we share in the fiction-maker's pretense of doing things (and being things; 328), we do this by mak- ...... specify thin
UNIVERSITY OF CALIFORNIA Santa Barbara
Signifying Nothing: How Fiction Represents
A dissertation submitted in partial satisfaction of the requirements for the degree Doctor of Philosophy in Philosophy
by
Luke Andrew Manning
Committee in charge: Professor Nathan Salmon, Chair Professor C. Anthony Anderson Professor Kevin Falvey
March 2012
The dissertation of Luke Andrew Manning is approved.
C. Anthony Anderson
Kevin Falvey
Nathan Salmon, Committee Chair
March 2012
ACKNOWLEDGEMENTS
In working on this dissertation, I benefited from discussions with several of my colleagues at the University of California, Santa Barbara, especially Nathan Salmon, C. Anthony Anderson, Kevin Falvey, Philip Atkins, Matthew Griffin, Ian Nance, and audiences in the Santa Barbarians reading group, the Philosophy Department, and the Interdisciplinary Humanities Center. The last two of these also supported my work with fellowships. And I thank Amie Thomasson of the University of Miami and Stuart Brock of Victoria University of Wellington for their helpful correspondence.
iii
VITA OF LUKE ANDREW MANNING March 2012 Education Ph.D., Philosophy, University of California, Santa Barbara, March 2012 M.A. and C.Phil, Philosophy, University of California, Santa Barbara, June 2007 B.A. with honors, Philosophy, Michigan State University, May 2003 Professional Appointments 2005–2010, Teaching Assistant, Department of Philosophy, University of California, Santa Barbara 2007–2009, Teaching Associate, Department of Philosophy, University of California, Santa Barbara 2006–2007, 2009, Grader, Department of Philosophy, University of California, Santa Barbara Areas of Specialization Philosophy of Language, with Professors Nathan Salmon and Michael Rescorla Metaphysics, with Professors Nathan Salmon, C. Anthony Anderson, and Michael Rescorla Areas of Competence Logic, with Professors C. Anthony Anderson, Nathan Salmon, and Michael Rescorla Philosophy of Mind, with Professors Kevin Falvey and Michael Rescorla Ethics (especially Metaethics), with Professors C. Anthony Anderson, Aaron Zimmerman, and Christopher McMahon History of Analytic Philosophy, with Professors Nathan Salmon, C. Anthony Anderson, and Michael Rescorla Thesis “Sherlock Holmes and the Adventure of the Bandersnatch: Kripke’s Theory of Fictional Names,” Unpublished thesis submitted in partial fulfillment of the requirements for the degrees of Master of Arts and Candidate in Philosophy, University of California, Santa Barbara, 2007 Presentation “No Hope for Fodor’s Asymmetric-Dependence Theory of Representation,” University of Alaska, Anchorage (invited, April 2007) Awards • Ralph W. Church Fellowship, University of California, Santa Barbara, Fall 2011 • Pre-doctoral Fellowship, Interdisciplinary Humanities Center, University of California, Santa Barbara, Fall 2010 iv
• Phillip and Aida Siff Award, Best graduate student paper, University of California, Santa Barbara, Spring 2007 • Ralph W. Church Fellowship, University of California, Santa Barbara, Fall 2004– Spring 2005 • Honors College member, Michigan State University Fall 1999–Spring 2003 • Lewis K. Zerby Prize, Outstanding undergraduate paper, Michigan State University, Spring 2002 • National Merit Scholarship, Fall 1999–Spring 2003 Service and Professional Activities • American Philosophical Association, Student Member (2005–) • Website co-developer and maintainer, University of California, Santa Barbara, Philosophy Department (Fall 2005–Spring 2006) • Philosophy Club President at University of California, Santa Barbara, organizing departmental functions (Fall 2007–Spring 2008) • Conference co-organizer, “Themes from Burge,” at University of California, Santa Barbara (November 2009–May 2010) • “The Guerrilla Radio Show” co-host; an informal philosophy talk show broadcast on KCSB and online at www.guerrillaradioshow.com (2006–2008)
v
ABSTRACT
Signifying Nothing? How Fiction Represents
by
Luke Manning
There are many stories about Sherlock Holmes. Or are there? There never was such a person as Sherlock Holmes— that is part of what makes those stories fiction. On the other hand, of course the fictional character Holmes exists, because Arthur Conan Doyle created him. I explore three problems related to this. First, how are fictional objects created? John Searle, Gregory Currie, and a few other theorists sketch accounts in terms of fiction-makers’ intentional actions or general propositions that are true in the relevant works. These fail for many kinds of fiction, because fictional objects need not be created intentionally, and need not correspond uniquely to such general propositions. A general account of creation must appeal to the function of fiction, namely to guide our imaginative activity. Kendall Walton uses that to sketch a better answer: works generate fictional objects by guiding us to imaginatively simulate de re mental representation. I develop this in terms of the psychology of mental simulation and de re representation, and argue it is superior
vi
to other accounts. Second, do such works that generate fictional objects also represent them? In general, do representations apparently about fictional objects really represent them? Surprisingly, some theorists accept fictional objects but deny they are represented in many cases. I reconstruct and criticize their arguments, concluding that they have not yet found an obstacle to such representation. Third, Peter Geach raises a puzzle: analyze the sentence ‘Hob thinks a witch has blighted Bob’s mare, and Nob wonders whether she (the same witch) killed Cob’s sow’ as true even if there are no witches. Dozens of philosophers and linguists see this as a problem in the semantics of anaphora. But based on careful exegesis, I argue that Geach misrepresents the puzzle: the sentence cannot be true if there is no relevant witch, and so the puzzle is to account for its seeming truth without witches. I thus reduce it to the problem of apparent representation of fictional (or mythical) objects. This rules out nearly every published response, but leaves open a few possible paths, which I sketch.
vii
TABLE OF CONTENTS
1 How to Create a Fictional Object 1 Introduction 1 2 Fiction-maker criteria 5 3 Work criteria 14 3.1 Formal criteria 14 3.2 Content criteria 17 4 Consumer criteria 22
1
2 Real Representation of Fictional Objects 1 Introduction 37 2 Fictional names 40 2.1 Kripke on fictional names 41 2.2 Thomasson on fictional names 45 3 The work’s pretense 49 3.1 General content 50 3.2 Mismatch 51 3.3 Generation 56 4 Further work 59
37
3 No Identity without an Entity 62 1 Introduction 62 2 Puzzle pieces 63 3 Geach’s program 66 4 The puzzle, as Geach understands it 77 5 Proposed solutions 85 5.1 Solutions diverging from Geach’s stipulations 85 5.2 Solutions that otherwise miss the puzzle’s point 87 5.3 Metarepresentational solutions 90 6 Assessment of Geach’s program 96 Bibliography
105
viii
Chapter 1 How to Create a Fictional Object 1 Introduction Gene Roddenberry created the fictional character Spock in creating the television show Star Trek, and in general, common sense says fictional objects are created — brought into existence — in fiction-making. But how are they created? This question has two readings. First, the “transcendental” question: how is it even possible they are created? Many metaphysicians and philosophers of representation (philosophers of logic, language, mind, and depiction) are skeptical of creation: they think either there is no such thing as Spock to have been created, or there is such a thing, but it is independent of human action, necessarily and eternally. In contrast to such antirealist and Platonist views, creationism says fictional objects are abstract artifacts of the same general kind as novels, names, and nations. Spock is a real thing, though he is not really a person, logical, etc.; instead he has properties like being a famous fictional object and being created by Roddenberry. This view, first defended in the 1930s by Polish phenomenologist Roman Ingarden, then independently in the 1970s by Saul Kripke and Peter van Inwagen, is now the prevailing form of realism about fictional objects.1 An answer to the transcendental question is a defense of creation1. See Ingarden (1973a,b, 1989), Kripke (1973a, 2011), John R. Searle (1975), van Inwagen (1977, 1983, 1985, 2000, 2003), Jeanette Emt (1992), Alberto Voltolini (1994, 2003, 2006, 2009, 2010a,b), Stephen Schiffer (1996; 2003, ch 2), Amie Thomasson (1996a,b, 1999, 2003a,b, 2010), Nathan Salmon (1998), Stefano Predelli (2002), Jeffrey Goodman (2003, 2004, 2005), and David Braun (2005). For years Kripke’s views were known only from unpublished lectures, most notably his 1973a and 1973b; for discussion see Richard Routley (1980, 148–49, 561–63), Gareth Evans (1982, §10.1), Salmon (1998, esp. §§V–VI; 2011), and Edward Zalta (2006, §§3–5). Dummett et al. (1974) discuss Kripke’s 2011, the recently published “precursor” to his 1973a. Though van Inwagen expresses doubt that we
1
ism. Although I accept creationism, I will not argue here that it best explains the host of phenomena that theories of fictional objects confront. Instead, I will consider the more “empirical” question: assuming fictional objects are created, in which sorts of case does this happen? An answer to this question is a creation criterion, specifying the conditions for a work of fiction, or an act of fictionmaking, or whatever else, to create a fictional object. But metaphysicians and philosophers of representation concerned with fiction rarely discuss this question, or do so only in passing. This is a significant oversight. We have intuitions about creation— about which sorts of works and fiction-making generate fictional objects and which do not— and even noncreationists must account for these.2 More generally, fictional objects’ origins can tell us much about their existence conditions, their identity, and their relations to representation in and around fiction. In short, even if we ultimately reject our creationist intuitions, reconstructing them will illuminate the metaphysics of fiction. Then why have so many theorists, including creationists, said so little about creation, beyond the occasional platitude? Perhaps they suspect we cannot cover the full range of cases without a sensitive understanding of fiction, potentially including the aesthetic issues of defining and interpreting it, and the psychology of our engagement with it— tricky subjects for theorists mostly working in philosophy of language create abstract objects (2003, 153–55), he frequently includes being created by so-and-so among fictional objects’ real properties, so I treat him as a creationist. Kit Fine (1982, 1984) defends creationism but does not endorse it. 2. For Platonists this is called the “creation problem,” following Harry Deutsch (1991); see also Fine (1982, 1984). Deutsch (ibid., 211) and Peter Lamarque (2003, 38) propose that by ‘create’ we mean ‘exercise creativity’ in a supposedly ontologically neutral sense. I doubt we are confused in that way.
