit in terms of properties that this fictional property has in the stories: its distribution, what its instances are like (the fact that these instances practise communal living, have furry feet, are of small stature), and so on.If so, just as there is a genetic dependence of artefactual fictional objects on first-order properties, there is a similar genetic dependence of artefactual fictional properties on properties of first-order properties: a second-order specification.
Enough has been said to allow us to make a number of further points.First, Artefactualism has no trouble meeting the Engagement Constraint: works of fiction generally provide a rich store of second-order information about fictional properties, which explains our ability to engage with fictional properties.Secondly, Artefactualism clearly meets the part of the Quasi-Property Constraint that declares that fictional properties are not properties; a further question is whether fictional properties so construed are nonetheless sufficiently property-like to do the work we should expect from the referents of predicates like 'is a hobbit' (e.g., helping to account for the truth of sentences like (i) 'In the Tolkien stories, Bilbo Baggins is a hobbit', (ii) 'Hobbits don't exist', (iii) 'Hobbits are merely fictional creatures', etc.) Following Braun, we shall assume that they are.(Here and in the case of the other theories to be discussed, we are putting aside certain more or less familiar worries about the ideological commitments of the theory, worries we briefly return to in Section 5.) Our third point is more general.Highlighting the way second-order properties give consumers of fiction a grip on the Artefactualist existence conditions for fictional properties allows us to see how all of the views under consideration (with the possible exception of Modal Meinongianism) are able to develop accounts of fictional properties that, arguably, satisfy the Engagement and Quasi-Property Constraints.Pace Sawyer, then, far from being stymied by fictional properties, all the versions of fictional realism she discusses seem to have the resources to show that there are fictional properties and that they have the sorts of features we should expect from them.Take Zalta's Object Theory.In claiming that realist theories of fictional objects do not take the problem of fictional properties seriously enough, Sawyer targets Zalta's and Parsons's theories together, taking any differences to be irrelevant to her purposes ( [6], p. 219).But that was a mistake.Zalta was, in fact, the first to develop a realist account of fictional properties that acknowledged their special nature, and he made use of secondorder properties to describe this nature.This is something that is entirely absent from Parsons's work.For Zalta, fictional properties are special "abstract" first-order properties: entities that encode rather than exemplify second-order properties.Here is how Zalta and Bueno summarize Zalta's earlier work on the subject in a recent paper criticizing Modal Meinongianism (we omit the formal details): [T]here is an abstract property (namely, an entity of type <i>) that encodes just the properties F of properties such that, in the legend l about unicorns, the property being a unicorn exemplifies F. Similarly, . . .there is an abstract property (namely, an entity of type <i>) that encodes just the properties F of properties such that, in the Tolkien novels about hobbits, the property being a hobbit exemplifies F. Since these abstract properties are unique, we can identify the properties being a unicorn and being a hobbit, respectively, with the abstract properties asserted to exist.This is completely analogous to what we did in the case of Holmes: whereas Holmes is an abstract entity of type i, the properties being a unicorn and being a hobbit are abstract entities of type <i> ( [22], p. 776).