2
and metaphysics. I welcome this connection, as a rare chance to make a metaphysical issue less abstract and obscure, and to ground it in cases that are intuitively familiar and well understood. In §§2–3 I argue that most published creation criteria fail because they are not sensitive enough to the nature and variety of fiction. Then in §4 I defend a criterion based on Kendall Walton’s aesthetically robust work on fiction, and recent work in psychology and philosophy of mind on mental simulation and de re mental representation. That criterion is the only one that handles all the cases; so like it or not, the route to fiction’s metaphysics leads through arts and minds. Philosophical use of fiction-related terms is inconsistent and often insensitive to ordinary usage and relevant aesthetic and cultural issues. To avoid confusion I will define some terms. Fictional objects include fictional characters (i.e., fictional persons) like Spock, fictional planets like his home Vulcan, and other objects “originating in” works of fiction. Fiction is a type of representational art, and work of fiction is a type of representational artifact. The term ‘fiction’ is sometimes used more broadly, to include myth, hallucination, logical constructions, thought experiments, etc. Though we often run these phenomena together, they are worth running apart. I am concerned with fiction in the sense that novels are fiction, but works of history or biography (however inaccurate) are not. And not only novels are fiction; although we most associate the term ‘fiction’ with literature, the fiction/nonfiction distinction is more general. There are works of fiction in many media, including movies, television, radio shows, comic books, paintings, sculptures, prints, interactive media like
3
computer games, etc. I will use the following medium-neutral definition of fiction: A work of fiction is an artifact of a type normally made for guiding a certain kind of imaginative activity; call this activity engagement.3 This is the function a work of fiction in the sense of what it is made for, and not necessarily what it is used for. Myth, hallucination, history, and the rest are sometimes put to that use, but it is not their function. So mythical objects like Zeus are not fictional objects, because they do not originate in works of fiction (hereafter called works). Each fictional object is native to the work it originates in, but fictional (and nonfictional) objects may also immigrate to other works (see Terence Parsons, 1980, 51).4 Since not all works are literature (fine writing or linguistic fiction) or stories (works with narrative content, concerning a sequence of events), not works all are written/told by authors/storytellers, or read by readers. In general, fiction-makers make works, and consumers consume them. Many theorists propose or sketch creation criteria, or indirectly suggest them. These criteria come in three main kinds, depending on the feature of works of fiction appealed to. Fiction-maker criteria appeal to a fiction-maker’s actions; work criteria appeal to a work’s representational properties; and consumer criteria appeal to the way consumers are to engage with the work.5 I discuss these in §§2–4, respectively. 3. Compare the more common linguistic definitions, by Searle (1975) and others. As Walton (1990, §§2.3–2.6) argues, those are too narrow; see also below, §§2 and 3.1. Walton seems to accept the above definition (1990, §1.5, ch 2), but his later discussion (87–89) may conflate function and use; see below, note 13. 4. Several theorists reject immigration (e.g., Voltolini, 2006, ch 4). These nativists instead allow surrogates for supposed immigrants (Parsons, 1980, §3.4), that is, newly created fictional objects “based on” nonnative objects. I am against any restrictions on immigration, but the issue is tangential: the criteria below concern only natives, and so are compatible with both views on immigration. 5. Two theorists appeal to further features that some works contingently gain after they are made. For a work to generate fictional objects, Kripke says it must be published (according to Salmon 1998,
4
2 Fiction-maker criteria On John R. Searle’s (1975) influential theory of fiction and fiction-making, writing fiction is merely pretending to perform illocutionary acts that comply with the normal semantic and pragmatic conditions operative outside fiction. (Many works also mix in real illocutionary acts, but these acts are irrelevant here.) For example, in writing the novel The Red and the Green, Iris Murdoch pretended to assert propositions, and so pretended to commit to their truth (322–25). She also pretended to refer, and so pretended there were objects she referred to (330). In this way she created the novel’s protagonist, Andrew Chase-White. “By pretending to refer to (and recount the adventures of) a person, Miss Murdoch creates a fictional character. Notice that she does not really refer to a fictional character because there was no such antecedently existing character; rather, by pretending to refer to a person she creates a fictional person” (330). How should we generalize from this case? For one thing, I take it Searle means to address fictional objects in general, not merely fictional persons (cf. 331). But to generalize beyond that, we must know which of two things Searle is doing. On the first reading, he states conditions in which a fictional object was created (in this case, 316 n 43), and Voltolini says someone must reflect on its relation to certain sets of properties (2006, 84–89; 2009, 45–47). Both restrictions are implausible: there is no common practice of denying natives to works lacking these features. (See also n 8 below.) R.M. Sainsbury misreads Schiffer and van Inwagen as holding theories of creation of none of the above three kinds. Although Schiffer (1996, see below, n 7) says fictional objects are “languagecreated entities,” he really thinks fiction-makers create fictional objects, by using language in certain ways, contra Sainsbury (2009, 224 n 2). And although van Inwagen says fictional objects are “theoretical entities of literary criticism” (1977, 302), he never says critics create fictional objects, contra Sainsbury (2009, 98, 224 n 2); indeed, he often mentions Dickens creating in making fiction (305 ff.).
5
Chase-White), and thereby illustrates the sort of creation conditions all fictional objects have. On the second reading, he states conditions for creating Chase-White, specifically— an action of Murdoch’s that is specific to that character. He thereby illustrates how to map distinct fictional objects to distinct (classes of) creation conditions— namely, fiction-makers’ actions (or classes of such). Either way, Searle illustrates a creation criterion, but the second one is especially strong, what I call a specific-creation criterion. (Creation criteria correspond to functions from fictional objects to (classes of) creation conditions; the functions corresponding to specificcreation criteria are one-to-one (injective).) If he meant to give a specific-creation criterion, it is grossly underspecified. Although he says Murdoch’s pretended reference created Chase-White, he does not say how that differs from the pretended reference by which she created any other object. And it is not obvious how best to distinguish them; there are many ways, as I discuss below. So I take the first, more charitable reading: Searle commits to a relatively weak, but still interesting criterion, one he adequately specifies. Searle says Murdoch did not create Chase-White by really referring to a fictional object or to a real person— she merely pretends to refer. I take it he would also deny that she created Chase-White by really referring to that fictional character. Indeed, that would be impossible, by what I call the independence constraint:6 6. Compare Robert Howell (1997, 434 n 39). Sainsbury may think creationism violates IC. He says realist accounts of creation must involve “having the object in mind,” (2009, 63), but also that “until the creation has occurred there is no object fit to be the target of a de re belief, or of the de re ascription by an author of a property to a character,” (97). (On ascription, see below, n 9.) The former is not required, nor do creationists accept it (except, on a literal interpretation, Thomasson (1999, 13);
6
IC x can only have the function of representing (e.g., refering to) y if there is such a thing as y independent of x’s representing it. Since creation by merely pretended referring is compatible with IC, I take the following to be Searle’s creation criterion. SCC A fiction-maker creates a fictional object iff in making a work of fiction she (merely) pretends to refer to an object. Many creationists endorse something like SCC.7 But SCC has its problems. I will present several objections from critics, and then my own. First, according to Takashi Yagisawa, SCC leaves the connection between pretense and creation “mysterious,” because a fiction-maker’s pretending to refer does not entail there is a fictional object (2001, 155). In effect, his objection is that Searle fails to answer the transcendental creation question. But it is not clear Searle even means to. He seems more concerned with the empirical question, the one I am discussing. R.M. Sainsbury also criticizes SCC. One can question whether [authors] Murdoch or le Carré were pretending (at least intentionally) to refer to a real person. A paradigm way of doing that would be to engage in ordinary assertive behavior but slip in a name without a bearer, pretending that it has a bearer. By contrast, authors are engaged in the but see the end of this section and n 16). And the latter is too strong, unduly precluding representation of merely future and merely possible fictional objects. As long as we can uniquely specify them (which we can only do if there are such things) we can represent them, and introduce conventional representations with that function (Salmon 1998, 301, 316 n 43). 7. See Schiffer (1996, 155–56; 2003, 50–52), Salmon (strongly suggested, 1998, 300), Thomasson (2003a, 148–49; but see below §3), and Braun (2005, 611). Kripke (1973a, 6.19) also speaks of pretended reference, but he rarely mentions fiction-makers, so he may intend to describe the work’s reference.
7
distinctive act of telling a story, inviting their audience to make-believe in Second Lieutenant Andrew Chase White and Jonathan Pine. They don’t need to pretend that these are real people. Clearly, there’s no real person that Murdoch or le Carré pretended to refer to [in creating those fictional characters]. (2009, 95) Of course, Searle denies there was a real person who Murdoch pretended to refer to in creating Andrew Chase White. Nor does he hold she was really asserting during this pretended referring. Rather, according to Searle, what Sainsbury calls “the distinctive act of telling a story” is a form of pretense— probably best understood as the fictionmaker pretending to be the narrator and to perform her/his illocutionary acts (Searle, 1975, 327–28). And when fiction-makers “invite their audience to make-believe in” certain individuals, it is by inviting us to share in their pretense (330). Sainsbury’s remarks give no reason to doubt SCC. Finally, Stuart Brock (2010, §4) attacks Searle’s account of creation, which he says is incomplete. Although he does not distinguish specific- from weaker creation criteria, his discussion shows he takes Searle to commit not to SCC, but to a stronger, underspecified specific-creation criterion. As I argued, this is uncharitable. But a specific-creation criterion would be nice to have, especially for creationists. Brock’s suggests three such criteria on Searle’s behalf, and argues against them. The first two suggestions involve “fictional names” (presumably, names not used to refer to an immigrant, which will refer to fictional objects); one maps distinct fictional objects to distinct tokens of fictional names (§4.1), the other to the first tokenings of distinct fictional name types (§4.2). Both criteria fail because the mappings, obviously, are not
8
one-to-one. Brock’s final suggestion is that distinct fictional objects are created by distinct acts of pretended reference, each of which is “a causal consequence of [a distinct] intention to create such an individual,” (360). He gives three counterexamples to this criterion. First, a fiction-maker may write a draft of a work, pretending therein to refer with two distinct creative intentions, then revise the work so that it is about a single fictional object. In that case, the distinct intentions do not correspond to distinct objects (§4.3.1). Second, a fiction-maker may just once pretend to refer with a creative intention, but introduce two names, while leaving no evidence for readers that there was one rather than two creative acts. In that case, one creative intention corresponds to two objects (§4.3.2). Third, an antirealist fiction-maker can create fictional objects (if anyone can), without intending to do what she thinks is impossible (§4.3.3). I agree that Brock refutes these specific-creation criteria. But his first two objections to the third criterion are debatable. In writing a work’s draft a fiction-maker may create objects she later abandons as false starts. And in writing the final version she may create new objects, perhaps based on those abandoned objects. In such a case, her distinct intentions would correspond to distinct objects, though not objects that appear in the final version.8 (Brock argues there are not distinct objects in the draft by arguing against three interpretations of the final version. He says we should not suppose one of the pair appears in the final version while the other is “destroyed,” 8. Kripke and Voltolini think fictional objects are not created until after a work is finished (see above, n 5), so they would suggest drafts (of eventually-completed works) are irrelevant to creation.
9
or that both appear, “fused” into one or ascribed9 identity (361). Fair enough, but I see no reason either object must appear in the final version.) The second counterexample assumes that a fiction-maker can only affect a work’s interpretation if she provides consumers evidence for it, that her acting with certain intentions it is not always enough.10 This is plausible, but somewhat controversial, and Brock does not argue for it. In any case, his third objection is decisive, as are his objections to his first two proposals. If those were the only options, special-creation criteria would be unattainable. But there is another, more Searlean option, concerning pretended illocutionary acts. I suggest Searle should say fiction-makers pretend not only to refer, but to corefer. The pretended referrings involved in making a work can be partitioned into classes of pretended coreferrings (cf. Walton, 1990, 400-04), each class corresponding to a distinct object native to that work. So consider this Searlean specific-creation criterion: SCC′ For each (maximally inclusive) class of pretended referrings a fiction-maker performs in making a work that are pretended coreferrings, she creates a distinct fictional object. This avoids Brock’s counterexamples: it does not correlate fictional objects with names (tokens or types), or require that fiction-makers be realists. 9. ‘Ascription’ is van Inwagen’s term (1977, 304–08) for the ternary relation between a fictional object α, a property φ, and a work (or something more specific, like part of a work) β, such that, intuitively, α has φ according to β. Ascription is not exemplification, nor is it the relation Zalta calls ‘encoding’ (see, e.g., Zalta, 2000), which is not work-relative or contingent on there being works with certain content. 10. This objection is one of many that accuse fiction-maker centered theories of fiction of the socalled “intentional fallacy,” taking a work’s properties to depend too much on its maker’s intentions. See Manuel García-Carpintero (2007, §VII) for a recent discussion.
10
In any case, Brock’s objections do not refute SCC. It is not a specific-creation criterion, and it does not presuppose fiction-makers pretend to refer by name, or with an ontological intention. If a fiction-maker revises her work, Searle can allow that she abandons some objects she created. And he can argue that a fiction-maker can affect a work’s interpretation without giving evidence to consumers. I turn now to my own objections. There are two major problems with SCC. First, it is too narrow, because not all works comprise pretended illocutionary (linguistic) acts (cf. Walton, 1990, §§2.3– 2.6). Consider the movie 2001: A Space Odyssey, specifically the first monolith, which appears in a long wordless sequence. No pretended illocutionary acts created that scene and the fictional monolith. Rather, Stanley Kubrick and his crew filmed a black box on their prehistoric-Africa set, and included shots of it in the movie’s final cut. But perhaps Kubrick pretended to represent a monolith, visually, on film.11 Unlike the illocutionary act of referring, representing is medium-neutral, and so appropriate to many kinds of works: comic books, paintings, sculptures, children’s books with tactile and olfactory parts, etc. More specifically, a fiction-maker can merely pretend to represent in the relational sense: it can be part of her pretense that there is a relation between a representation and a representatum— that is, that the representing succeeds.12 So I attempt to revive Searle’s account with creation by pretended (relational) representation: 11. Or pretended to present it; see note 27 and the attached text. 12. ‘Relational’ is W.V.O. Quine’s term (1966, 177). Searle (1975, 330) says referring is relational. Compare the discussion of de re representation below, in §4.
11
CPR One creates a fictional object iff in making a work of fiction one merely pretends to (relationally) represent something. CPR is plausible, and pretended representation may be sufficient for creation. But there is still the second problem for SCC, which affects CPR as well. If fictional objects are created in fiction-making, and this requires that someone pretend to refer or represent, then all works and fictional objects are intentionally created by humans. Against this, Walton argues there could be a “naturally occurring [fiction] story: cracks in a rock spelling out ‘Once upon a time there were three bears . . . ’,” (1990, 87–88). I agree with Gregory Currie (1990, 35–36) and Peter Lamarque (1991, 126) that this is not a work. Stories and other works are constitutively artifacts.13 But there are stronger cases, because not all artifacts are human-made— they can be created by other artifacts with the right function. Even artifacts with derived intentionality— meanings, functions, etc.— can be made by other artifacts. The products still derive their intentionality from us, but indirectly. For example, a lowly staple has derived intentionality, the function of fastening. It derives this, not directly from anyone’s plans for that staple, but from a machine that makes them in bulk. The machine derives its intentionality, the function of making staples, from its human makers. Perhaps, normally, the machine’s human operators make staples. But if the machine is kept in order and supplied with material, it can be run “on autopilot,” or even accidentally, by something falling on its on-switch. In such a case the machine, rather than any person, 13. Walton ambivalently describes such cases as both fiction and merely treated as fiction (86–89), so his position is unclear.
12
makes staples. The same goes for works. There are computer programs for generating stories. Imagine Ada is a pulp writer and computer programmer who writes such a program. She specifies formulas to generate genre-appropriate plots, names, etc., and readers cannot distinguish its output from her formulaic fiction writing. If the program runs because something falls on the keyboard, then the program, not Ada, makes a work— something with the function to guide consumers’ imaginative activity. Yes, Ada is in some sense responsible for the story; it would not have existed without her writing the program, and if anyone deserves credit (or blame) for its aesthetic properties, she does. But she did not write or otherwise create it, and thus she did not create it as part of any pretense. Now imagine that the story accidentally generated by her program starts, “They say one man’s misery is another man’s mystery. My latest case was no exception.” Assuming the rest of the work supports a straightforward reading of this passage— for example, we do not discover later in the story that a monkey typed those lines at random— the program creates a fictional object, namely the narrator. And so a fictional object was created by machine, without any fiction-maker’s actions, thoughts, or pretense.14 Such machine-made works and fictional objects are counterexamples to CPR, to SCC, and to other fiction-maker criteria. For example, Amie Thomasson suggests writing certain words in a fiction text is a performative utterance, creating and nam14. Compare Peter David’s (1994, 104–08) argument that the unimaginative superhero character Cable was created not by any person(s) but by the institution Marvel Comics, effectively running on autopilot.
13
ing a fictional object (1999, 12–13, 47–48). But Ada’s program creates a fictional object without performative utterances, so they cannot be necessary. In short, while the fiction-maker’s role is often historically and aesthetically interesting, in general it is not the thought that counts; what counts is the product, the work (cf. Walton, 1990, 87). So I move on to work criteria. 3 Work criteria If a work satisfies a creation criterion, I will say it generates a fictional object. A person (or group of persons, institution, computer program, etc.) creates a fictional object by making it the case that a work generates one.15 Work criteria specify the representational properties by which works generate fictional objects, and they come in two kinds. One specifies the work’s content — what is true in it. The other specifies the work’s formal properties— what sorts of representational elements it contains, independent of its content. I begin with formal criteria. 3.1
Formal criteria
Ingarden, Jeanette Emt, and Jeffrey Goodman sketch formal criteria. Ingarden’s criterion for creation in linguistic works seems to be the occurrence in the text of a noun phrase.16 Emt mentions “language . . . where genuine reference is lacking,” (1992, 15. Normally, this is done by creating the work, but can be done by changing a preexisting work, for example by revising it or making a canonical sequel. 16. Ingarden’s semantics, using a traditional grammatical framework, gives noun phrases (especially the subjects of sentences) a special role. They “project” “purely intentional objects” (PIOs), objects that exist because (and while) they are represented (1973a, §§15, 20). Outside fiction, PIOs mediate representation; their qualitative contents determine the further representata (if any) of the representations that project them. Fiction merely simulates this; PIOs are a work’s ultimate representata, its native fictional objects (§§25, 28, 38). Though most philosophers of representation reject PIOs, we can evaluate Ingarden’s criterion independently. Thomasson (in early work: 1996a, §VII; 1999, 88–92; cf. 1996b, §II) also reduces fictional objects to PIOs, but gives no creation/projection criterion.
14
154)— presumably expressions whose function is to refer, like singular terms. Goodman mentions the “association” of predicates with names, which suggests the text must contain both.17 These sketches need development. Which sorts of singular terms are sufficient? Which are necessary? (Proper names are not; Frankenstein’s monster is never named in Mary Wollstonecraft Shelley’s novel.) Does sentential context (e.g., quotation) matter? How do I create one fictional object with two names?18 I will head this inquiry off at the pass: formal criteria will fail on general grounds, however they are developed. First, the above suggestions are linguistic, but many works use other modes of representation besides (or along with) language, modes lacking the kind of purely formal properties cited here.19 Linguistic expressions have formal (syntactic) properties we can identify independent of the expressions’ content. For example, some of them are singular terms: their function is to designate one individual, and we can tell this without knowing what (if anything) they designate. But depiction (pictorial representation), auditory representation, and other perception-based modes of representation have elements without such formal properties. Call a display an artifact that represents, at least in part, by one of those modes— a photograph, a map, a sound or video recording, a statue, etc. Displays represent as they do (at least in part) because they appear a certain way (in appropriate observation conditions to observers with 17. See Goodman (2004, 144–45, 146–47 n 26), and below, note 24. 18. Compare Brock (2010, §4.2; discussed above in §2). Schiffer seems to doubt one can (1996, 162). 19. Ingarden also discusses nonlinguistic works (1989, pts. 2, 4), but gives no corresponding criteria.