Such an account shows why fictional properties are not properties in the standard sense: [I]n OT, there is a parallel between fictional individuals and fictional properties.Just as fictional individuals are abstract and so, by definition, not possibly concrete, similarly, fictional properties are abstract properties and therefore not possibly concrete properties [and so distinct from any ordinary property] ( [22], p. 776) (The reason in brief: as an encoding entity, a fictional property is simply not identical to any of the ordinary properties of type <i>; of these, we can coherently ask what, if anything, exemplifies them, not so in the case of encoding objects of type <i>.)Finally, the details of the theory (arguably) show that it nonetheless meets the Quasi-Property Constraint: despite not being properties in the standard sense, fictional properties so construed confirm the truth of sentences like (i)-(iv) in Section 1. 10  So Zalta has a straightforward reply to Sawyer's argument that the view does not take fictional properties seriously enough. 11(Whether the resulting theory is ultimately defensible is, of course, a wholly different matter.)What about Parsons's version of Meinongianism?Recall that according to Parsons's 2KP Meinongianism, the Φ of story s is the object x that has exactly those nuclear properties that the Φ has in s.As we saw earlier, Parsons seems to assume that this view straightforwardly applies to objects like Bilbo Baggins; as it stands, the theory does not acknowledge a special category of fictional properties.But if it is to deal with cases where the central constitutive properties are fictional, it needs to be able to identify fictional properties like being a hobbit, and in a way that respects the Quasi-Property Constraint.Explaining how this might be done follows the same contours as before.Where the predicates are fictional, we should look to the relevant works of fiction to identify the properties in question.The fictional object Bilbo is correlated with a set of nuclear properties that specify the criterion of identity for Bilbo, namely those nuclear properties that Bilbo has in the stories.This includes being a hobbit, a fictional property.As a fictional property, the latter is, in turn, correlated with a set of properties that specify the criterion of identity for the property of being a hobbit.These are the nuclear second-order properties that the property of being a hobbit has in the stories, where a second-order property is nuclear if all its instances are necessarily nuclear.This includes second-order properties, like being a species or race property, being a property instantiated by creatures descended from humans and elves, being a property instantiated by many short humanoid creatures who live in shires, love comfort, have hairy feet, and so on.(Note that being a hobbit also has extranuclear second-order properties, like being a property often thought about by readers of The Hobbit and The Lord of the Rings (the Hobbit stories, for short), being a property instantiated by more creatures in the Hobbit stories than there are US states, and so on.Such extra-nuclear second-order properties have no role in identifying what it is to be a hobbit.) In short, we can identify being a hobbit as the unique (nuclear) property h that has the (nuclear) second-order properties that are ascribed to h in the Hobbit stories, rather than, as in Zalta's Object Theory, the abstract property h that encodes these second-order properties.(Note that this is an even more straightforward way of applying the idea of Ramsification to fiction than our two earlier examples.) Far from being unable to make room for fictional properties, then, even 2KP Meinongianism has the resources to show that there are such things as fictional properties.Note that the account meets the Quasi-Property Constraint.First, the view shows why such fictional properties are not properties in the standard sense.Just as an object like Holmes does not exist because it is incomplete-there are numerous properties P such that it lacks both P and not-P (for example, having the right leg slightly longer than the left)-we should say that a fictional property like being a hobbit does not exist because there are numerous nuclear second-order properties Q such that being a hobbit lacks both Q and not-Q (having more than a hundred instances who have the right leg slightly longer than the left, say).Secondly, a fictional property like being a hobbit is nonetheless like a first-order property, being defined as something that possesses certain second-order properties.Once again, the details of the theory arguably show how this enables the theory to account for the truth of sentences like (i)-(iv) of Section 1.
What, finally, about Modal Meinongianism?Here, we are less confident.Recall that for Modal Meinongianism, fictional objects are selected from the realm of non-existents through the imaginative endeavours of authors as they describe their characters.So, a preliminary suggestion is this: Modal Meinongianism could stipulate that fictional properties are selected from the realm of all properties that have the second-order properties in terms of which they are characterized in the relevant work of fiction.But as it stands, this suggestion falls foul of one part of the Quasi-Property Constraint: we need the selected property to be unreal, not a property in the standard sense.So, we should add the restriction that such properties are to be selected from the realm of non-existent properties, properties that do not exist in the actual world but exist in other worlds.It is far from clear, however, that Modal Meinongianism can find room for this idea.As Priest and others put it in a recent publication critical of Zalta's Object Theory, "one is naturally inclined to take all properties to be the same kind of thing: abstract entities.A fictional property, in particular, is exactly the same kind of thing as a non-fictional property, like being in London" ( [18], p. 18). 12 This is a mark against Modal Meinongianism, in our view.