15
appropriate perceptual faculties).20 For example, in Salvador Dalí’s painting The Persistence of Memory there is a dark brown patch, which represents a tree (or the parts of a tree not occluded by a melting watch). We know it represents an object only because on the painting’s proper interpretation, looking at that area is like looking at a tree (in some relevant way I will not attempt to define)— because we understand the whole painting’s content. In another context, a patch of the same shape and color could lack that object-representing function; it could be part of an abstract design, or a mottled background, or a picture within the picture. Unlike linguistic expressions, (some) parts of displays have no representational function that we can identify without knowing their content. So no purely formal criterion tells us which nonlinguistic works generate fictional objects. Second, formal criteria are insufficient even for linguistic works. A work can contain singular terms but fail to generate corresponding fictional objects, if on the work’s proper interpretation the terms do not fictionally refer. Consider this short Sherlock Holmes story. 1 Holmes found a note with the name ‘Hugh Fitzhugh’. He asked if I had heard of the fellow, but I was no help. Then he learned the note’s author knew the name was not in use, so he gave up the search. 1 contains the proper name ‘Hugh Fitzhugh’, so a formal criterion might predict that 1 generates a fictional object. (Holmes and the narrator, Watson, are immigrants.) 20. Most displays do not generate all their content this way. Notably, many are also de re representations of individuals and kinds, which I discuss in §4 (see n 40).
16
But there is no such character, because according to the story, ‘Hugh Fitzhugh’ is an empty name. 1 fails to generate a fictional object because the wrong things are true in it— because of its content — so formal criteria are insufficient.21 A work’s formal properties are only a rough guide to interpretation, to its content. Since content matters to creation, purely formal criteria will not work. 3.2
Content criteria
Content criteria say a work generates a fictional object iff certain things are true in it. Which propositions— ones about the object generated? Such propositions may be true in a work as a result of its generating the object, but by IC a work cannot generate an object by representing it. So which other propositions must be true in a work for it to generate fictional objects? Setting aside the interpretive question which propositions are true in a work, most theorists agree that consumers should make believe them while consuming it (see §4). Granting this, we can state a content criterion in terms of make-believe, as Alberto Voltolini does. “The kind of make-believe process that leads to the constitution of a fictum is what [Gareth] Evans would label an existentially creative game of make-believe. In other words, those who take part in such a process . . . make believe that there is an individual, typically a concrete one, that (explicitly or implicitly) has a certain set of properties” (2006, 74). I take it that by ‘an existentially creative game of make-believe’ he means a game that involves mak21. Salmon (in correspondence) suggests a formalist reply: the criterion should require an occurrence in the work of a specific name (like ‘Aristotle’, the philosopher’s name), not a generic one (like ‘Aristotle’, the name several men share). But which sort of name ‘Hugh Fitzhugh’ is in 1 depends on what is true according to 1, not on the work’s purely formal properties, and so this reply is not available to formalists.
17
ing believe there is some object distinct from any immigrants the game is about (see Evans, 1982, 358). (I ignore here his claim that creation also requires reflection on the work— see above, n 5— which makes this criterion only necessary.) In non-makebelieve terms, we have a naive content criterion: NCC A work generates a fictional object iff it is true in that work that there is something other than an immigrant. Since NCC only gives the conditions for creating at least one object, it is difficult to test with works that may generate multiple objects. It is not obvious which corresponding specific-creation criterion to endorse, so I will consider something else.22 A fictional-φ creation schema states the conditions for creating a fictional object that is ascribed, by its native work, a certain property. Call a fictional object ascribed φ by its native work a natively fictional φ. Consider this schema: NCΦ A work generates a natively fictional φ iff it is true in that work that there is a φ, other than an immigrant.23 22. Voltolini nearly endorses a specific-creation criterion. According to his reductive account of fictional objects, “ficta are compound entities consisting of both . . . a certain set of properties, the properties corresponding to those mobilized in the make-believe process-type that leads to the generation of a certain fictum— the relevant part in a particular game of storytelling — and that very process-type itself ” (2006, XV). But as far as I can tell, he never says how to divide up the “game of storytelling” of a work, except into parts corresponding to fictional objects (73). So at most he commits to there being a mapping of fictional objects to these elements. (He may not even commit to that, since he is explicitly neutral about whether some kinds of make-believe correspond to fictional objects; see 209–11.) The only well-defined criterion he endorses (ignoring his reflection requirement) is NCC, so I evaluate that instead. 23. Without the ‘natively’ qualification, there would be counterexamples. For example, Action Comics 1 (1938) generates Superman, and Superman is a fictional flying man. But it is not true in Action Comics 1 that there is a flying man; it ascribes Superman the power to leap an eighth of a mile, not the power of flight.
18
NCΦ has counterexamples. For example, it may be true in J.R.R. Tolkien’s novel The Lord of the Rings that there is a tallest Uruk-hai, which is distinct from any immigrant (as all the Uruk-hai are). But the work does not say which is tallest, so it does not generate a fictional object to which it ascribes being the tallest Uruk-hai. Furthermore, a work can ascribe nonexistence to one of its native objects. For example, according to Hamlet, Yorick is dead, so he does not exist (anymore). But the play still generates a natively fictional jester-called-‘Yorick’ (assuming Yorick is native to Shakespeare’s version), so NCΦ is false. But perhaps there is a difference between existing and being, so that according to Hamlet, there is a jester called ‘Yorick’. Then this is not a counterexample to NCΦ. But consider a story that denies being to a native: 2 Mariko lived in Kansas — in the weak, Meinongian sense that something can live in Kansas even though it lacks being of any sort. Yes, poor Mariko was a beingless object. She had been in love and she had been experienced, but she had never been. It is sometimes difficult to interpret philosophically odd stories, but the oddity here is not a mistake, something we should “interpret around.” So it is not true-accordingto-2 (on any plausible interpretation) that there is a Kansan called ‘Mariko’. But 2 generates a natively fictional Kansan-called-‘Mariko’, so NCΦ is false.24 Further24. Goodman’s (seemingly) formal criterion suggests a variant of NCΦ. He says that for a name occurring in a work to generate a fictional object, it must have “enough properties associated with it,” where “enough” is a vague amount (2004, 146–47 n 26). So perhaps a work w generates a natively fictional object-named-n iff according to w something named n (other than an immigrant) is φ, for some sufficiently meaty property φ. This fails not only for 2, but for natively unnamed fictional objects like
19
more, 2 generates a fictional object (namely Mariko), but it is not true in 2 that there is something other than an immigrant — so NCC is false as well.25 So what kind of content makes 2 generate a fictional object? Perhaps it concerns the work’s narrator. G.E. Moore and Currie hold that each work has a narrator who recounts her knowledge of objects she has in mind. Moore (1933, 68–70) says the narrator is the author, while Currie says it is sometimes a fictional character.26 In either case, as Currie argues, the narrator must be a perfectly reliable one, what he calls a teller, and so may be distinct from any ordinary character in the story, like A Study in Scarlet’s Watson (Currie, 1990, §3.7). That all works have tellers is controversial interpretive claim. And it is not general enough, since narration is linguistic presentation of narrative (i.e., sequential) content. But let us grant Currie the related assumption that every work has a perfectly reliable presenter, someone who presents its content to consumers, knowing that content to be veridical.27 Now consider this Moore/Currie-style creation criterion: MCC A work generates a fictional object iff according to it the presenter recounts her knowledge of something she has in mind, other than an immigrant.28 Frankenstein’s monster, and minimally detailed ones like Walton’s character George (see example (A) below, p. 25). 25. NCC is a restricted analogue, for fictional objects, of what Parsons (1980, 30–32) calls Alexius Meinong’s “unrestricted satisfaction principle.” Compare the Yorick and Mariko objections to Bertrand Russell’s (1905, 482–83) criticism of Meinong’s treatment of statements like “The existent present King of France does not exist.” 26. In earlier work (1990, 150–54) he says it is always a fictional character, but he now says in some cases it is the author (2010, ch 4, esp. 67). 27. On presenters, see George Wilson (1997) and his references. 28. Currie appeals to narrators to support a variant of NCC, that a work w generates a fictional object iff according to w there is a unique φ, other than an immigrant. (Parsons seems to endorse this variant (1975, 82), but later rejects it (1980, 181). Lamarque (2003, 43) also suggests unique individuation is
20
This is difficult to evaluate without evaluating the presenter assumption. If, as Currie suggests, presenters also present themselves (perhaps only indirectly, by presenting other things; cf. Currie 1990, 123, 157–58), then the assumption guarantees that all works satisfy both sides of MCC. But consider this Moore/Currie-style fictional-φ creation schema: MCΦ A work w generates a natively fictional φ iff according to w, the presenter recounts her knowledge of a φ, other than an immigrant, that she has in mind. Granting the presenter assumption, this does better than NCΦ. According to 2, there is a presenter who has in mind and recounts his knowledge of a Kansan named ‘Mariko’, and so by MCΦ the work generates a fictional Kansan-named-‘Mariko’. But there are counterexamples. It can be true in a work that the presenter presents nonfictional photographed scenes, knowing them to be veridical, without having in mind all the objects they are pictures of. So there can be fictional pictured-φs without it being fictional that the presenter has a φ in mind. And it can be true in a work that the presenter reports the text of a song, unwittingly giving consumers conclusive evidence that it is about a historical person (other than an immigrant), and so without having that person in mind. In that case, it is fictional there is a subject of that song, and there really is a fictional subject-of-that-song, though fictionally the presenter has no one in mind. Since presenters can overlook — fail to have in mind— objects implicated in the representations they know to be veridical, then even granting the necessary. Cf. Voltolini 2006, 209–11.) Currie argues that objections to this variant overlook a powerful substitution for φ: ψ such that the teller has it in mind and recounts her knowledge of it (1990, §§4.6–4.7). But this variant shares NCC’s counterexamples, so I discuss MCC (and MCΦ) instead.
21
presenter assumption, MCΦ fails. Currie might respond by strengthening the assumption: suppose that the presenter not only knows that everything she presents is as she presents it, but has in mind every object implicated in what she presents. This super-presenter assumption makes MCΦ more plausible, but why should we accept it? It is Currie’s burden to support such a controversial claim. His support for the teller claim is that (i) it is needed to account for consumers’ engagement with fiction (1990, §3.7), and (ii) it is part of the right creation criterion (§4.7; see above, n 28). In this context, he can only appeal to (i). I will argue against this appeal in §4: we need not posit a teller, presenter, or super-presenter to make sense of engagement, and so we need not accept MCC and MCΦ. Having argued against all content criteria I know of (see nn 22, 24, and 28), I now move on to consumer criteria. 4 Consumer criteria A consumer criterion says a work generates a fictional object iff it prescribes that consumers engage with it in a certain way. (The work’s prescriptions are the ways consumers are to engage with it, on its proper interpretation; see Walton 1990, 39– 41.)29 In the literature on fiction in metaphysics and philosophy of representation, most theorists assume engagement is taking a certain attitude (e.g., make-believe) to the propositions true in a work. This makes content criteria such as NCC equivalent to consumer criteria. And according to many theorists, in engagement we continue the 29. It may be indeterminate what some works prescribe, and particularly whether they satisfy a given consumer criterion. I suggest that if a work fails to determinately make certain prescriptions, it determinately fails to make them.
22
fiction-maker’s game of make-believe, the game whose initiation created the work.30 This makes consumer criteria equivalent to both content and fiction-maker criteria. But engagement cannot always require continuing a fiction-maker’s game, because some works have no fiction-maker, as I argued in §2. Furthermore, it is doubtful that engagement is merely making believe certain propositions. Many aestheticians and psychologists also discuss affect, perception, and action in engagement. Most of that literature does not mention creation, but there are two notable exceptions. First, as I noted, Currie supports a content criterion with his theory that consumers “imagine themselves in the presence of one who tells a story he knows to be true,” (1990, 125). Second, there is Walton’s consumer criterion: WCC A work generates a fictional object iff it prescribes that consumers imagine having de re knowledge of something other than an immigrant.31 Unlike most of Walton’s work, WCC has been ignored in the literature.32 I think it is basically right, but it needs development. In the rest of this section I will explain it in terms of Walton’s theory of engagement, then precisify it in terms of recent work in 30. See, for example, Searle (1975, 330) and Evans (1982, 353). Searle does not say whether, when we share in the fiction-maker’s pretense of doing things (and being things; 328), we do this by making believe propositions that are true in the work. If we do not, his criterion is equivalent only to a consumer criterion, not also a content criterion. 31. See Walton (1990, §3.8, esp. 133–36). Compare Lamarque’s brief comment (2003, 43) that a “crucial feature of character initiation . . . is something akin to indexicality. A reader of the narrative or someone reporting it can refer to that character.” Of course, if a work is about a character, then readers can refer to it. But I take it Lamarque is stating how a work gets to be about a character. Unfortunately he does not expand on this (cf. above, n 28). Compare also Searle’s, Moore’s, and Currie’s criteria, which concern a fiction-maker’s or character’s de re referrings, rather than the consumer’s de re knowledge. 32. Oddly, even Walton ignores it in his later discussion of the metaphysics of fiction (1990, chs. 10– 11). He adopts a fictionalist version of Meinongianism rather than creationism, but does not explain how Meinongianism or antirealism relates to WCC. (See above, n 2 and the attached text.)
23
psychology and philosophy of mind on mental simulation and de re mental representation. Finally, I will argue it handles the cases other criteria could not, and extend it to a specific-creation criterion. According to Walton (1990, ch 1), we imagine having de re knowledge while participating in games of make-believe, imaginative activities guided by props. Props are objects whose real properties contribute to determining what is true in the game (or as Walton misleadingly says, “fictional in the game”). A work is something whose function is to be a prop (but see above, n 13). In engagement with a work, both it and the consumer are props, and so works prescribe two kinds of imagining besides making believe the propositions true in them. First, they prescribe making believe propositions about oneself, which are true in one’s game, but not in the work. Second, they prescribe nonpropositional imagining, imagining doing something or being some way. For example, movies prescribe that we imagine seeing things, and if we imagine accordingly then it is true in our respective games that we see things. (This is not to say we deliberately follow such prescriptions; proper engagement is usually automatic, except with works that impede it, by defect or design.) Works prescribe (and permit) many kinds of engagement— affective, perceptual, and behavioral, including verbal— each of which Walton analyzes in terms of (a) propositions true in the consumer’s game, and/or (b) nonpropositional imagining within that game (1990, chs. 5– 9).
24
Walton says imagining having de re knowledge is a kind of nonpropositional imagining (136). He does not analyze it, but gives two examples. First, consider “a picture (X) showing a fish or several of them swimming in a lake,” (134). “When one sees picture (X), not only is one to imagine that one sees a fish, but . . . one is to imagine seeing a fish, and thus knowing about one,” (136). Second, consider this work: (A) George was an old and almost worn out ghost who lived in the rundown mansion on Spruce Street. The End (133) Walton says, “The reader of (A) imagines knowing about a ghost, not just that he knows about one,” (136). And how, within his game, does he know about it? “If story (A) has a narrator, it may be fictional [i.e., true in the reader’s game] that the reader is told about a ghost by him, that the reader knows it as the ghost the narrator speaks of. If there is no narrator, it will still be [true in the reader’s game] that the reader knows of a ghost, although it may be indeterminate how [in his game] he knows of it,” (135). Thus Walton allows that a work’s prescriptions regarding de re knowledge may be incomplete: works do not always prescribe that we imagine getting our de re knowledge in any particular way.33 (And so we do not always imagine getting it through communication with a teller or super-presenter.) These examples suggest that imagining having de re knowledge is imagining being in conditions for having de re knowledge, conditions like seeing things and hearing tell of things. And where we do not imagine being in specific conditions, we still imagine being in some such conditions or other. 33. On incomplete prescriptions more generally, see Walton (1990, §§6.4–6.6, 7.2) and Wilson (1997).
25
This is a start, but since imagining is a vague notion, so is imagining having de re knowledge. To make WCC precise, I must detour through philosophy of mind and psychology. At first Walton declines to analyze imagining (1990, 21), but later he says “the participation in make-believe that I described is . . . a form of mental simulation,” (1997, 38). Mental simulation is a way of mentally representing mental states, processes, etc.; namely, matching them by exercising some of the psychological mechanisms underlying them. (Some qualifications: The matching may fail. Simulating a state may or may not be having the state itself. The simulated state may or may not be anyone’s actual state— even the simulator’s. Cf. Karen Shanton and Alvin Goldman 2010, 528.) Philosophers are most familiar with simulation as an explanation of mindreading, our ability to mentally represent others’ mental states. (It is often cast as a rival to the propositional-attitude explanation called the “theory theory.”) But a variety of experimental data show we simulate in many other contexts, even in some kinds of engagement.34 Simulation is especially useful in analyzing nonpropositional imagining, which cannot be merely holding propositional attitudes.35 Walton does not restate WCC in terms of simulation, but he might say that simulating having de re knowledge is simulating being in conditions for having de re knowledge. Unfortunately, it cannot be this simple, given certain assumptions about 34. See Shanton and Goldman (2010) and Goldman (2006, 287–88). He also cites (2006, §11.4) several theorists, including Currie, who take engagement to involve simulation. 35. Nonpropositional imagining may somehow involve propositional attitudes, or at least propositional imagining— make-believe. And make-believe may itself be simulated belief; cf. Goldman 2006, 61.
26
de re knowledge and simulation. By ‘de re knowledge’ I take it Walton means not only propositional knowledge, but more generally de re mental representation, which for short I will call beholding.36 And I assume the conditions for beholding are externalist. Specifically they are, as Burge says, anti-individualist: to behold something, a subject must bear certain nonrepresentational (coarsely, causal) relations to her environment; she need not represent conditions for representing that thing (e.g., a qualitative specification of it) or for objective representation in general.37 The problem is that we cannot simulate being in these external conditions, because they are not themselves mental states. So simulating beholding cannot be simulating being in beholding conditions. However, we can make believe we are in certain external conditions. And if we do that, we can simulate being in corresponding internal conditions, undergoing the kind of mental activity that, in those external conditions, underlies beholding. This is part of Currie’s view: in engaging with any work, we make believe there is a teller (or super-presenter), and that she presents things to us, which we simulate beholding by understanding her communication. I doubt we always posit such intermediaries, but some works prescribe that we make believe we are in certain external beholding conditions. The best case is a first-person-perspective computer game like Doom; these 36. This includes cognitive representation of individuals— “singular thought,” properly so-called — and, if there is any, plural, attributive, or noncognitive representation, for example in perception. On beholding beyond singular thought, see Tyler Burge (2009, 247; 2010, 31–32, 537 ff.). 37. See Burge (esp. 2010, ch 3). Walton might agree — he endorses Kripke’s analogous externalism for the semantics of names (1990, 402). And since he says imagining having de re knowledge is nonpropositional imagining, he would likely reject the individualist views that reduce beholding to non-de re propositional attitudes.