Room for Doubt
We have argued that different fictional realist views (apart, perhaps, from MM) have the resources to capture the idea of fictional properties, in a way that satisfies both the Quasi-Property and Engagement Constraints.But none of this shows that the theories are in the end defensible.Not surprisingly, in adopting the suggested accounts of fictional properties, each of the theories inherits familiar problems.2KP Meinongianism, for example, requires the second-order properties that are ascribed to h in the Hobbit stories to be nuclear; after all, if any of them were extranuclear, we might be able to derive obvious falsehoods (take existence to be extranuclear; Bilbo is a hobbit, so if being hobbit involved the extranuclear second-order property of being a property only possessed by existing creatures, it would follow that Bilbo exists).But critics of 2KP Meinongianism will find the distinction between nuclear and extranuclear second-order properties just as invidious as the distinction among nuclear and extranuclear first-order properties whose rejection lies behind the emergence of newer realist views, like Object Theory and Modal Meinongianism.(The other theories face similar difficulties.Take Object Theory.Zalta's notion of an "abstract" property involves an application, at a higher-type level, of the familiar encoding-exemplifying distinction that was defended in early work.Reject that distinction as ill-defined or defective in some other way, as many have done ( [18] is just the most recent), and the same will certainly be true of the generalized version used to define fictional properties.) 13  In short, although we think that there are promising ways in which these various theories can account for the unreality of fictional properties, as well their contribution to the truth of sentences involving fictional predicates, these ways all harbor (and perhaps even exacerbate) familiar difficulties for the theories.
Philosophers already impressed by the (alleged) difficulties facing realist theories of fictional objects might prefer to go in a wholly different direction.They might choose to opt instead for the kind of anti-realist view of fictional properties favoured by Sawyer.Sawyer thinks that fictional predicates ought to be treated in exactly the same way as fictional names-in terms of pretence: Fictional names and predicates are to be treated simply as if they refer to individuals and properties even though in fact they are empty terms. . ..And treating fictional names and predicates as if they refer to individuals and properties involves treating them as if they are terms that are meaningful and about which questions of co-reference can meaningfully be raised.This is sufficient . . . to account for intuitions about meaning and reference.But what of intuitions about truth? . . .[J]ust as fictional names and fictional predicates are to be treated as if they refer to individuals and properties, engaging with object-fictional sentences requires that they be treated as if they express complete propositions about individuals with properties, and hence as if they are true ( [6], pp.225-226).
Walton would agree.While Mimesis [24] does not tackle the issue of fictional properties, he tackles the issue of fictitious properties in a postscript to his 'Existence as Metaphor?' [25].There, he expresses his disagreement with Simon Blackburn's attack [26] on fictionalism about such (alleged) properties as colour and goodness.Blackburn asks what it would be for there actually to be colours if we are merely pretending to assert that roses really are red or canaries really are yellow: how would our world differ from some other possible world in which there really are colours?Fictionalists have no answer, Blackburn thinks, and so fictionalist theories about colour and goodness simply cannot get off the ground.But Walton thinks that the fictionalist he envisages has a ready response: There is no need to suppose that the speaker's words ("Canaries are yellow" or "Neglecting children is wrong"), taken literally, express a proposition, one that is true in some possible world.Fictionalism requires only that it be fictional in the implied game that these words express a proposition, whether or not they actually do, and the fictionalist can have the speaker pretending to be asserting a proposition even if there is no actual proposition that she pretends to assert.Thus can fictionalism about colours or moral discourse get off the ground ( [25], pp.115-116).
(Everything suggests that Walton would say the same about alleged fictional and mythical properties, like being a hobbit or unicorn.) On this kind of account, fictional predicates, like fictional names, are empty-no room for any doubt, then, about the unreality of fictional properties (nor, let us assume, the rest of the Quasi-Property Constraint). 14But it is a different matter where the Engagement Constraint is concerned.According to pretence theory, as described above, in telling or recounting science fiction or fantasy stories, we pretend that they are records of fact and that sentences that contain them express propositions: complete propositions.Perhaps there is room for such a theory.If one is a Millian about names and propositions, such an appeal to pretend propositions is precisely what one should expect.It is the form of the theory that strikes us as problematic.On Sawyer's and Walton's suggestion, speakers and readers pretend that there is a proposition by pretending that fictional names and fictional predicates refer to objects and properties.But that is not what happens in the typical games of make-believe that Walton has had much to say about.We do not pretend that there is something that the expression 'that bear' refers to; we pretend that there is a bear over there.We do not pretend that the name 'Holmes' refers to a person.We pretend that there is someone we are being told about who is a detective called 'Holmes', lives in Baker Street, has a friend called 'Dr Watson', and so on.In brief, it is phenomenologically incorrect to represent the pretence underlying our use of fictional names as metalinguistic because it misrepresents the way we engage with fiction.Perhaps a metalinguistic account is appropriate for the kind of pretence we are involved in when reading or composing lines in a nonsense poem (e.g., "the slithy toves did gyre and gimble"); it is not, we think, for the kind of pretence we are involved in when engaging with fiction.(Ironically, one of the antirealist views Sawyer discusses and dismisses, Greg Currie's make-believe account of the occurrence of fictional names in fictional contexts ( [28]), does far better on this front than her own, since it lends itself naturally to an application of the same strategy of Ramsification that proved useful for realist treatments of fictional properties.) 15

Conclusions
Our main conclusions can be summarized as follows.First of all, fictional predicates and properties are important.They are distinctive but philosophically neglected elements of certain genres of fiction, and they are an essential feature of the way fictional objects are presented to consumers of these genres.Furthermore, consumers engage with fictional properties just as they do with ordinary properties.Despite this, they are in an important sense unreal, just like fictional objects.
Secondly, contrary to Sawyer, they do not present a new problem for fictional realists.Rather, the ability of fictional realism to deal with fictional objects, by and large, carries over to fictional properties (as we saw, Modal Meinongianism may be an exception).Likewise, however, the most familiar challenges that have been mounted against realism about fictional entities also carry over to realism about fictional properties.A consideration of fictional properties will, therefore, not provide pretence theorists or other anti-realists about fictional entities with a knockdown argument against fictional realists, but neither will it give fictional realists much to celebrate.The problems (or alleged problems) that the different varieties of fictional realism face will still have the power to persuade philosophers of the virtues of something like a pretence approach.
But we have also seen that the way in which pretence theorists have dealt with fictional properties has its own drawbacks.Pretence theorists may have the solution par excellence to the problem of how to understand the unreality of fictional properties, but familiar versions of the approach struggle with the problem of how to understand our engagement with them.We suspect that which account of fictional properties wins out, in the end, is going to depend on a much more holistic view of the terrain.Many will resist such a conclusion, of course.Despite what we see as the power of this generalized Kripkean argument, the view that such fictional predicates refer to ordinary, if unusual, properties continues to be widespread.In this paper, we simply accept the Kripkean conclusion and focus on its consequences.A very different kind of case is the case of novel fictional predicates where the work in question makes it clear that they have standard (if perhaps unusual) application conditions able to be satisfied outside of the fiction by ordinary individuals.See, for example, Goodman's example of the fictional predicate "flurgish", introduced in a way that makes it clear that something is flurgish if it is a 30-foot tall human, and where only humans satisfy the predicate in the story ( [5]).That suggests that the predicate stands for a genuine property (a fictional universal, in Goodman's terms) whose possible instances include ordinary humans.(Goodman accepts the generalized Kripkean argument for the sorts of fictional predicates that are the focus of the present paper.)3 There is a more liberal understanding of 'fictional predicate' and 'fictional property' ('fictional+ predicate / property') on which matters are not so simple.On this understanding, a fictional+ predicate is simply a predicate that first appears in fiction and a fictional+ property or universal is whatever such a predicate designates, if anything.A more liberal usage of this sort has some advantages.For one thing, it does not prejudge the issue of whether predicates like 'hobbit' or 'unicorn' stand for genuine properties.(There are many who would contest Kripkean intuitions about such cases, something we discuss below.)It also becomes an interesting question whether there are fictional+ properties that should uncontestably be classed as genuine properties (see [5] for an argument that there are such fictional+ properties or "fictional universals").In the present paper, however, we assume that 'fictional predicate' is understood in the narrower, Kripke-inspired sense above.Cf. also the work of Fontaine and his colleagues, who propose an interesting way of combining game-theoretic modal semantics with constructive type theory to provide a background for treating fictional individuals as abstract artefacts.(See, for example, [16].)In a sense this work straddles Zalta's approach, with its use of type theory, and standard forms of Artefactualism.We are grateful to an anonymous referee for bringing this work to our attention.5 Thus, "During Doyle's storytelling, at some (early?)moment, the referent of 'Holmes' is arbitrarily chosen from the objects that, in the worlds that realize the story so far, s 1 , have the properties that s 1 describes Holmes as having" ( [18], p. 8).