27
prescribe that we make believe we inhabit an environment, and so prescribe that we simulate beholding by simulating perceiving things, in person. But most works do not fully specify the external conditions we should make believe we are in. A written work with no narrator does not prescribe that we make believe we learn of things in any specific way, for example, by being told of them or by reading a nonfiction document. Even Vladimir Nabokov’s novel Lolita, which prescribes we make believe it to be its main narrator’s memoir, published by a psychiatrist, does not specify how we should make believe we come to access the memoir. A movie like 2001 does not prescribe that we make believe we see the first monolith in person, or in a nonfiction recording, or in any other specific way. Even the movie This is Spı¨nal Tap, which prescribes we make believe it is a documentary, does not specify how we should make believe we came to see it. So unless Currie is right, most works’ prescriptions to make believe fail to determine what mental states we should simulate to simulate beholding. We neither simulate being in, nor make believe we are in, specific external beholding conditions, but we do not need to. Of course one must be in specific external conditions to behold something— that is the point of externalism. But beholding conditions can be divided more finely than internal/external. Consider perceptual beholding: on a less simplistic view it has four parts. (i) There are things in our environment, the distal simuli. (ii) By some causal route, these affect our sense organs. For example, light rays scattered from objects may travel straight to our eyes, or they may be reflected or refracted in various ways. (iii) We get proximal stimuli, for ex-
28
ample retinal images. (iv) Our subpersonal perceptual system processes the proximal stimuli. This last part is the subject of perceptual psychology. It is quite complex, but a brief summary is sufficient here.38 The perceptual system produces representations of distal stimuli by processing proximal stimuli. The latter greatly underdetermine the former (as well as the causal route between them). For example, a retinal image underdetermines the shape, color, and other properties of a seen object. Still, the system outputs relatively rich and determinate representations. Occasionally, the subject’s propositional attitudes affect certain kinds of processing—for example, if he believes part (ii) introduced distortion. But mostly the system works without such influence. It automatically and unconsciously compensates for many variations in perspective, lightness, etc. by exercising perceptual constancies, “capacities systematically to represent a given particular or attribute as the same despite significant variations in proximal stimulation,” (Burge, 2010, 274). These are capacities to behold things in the distal stimuli. So for the most part, the perceptual system is sophisticated enough to reconstruct (i), and anything we might care about in (ii), entirely from the proximal stimuli. Now consider again the monolith scene from 2001 in terms of this division. Beginning with (iii), it prescribes that we simulate getting certain proximal visual stimuli, those we really get by viewing the movie in appropriate conditions. (iv) It thereby prescribes that we simulate processing those stimuli and beholding things in the dis38. See Burge (2010, esp. ch 3, 87–100, 396–416) for more.
29
tal stimulus, namely a monolith and some hominids. As a result, (i) it prescribes we make believe there is a tall black monolith surrounded by primitive hominids. Regarding (ii), it does not prescribe our making believe anything much about the causal route from the distal to the proximal stimuli. So the movie, by its prescriptions to make believe and simulate, specifies part (iii), and thereby parts (iv) and (i). It does not specify much of part (ii), but it does not have to, in order to prescribe simulating beholding. We simply simulate processing proximal stimuli as if they came from the distal stimuli they appear to, rather than from a fiction movie, and in most cases these determine anything we care about in part (ii). Occasionally they do not; some works prescribe that we simulate perceiving in distorting conditions that our perceptual system cannot compensate for. For example, in the sequence at 2001’s end, some of the visuals are so alien we cannot tell what we are to simulate seeing. In such cases, the work may prescribe making believe about part (ii), rather than simulating reconstructing its important aspects from proximal stimuli. Or it may not prescribe anything in particular, leaving open what we simulate perceiving. What exactly a work prescribes may be a difficult interpretive question. But most works prescribe only perceptual simulations within the (wide) range of variation our perceptual system can compensate for, and so clearly prescribe simulating perceptual beholding. That is a sketch of simulated perceptual beholding, but we also simulate beholding things not through (direct) perception. We may simulate perceiving things on television monitors, hearing of them in dialogue, etc. To explain this I must explain
30
beholding mediated by de re artifactual representations. By ‘de re’ here I mean there is something such that the representation (absolutely, or on a certain conventional use) has the function to represent it. These include some linguistic expressions (e.g., typical proper names)39 and utterances (e.g., demonstratives with demonstrations, definite descriptions on Keith Donnellan’s “referential use”), as well as insignia, portrait paintings, impersonations, and machine-made representations like photographs.40 The process of beholding through such representations has four parts. There are (i) the representata, (ii) the making of a de re artifactual representation about those representata— introducing a proper name, taking a photograph, etc.— and its then being made accessible to us, (iii) our accessing the representation— reading text, viewing an image, etc.— and (iv) our decoding it (typically automatically) and thereby coming to behold its representata. As in the perception case, part (iii) typically gives us all the input we need. As long as nothing actually went wrong in part (ii), we can behold the representata without knowing how the sausage got made. Now consider a work like Jane Austen’s novel Pride and Prejudice, which has no conspicuous narrator. The first chapter contains dialogue between Mr. and Mrs. Bennet, and occurrences of those names (outside quoted speech). In terms of the above division, (iii) by reading the text we we simulate reading a nonfiction text. (iv) We sim39. Linguistic expressions are de re in this sense iff there is something they refer to by what Salmon calls pure semantics (1993, 128–29). While typical names do this, in David Kaplan’s example, ‘Newman-1’ may not have the function of representing Newman-1, the first child to be born in the twenty-second century (cf. Salmon, ibid., 133, n 15). 40. Though displays, referentially used definite descriptions, and some other de re representations represent things as being a certain way, it is their de re representational function, rather than this attributive content, that determines their representata.
31
ulate understanding the text as if it were a nonfiction representation, and so coming to behold the bearers of those names. Thus, (i) we make believe people by those names are talking. Regarding (ii), the story does not prescribe we make believe much in particular, except what is obvious from the text, that ‘Mr. Bennet’ is someone’s name. Other cases prescribe making believe more about part (ii). For example, in Frankenstein, Frankenstein says, “I perceived, as the shape came nearer (sight tremendous and abhorred!) that it was the wretch whom I had created.” In a typical first-person narrative, we would make believe the speaker uttered the description ‘the wretch whom I had created’ as a de re representation of something, and simulate beholding the representatum by understanding that utterance. (Frankenstein is more complex: we make believe the story was written out by the ship captain to whom the main narrator recounted it. It probably leaves open whether ‘the wretch whom I had created’ were Frankenstein’s exact words or only a good approximation. Either way, someone uttered them as a de re representation of the monster, and by simulating understanding them, we simulate beholding the monster.) And in some cases we make believe part (ii) distorts things. We make believe the representation we access comes from an unreliable source, and simulate compensating for that by, for example, “reading people.” But mostly, we simulate beholding things without making believe anything much about how information about those things gets to us. If there are other ways of beholding things, there may be analogous accounts of their simulating such ways by simulating processing proximal stimuli. But the per-
32
ceptual and artifactual cases show well enough that we need not posit a Currie-style teller or super-presenter to make sense of our simulating beholding. Doing so would only fill in part (ii) of the above pictures, but all we really need is that the work not prescribe that we make believe the proximal stimuli are misleading in ways we do not automatically compensate for. Works that do prescribe this thereby prescribe simulating more complex processing of proximal stimuli— taking them with a grain of salt— or in some cases do not prescribe simulating understanding the proximal stimuli at all. But most works prescribe that we simulate beholding by simulating processing the real proximal stimuli they provide us (images, text, etc.) as if they were not part of a work of fiction. I suggest this is how Walton should analyze imagining having de re knowledge. I can now restate WCC in terms of prescribed beholding simulation: PBS A work generates a fictional object iff it prescribes that its consumers simulate beholding something other than an immigrant. I will support this criterion by summarizing how it covers the cases other criteria could not handle. PBS is a consumer criterion, not a fiction-maker criterion, so unlike SCC and CPR it applies equally to human-made and machine-made works. It also outdoes work criteria. For one thing, unlike formal criteria (and SCC), PBS applies directly to (wholly or partly) nonlinguistic works. Linguistic formal criteria associate generation with the occurrence in a work of certain singular terms, especially proper names. PBS ex-
33
plains generation with proper names (as above, in the case of Pride and Prejudice), and without them (as in the cases of Frankenstein and 2001). It also explains why works like 1 do not generate fictional objects; even though the name ‘Hugh Fitzhugh’ occurs in it, we make-believe it is empty, and so we should not simulate beholding its bearer. Unlike the content criteria NCC and NCΦ, PBS allows for fictionally nonexistent or beingless objects, like Yorick and Mariko. Hamlet prescribes we simulate beholding the bearer of the name ‘Yorick’, because it prescribes we simulate understanding Hamlet’s utterance of it. Likewise 2 prescribes we simulate understanding the occurrences of ‘Mariko’ (despite not making believe anything about who uttered them), and so beholding the name’s bearer. It is irrelevant that we make believe these are the names of nonexistent or beingless people. Finally, unlike MCC and MCΦ, PBS does not require that each work have a narrator, teller, presenter, or super-presenter. Indeed, Currie has the order of explanation backwards: we do not assume works have presenters in order to determine what objects they generate; rather, a work has a (native) presenter only if it generates such a fictional character, by prescribing that we behold someone who presents things to us. Since we can explain engagement and creation without positing such a character for every work, Currie has not given sufficient evidence that all works generate them.41 41. Nor have he and Moore given sufficient evidence that all works either generate native superpresenters or have immigrant super-presenters (namely, the works’ makers).
34
PBS, like many other creation criteria, only says when a work generates a object. It does not map fictional objects one-to-one to creation conditions, so it is difficult to test on works that generate multiple fictional objects (cf. SCC/SCC′ and the content criteria in §3.2). But there is an analogous specific-creation criterion, appealing to simulating taking things beheld in two ways (e.g., perceptually and by name) to be identical— in other words, taking two ways of beholding to be ways of beholding the same thing.42 PBS′ A work generates exactly one fictional object for each (maximally inclusive) class of ways of beholding things (other than immigrants) that it prescribes we simulate forming, such that it also prescribes we simulate taking them to corepresent. For example, when reading the first Batman story, we simulate beholding a masked crime-fighter, and beholding a rich socialite. On reaching the surprise ending that reveals Batman is Bruce Wayne, we simulate recognizing one man as the other, taking the two ways of beholding to corepresent. (I assume here that in simulating forming a way of beholding something, we also really form one, and that ways of beholding, including those formed in simulation, can be shared.43 Those who doubt these assumptions can replace classes of ways of beholding things with classes of a work’s 42. There are multiple philosophical and psychological literatures on this recognitional “taking.” See, for example, Salmon (1986) and Burge’s (2010, 411-12) connection of “Fregean informative identities” to perception science. 43. Compare Burge (2009, 248–49). There is a further question whether these are (a) essentially empty mental representations, or (b) as a result of generation— not a part of it, by IC— real ways of beholding fictional objects. I discuss this in chapter two.
35
prescriptions to simulate forming ways of beholding, such that it also prescribes simulating taking the resulting ways to corepresent.) PBS, more than any other creation criterion, fits our intuitions about when works generate fictional objects. Because it appeals to the proper interpretation of whole works, and covers works that generate their content and objects in different ways, it is more aesthetically sophisticated than most criteria. And because it appeals to simulation and the psychology of beholding, it is more psychologically sophisticated than Walton’s imagining-based criterion, WCC. If creationism is right, PBS states the (initial) existence conditions for fictional objects, and so it connects a popular metaphysics of fictional objects as cultural artifacts substantially with aesthetics, philosophy of mind, and psychology. And even if creationism is wrong, PBS embodies creationist intuitions that Platonists and antirealists must explain away. Knowing how fictional objects originate helps us understand their existence and identity, and lays the groundwork for an account of the conditions for representing them, in fiction, engagement, and outside any make-believe or simulation. I discuss that in chapter two.
36
Chapter 2 Real Representation of Fictional Objects 1 Introduction There are many stories about Sherlock Holmes. Or are there? Are the stories really about him, or are they about nothing? In general, lots of things seem to represent fictional objects— not just stories, but names, pictures, mental states, etc. But do these apparent representations of fictional objects (ARFOs for short) really represent fictional objects? This is not a metaphysical question, although it may be tempting to read it as one. Most philosophical discussions of fictional objects (outside aesthetics) are about metaphysics, especially ontology. And of course, if there are no fictional objects to represent, then none are represented; and contrapositively, if some are represented, then there are some. But being and being represented are quite different, and I am asking about the latter. Besides the work on ARFOs that is really about metaphysics rather than representation, at lot of work simply assumes that they represent fictional objects, or that they do not (e.g., that fictional names do not refer to anything). The work that most substantially treats my question assumes a certain metaphysics of fictional objects— either endorsing it or granting it for the sake of discussion— namely, creationism (aka artifactualism). That is the intuitive view that fictional objects are artifacts, really created in making works of fiction, and so really existing. They are not real people or animals or whatever, and do not really have a spatial location; in
37
that sense they are abstract, like certain other cultural artifacts such as languages, laws, and limericks. So I will assume creationism in the following. (In fact I accept it. And I think even noncreationists should consider it our starting point for the philosophy of fiction, to be revised or “explained away” only with good reason.) All creationists agree that some ARFOs represent fictional objects; e.g., fictional names in “discourse about fiction” such as ‘Conan Doyle created the fictional character Sherlock Holmes’. But there are some disputed cases, which some theorists say do represent fictional objects, and some say do not. Those cases include: 1 negative existential statements such as ‘Sherlock Holmes does not exist’ 2 fiction content reports, such as ‘Holmes solved the case’ or ‘According to the story, Holmes solved the case’ 3 fiction-consumers’ mental representations such as my admiration for Holmes, or reports of them, such as ‘Manning admires Holmes’ 4 works of fiction (novels, comic books, movies, etc.)1 5 fiction-making activities (e.g., writing the text of a novel, creating fictional objects) 6 fictional names such as ‘Sherlock Holmes’, as they occur in the above sorts of cases For example, Saul Kripke (according to commentators; see §2.1) seems to think 1 1. I will ignore works’ apparent representation of immigrant objects, such as The Hound of the Baskervilles’s representation of Sherlock Holmes (whose native work is the earlier A Study in Scarlet) or of London (which is a nonfictional object). Representation of nonfictional immigrants is unproblematic (even if a few philosophers say otherwise); and representation of fictional immigrants is more complicated than the core case of representation of natives. I will discuss them elsewhere.
38
and 4 do not represent fictional objects, because fiction or fiction-making introduces a nonreferring use of fictional names; since the names also have a use referring to fictional objects, he says 6 and 2 are ambiguous, in one sense representing fictional objects and in another sense not. On the other hand, Nathan Salmon says all those cases unambiguously represent fictional objects. To discuss all of 1–6 in detail would take a long paper with a lot of detours, especially for 1 and 2. But there is another way. The theorists in this dispute agree that a certain ARFO represents a fictional object if and only if that ARFO has a certain history. For example, the content report ‘Holmes is a detective’ includes an occurrence of ‘Holmes’. This is not a new, ad hoc use of the name, but one with a history. That history might show that the name on this use does not refer, in which case neither does the content report; or if the name on this use does refer to a fictional object, then so does the content report. So this is one sort of argument that certain ARFOs do not represent fictional objects: A1 Certain ARFOs reuse a representation according to its preexisting semantics, and thus whether such an ARFO really represents a fictional object depends on that semantics.2 A2 According to that preexisting semantics, the representation does not represent 2. There is an irrelevant sense in which we reuse names, namely introducing new instances of preexisting generic names (Kaplan, 1990, 110 ff). For example, instances of the generic name ‘David’ were introduced for David Hume and for David Kaplan, and instances of the generic name ‘Sherlock’ were introduced for the 18th-century bishop Thomas Sherlock and for Sherlock Holmes. But unlike their instances, generic names as such have no semantics (or no univocal semantics), and thus to introduce a new instance of one is not to reuse a representation according to a preexisting (univocal) semantics. (Thanks to Salmon for bringing generic names to my attention.)
39
fictional objects. .·.A3 Certain ARFOs do not really represent fictional objects. Since the argument is valid and everyone accepts A1, the key is justifying A2. There are two kinds of arguments for A2, depending on which kind of prior representation is at issue, either fictional names or a “pretense.” As I will show, the published arguments are unconvincing. Then do all ARFOs really represent fictional objects? I am not sure. But at the end of the chapter I will sketch some issues relevant to finding out. 2 Fictional names More than any other kind of ARFO, fictional names are the focus of philosophical discussion of fictional objects. This is due more to trends in the philosophy of language than to the degree of importance of names to fiction. I argued in chapter one that names are inessential to generating fictional objects. Names are also inessential to representing them. Most works of fiction are partly or wholly nonverbal — in media such as movies, television, paintings, sculptures, plays, comic books, etc.— and even the verbal ones can fail to name their natives (e.g., Frankenstein’s monster is unnamed in Mary Wollstonecraft Shelley’s novel). Plus we can refer to fictional objects by description, e.g., as “the one introduced in chapter 12” (cf. Peter van Inwagen, 1977, 307). But let us ignore that for now. Two arguments for A2 take the prior representation to be fictional names: one attributed to Kripke, the other from Amie
40
Thomasson.3 2.1
Kripke on fictional names
Kripke says that “to write a work of fiction is to imagine— spin a certain romance, say— that there really is a Sherlock Holmes, that the name ‘Sherlock Holmes’ as used in this story really refers to some man, Sherlock Holmes, and so on” (2011, 58).4 In other words, works that introduce fictional names involve a pretense that those names satisfy the normal conditions for names to refer. “If this is so, the name, of course, doesn’t really have any referent, it is pretended to have a referent” (59). The other use of fictional names, to really refer to fictional objects, is “parasitic on” this prior use (1973a, 3.25). And it is the nonreferring use that is invoked in content reports (on one of his analyses) and in negative existentials. For example, in uttering ‘Holmes really exists’, either as a content report or to deny it in a negative existential, “I am using the sentence to express an alleged proposition about a detective, given to us by the story,” or more precisely “pretending to express a proposition,” even though really there is no such proposition (2011, 68; cf. 61, 64–65; and 1973a, 3.3–5, 4.24, 6.26–31). And since ARFOs such as content reports and negative existentials invoke a prior, nonreferring use of fictional names, those ARFOs do not represent fictional objects. As 3. I will mostly ignore the question whether occurrences of expressions in a work’s text (such as occurrences of ‘Holmes’ in Conan Doyle’s works) represent fictional objects. A few theorists in this literature discuss it (e.g., van Inwagen 1977, 301; and Salmon 1998, 299, 301), but resolving it would take a deeper study of the function of those utterances— deeper than any in this literature, and more aware of the issues I mention in §4. (I will not even discuss it indirectly, because while van Inwagen denies that those utterances represent fictional objects, he does not explicitly deny that any other ARFOs do.) 4. Of course fiction-making is not just imagining; it is really making an artifact. I take it he means that imagining is part of fiction-making.
41
Salmon puts it: On this account, the name ‘Sherlock Holmes’ is ambiguous. In its original use as a name for a human being — its use by Conan Doyle in writing the fiction, and presumably by the reader reading the fiction— it merely pretends to name someone and actually names nothing at all. But in its nonpretend use as a name for the fictional character thereby created by Conan Doyle, it genuinely refers to that particular artifactual entity. In effect, there are two names. Though spelled the same, they would be better spelled differently, as ‘Holmes1’ for the man and ‘Holmes2’ for the fictional character. Neither names a real man. The latter names an abstract artifact, the former nothing at all. (1998, 294) Several other commentators agree: Kripke’s view is that fictional names have uses of two kinds, one to refer to a fictional object, the other nonreferring.5 So this is an argument for A2: one kind of use of fictional names in certain ARFOs is not the kind on which they refer to fictional objects; it is another kind, originating in pretense, and on which the fictional names do not refer. I will grant that if there is such a kind of use besides the one to refer to fictional objects, it is a nonreferring one. The question is whether there is such a kind of use at all, e.g., whether there is a use of ‘Holmes’ besides the one on which it refers to the fictional character Holmes. Gareth Evans and Salmon say Kripke has not shown there is one. Evans says that if there is such a use of ‘Holmes’, “we are entitled to ask for an account of what is required for understanding” ‘Holmes’ on that use, which Kripke has not given us (1982, 350). But I do not see why we need an account of understanding the name. Evans distinguishes understanding a name on a given use— which requires thinking of the referent (400)— from “name-using practices,” which one can participate in without 5. See Richard Routley (1980, 564), Gareth Evans (1982, 349–50), Salmon (1998, 294–304), Alan Berger (2002, 98, 151), and David Braun (2005, 624 n 21).