6
Nothing in what follows hinges on any difference between fictional characters and objects.An artefactualist, like Voltolini, for example, thinks that there is nothing special about predicates like '[is a] unicorn/hobbit', and allows them the same role in criteria of identity for fictional objects as ordinary predicates ( [7], p. 68).Thomasson, on the other hand, seems to accept that fictional predicates are in some sense special, but she denies that this prevents them from expressing properties.She claims that we can justify their being such properties by means of pleonastic inferences; e.g., we can move from "The wand is not magical" to "the property of magicalness is not possessed by the wand" ( [19]).We think this style of argument is problematic because of the way it severs the link between the idea of a property and the notion of (more or less) determinate application-conditions for the corresponding predicate.As noted above, Thomasson ([19]) in fact provides a different argument, one not subject to this criticism.9 More generally, if T introduces a number of different theoretical terms F 1 , F 2 , . . .F n , let T(x 1 , . .., x n ) be the realization formula that results from T when we replace the newly introduced terms F 1 , F 2 , . . .F n with different variables x 1 , . ..x n .We can then let F i be defined as "the ith entity in the unique sequence of entities x 1 , . ..x n such that T(x 1 , . ..x n )." (If there is no such sequence, F i fails to denote.) In [23], we argue that the theory performs better on this front than Artefactualism.
It may be that Fontaine's type-theoretic approach to Artefactualism, mentioned previously, can be adapted to yield a similar result, but we do not have space to address this work in the present paper.12 Priest himself takes all abstract entities, and so all properties, to be non-existent ( [12]).13 Artefactualism faces its own problem.For Braun, fictional and mythical properties are "abstract artefacts of roughly the same ontological type as fictional and mythical objects"; because they are not properties, "they are not the sorts of things that can be exemplified or instantiated" ( [20], p. 101).It follows that in saying 'According to the Hobbit stories, Bilbo Baggins is a hobbit' the speaker is taking two abstract objects, B and H, and making the de re claim that according to these stories, B is a certain human-like creature and H a property exemplified by the first (rather than abstract artefacts created by authors).From the point of view of critics, this will surely be regarded as exacerbating what is already an unfortunate theoretical commitment to the idea that stories inevitably represent objects as categorically different from what they are really like.(For discussion, see [23].) 14 The view holds that it is as if fictional predicates stand for properties.Arguably, by suitably extending the pretence framework of Walton ([24]) or Everett ([27]), the view will then be able to account for the truth of sentences involving such predicates.15 Currie offers the following antirealist account of the semantics of fictional names as used in stories.Suppose S contains the fictional names N 1 , N 2 , . . .N n (and only these) and that S(x 1 , x 2 , . .., x n ) is the (realization) formula that results from S when we replace N 1 , N 2 , . . .N n with different variables x 1 , . ..x n .Currie assumes that when we make-believe the content of a story we make-believe that we are reading the words of someone, the fictional author, who is responsible for the text S, where S sets out her knowledge of the activities of the entities x 1 , . .., x n that satisfy S(x 1 , x 2 , . .., x n ).For Currie, we can then take N 1 as semantically equivalent to the definite description: the first entity in the unique sequence of entities x 1 , . ..x 1 , x n+1 such that [S(x 1 , x 2 , . .., x n ) and x n+1 is responsible for the text S and S sets out x n+1 's knowledge of the activities of x 1 , . .., x n ] ( [28], p. 149ff.)As Sawyer points out, however, story S might well contain fictional predicates, as well as fictional names, and Currie offers no justification for the crucial assumption that such predicates refer to fictional properties ( [6], p. 212).But whatever other problems Currie's account has, this is not one of them.The Ramsifying strategy, discussed earlier, can simply be extended to the case of fictional predicates in this new setting.Just replace the fictional predicates in S with predicate variables, and then define fictional properties in terms of their second-order properties, as in the case of theoretical predicates.Roughly speaking, being a hobbit, on this account, becomes something like the unique species property whose members are small, companionable, furry humanoid creatures, including one called 'Bilbo' who did . .., etc., such that the fictional author sets out her knowledge of this property in the Hobbit stories.And this fictional property does not exist because there is no property like that.