42
understanding the name (398–402). As far as I can tell, at most we are entitled to is an argument that there is such a name-using practice, not to an account of understanding. Salmon’s objection is about name-using practices: “The problem with saying that ‘Sherlock Holmes’ is nonreferring on Conan Doyle’s use is that in merely pretending that the name had a particular use, no real use was yet attached to the name on which it may be said to refer or not to refer” (1998, 299). Two points need clarifying. First, what does Salmon mean by saying that Conan Doyle’s “use” of ‘Holmes’ was not a “real use?” A “real use” is what Salmon elsewhere calls a “correct use” of an expression “from the point of view of pure semantics” (2004, 320–21; emphasis removed); I will call it a conventional semantic use (CSU for short). A semantic convention for an expression assigns to it a semantic value, or conditions determining which thing is the semantic value. For example, one semantic convention assigns to ‘London’ as its value the city of London, England; David Kaplan’s convention assigns to ‘Newman-1’ the condition being the first child born in the 22nd century (Kaplan, 1969, 201). Aside from a CSU, ‘use’ can also mean what I call a mode of utterance, something we do with a linguistic expression. Modes of utterance include uttering an expression on a certain CSU. But other modes have no built-in semantics, such as my (hypothetical) use of ‘Holmes’ as a password. Salmon is saying that Conan Doyle’s “use” is merely a mode of utterance, and not a CSU. Second, if Conan Doyle’s use is not a use on which ‘Holmes’ refers, is it not therefore a use on which it does not refer— exactly as Kripke allegedly holds? As Salmon
43
seems to understand Kripke’s view, the nonreferring use would have to be a CSU (more on that below). But as I just explained, Conan Doyle’s “use” is a mode of utterance rather than a CSU, and so, Salmon apparently argues, Kripke does not show there is a nonreferring CSU. Kripke might respond that Conan Doyle’s mode of utterance, though not itself a CSU, generates a nonreferring CSU for ‘Holmes’. Conan Doyle (we are supposing) pretended that ‘Holmes’ rigidly designates a man (who is a detective, etc.); a name could have a CSU to refer to something fitting that description. (More generally, a name could have a CSU to refer to whatever is identified in a way Conan Doyle pretends to identify something, whether he does so by description or otherwise.) The problem is that pretending anything about a name need not really establish semantic conventions about it. Kripke has not established, or even made plausible, the claim that Conan Doyle’s pretense would establish such conventions. So this argument for A2 is unconvincing. But I doubt Kripke accepts the argument. Judging from his 2011 and 1973a, he does not claim that fictional names are ambiguous, or specifically that the nonreferring “use” is a CSU, rather than a mode of utterance.6 He does say that sentences like ‘Holmes exists’ have “multiple ambiguities” (2011, 68 n 30), but he does not say that one of those ambiguities is fictional names’ multiple CSUs. The only expressions he says are ambiguous or have multiple (kinds of) uses are predicates and whole sentences. As far as I can tell, everything he says about the nonreferring “use” 6. Salmon’s reading of Kripke on this issue may depend on sources besides those lectures (1998, 314 n 32).
44
of fictional names, including the passages I quoted earlier in this section, can be understood in terms of modes of utterance that are not CSUs. Thus I have no evidence from Kripke’s lectures that he accepts the above argument for A2. On the other hand, he probably does accept one of those I discuss in §3, so we will see him again soon. 2.2
Thomasson on fictional names
Thomasson too says that ARFOs such as content reports and negative existentials fail to represent fictional objects because in those ARFOs we use fictional names on a nonreferring use (2010, 123–25). She says a bit more than Kripke about what distinguishes the referring and nonreferring uses of fictional names, and so a bit more in support of A2. Unlike Kripke, she says that proper names have some descriptive content. It is not a full-on, reference-fixing description, but rather what she calls a “basic application condition” (BAC) which associates the name with “a highly general category of entity to be referred to” (2010, 118). These BACs are “a genuine part of the conceptual requirements associated with the name, not mere referencefixers” (119; cf. her 2007, 39–40, 42–44). But like a reference-fixing description, for the name to refer to something, that thing must satisfy the name’s BAC in the context of the name’s introduction. In the case of fictional names, she says they have two BACs: one for their referring use, where the BAC’s category is something like fictional object; and one for their nonreferring use, where the BAC’s category is (in her examples) person.7 She says: 7. Of course, it is person only for the nonreferring uses of fictional person-names. For the nonreferring uses of fictional fictional-object-names, such as ‘Itchy’ and ‘Scratchy’ from The Simpsons (according to which those are names of a fictional cartoon mouse and cat) Thomasson might say the
45
There . . . is a tradition of using the name as a person-name rather than as a character-name: the tradition readers engage in when they discuss what happens in the story. There, the name is used as a person-name, it’s just that the readers do not genuinely intend to refer using the name in that way; they only pretend to make assertions involving it. But while the force of their utterances differs from that of an assertion, the meanings of the terms used in them (including the meaning of the name — the application conditions with which it is associated) don’t change. (123; cf. the early version in her 2003b, 215–18) In earlier work she says that consumers continue a pretense begun in “the fictionalizing discourse of the storyteller” (2003b, 207), and probably she still holds that fiction-makers begin such nonreferring traditions of use of fictional names. So A2 is true because some occurrences of fictional names in ARFOs continue the tradition of use of them as, e.g., person-names, rather than fictional-object-names. The problem is much like the one discussed in §2.1: Thomasson has not shown there is a kind of use of fictional names besides the one on which they refer to fictional objects, for example a kind of use of ‘Holmes’ as a person-name. Her discussion suggests two defenses of that claim. First, a “tradition of use” associates the name ‘Holmes’ with a BAC whose category is person, and if a name has a BAC then it has a “meaning,” even if it lacks a referent. But it is doubtful that names have BACs at all. Her evidence that they do is that it explains why names do not refer if nothing of a certain category was suitably related to their introduction. Her only examples are cases where someone introduces a name under a false assumption about what is around to be referred to. “If, e.g., a speaker attempts to ground the reference of a name ‘Squeaky’ as the name for a mouse behind the wall, and the noise was caused category is fictional object; for their referring uses, maybe it is fictional fictional object.
46
only by a short in the electrical system, ‘Squeaky’ does not refer (not even to the wall or the cord), since there is nothing the speaker is causally related to that is of the category (animal) the speaker intended — it is this sort of mistake that seems critical to whether or not the term refers” (2010, 118; cf. 2007, 39–40, 49–50). I agree that in case of such a mistake, the speaker failed to introduce a use of the name referring to anything in the environment.8 But this is not because the category animal is part of the conceptual or semantic content of the name. That confuses the name’s semantics with conditions for introducing the name to refer to something that one beholds (in the sense I introduced in chapter one). To behold something (i.e., to mentally represent it de re), one (or some sub-personal, representation-forming part of one, such as a module in one’s perceptual system) must correctly represent it in some respect. At least, that is what Tyler Burge argues for the case of perceptual beholding, in which one forms a representational state S of entity E: “For it to be psychologically explainable how the individual or perceptual system discriminates E, some of E’s properties that help cause S . . . must ground an explanation of how the perceptual system perceptually discriminates the particular E” (2009, 294). In the ‘Squeaky’ case, something like the following might happen: the speaker’s auditory system, or some part of it, has the function of generating representations of animals (or maybe animate objects) by processing stimuli that they cause. But the stimulus processed was not caused by an 8. I am not sure he fails to introduce any use of the name on which it refers. Since he was not making fiction but instead believing a myth, he may have accidentally introduced a use of the name to refer to a mythical object (somewhat like a fictional object, but generated by a myth), which he wrongly thinks is a mouse (see Salmon, 1998, 304–6). But Thomasson might not agree, because the referent would belong not to the speaker’s intended category mouse, but to mythical object.
47
animal, and so the representation generated fails to represent anything. The speaker tries to introduce a use of ‘Squeaky’ to refer to what he beholds by having that representation, but he does not behold anything, so he fails to assign ‘Squeaky’ a referent. None of this shows that animal becomes part of the name’s semantic or conceptual content: subsequent users of the name need not know how its referent was originally identified, but only which use of the name they employ (cf. Burge, 2009, 309 n 67). Names (on uses) have no BACs, so ‘Holmes’s allegedly having two BACs is no evidence that it has two uses. Thomasson’s discussion also suggests a second argument that there is a use of ‘Holmes’ besides the one referring to the fictional character: there is “a tradition of using the name as a person-name,” a tradition on which the name has a meaning. (This argument does not appeal to names having BACs.) The problem (as in §2.1) is that this ignores the difference between modes of utterance and CSUs. A “tradition” could be a way things have been done, or a way things are to be done. Ways a word has been used are modes of utterance, but (certain) ways it is to be used are its CSUs. Using a name in pretense is a mode of utterance, not a CSU. Nor do we have a reason to think the mode of utterance establishes a CSU (other than the one referring to a fictional character). Even if all that took was “associating” the name with a category (as she puts it), the bar is still too high. ‘Associate’ is ambiguous between normative and nonnormative senses; normatively, the things are to be together, while nonnormatively they simply are together. Modes of utterance always establish nonnormative
48
associations, but what a CSU requires is a (semantically) normative association. We have yet to see any argument that pretenses about names establish real semantic conventions for those names, other than the CSUs on which the names refer to fictional objects. So Thomasson’s argument for A2 is no better than the Kripkean one. 3 The work’s pretense The first two arguments for A2 assumed that ARFOs such as content reports and negative existentials employ a real nonreferring use of fictional names, besides their use to refer to fictional objects. The second batch of arguments assume that certain ARFOs reuse a sort of pretense associated with the relevant work— call it the work’s pretense. For example, a typical utterance of ‘Holmes is a detective’ should be understood as not really asserted, but as asserted within a certain work’s (or works’) pretense (according to which it is a true assertion). And ‘Holmes does not (really) exist’ should (sometimes) also be understood as engaging in that pretense, in order to reject it as mere pretense, in order to correct or prevent a confusion of fiction with reality.9 Theorists understand the work’s pretense in various ways, for example as a fiction-maker’s pretense, or as consumers’ pretense, which either continues the fiction-maker’s or is otherwise guided by the work. Such variations will be unimportant to the following arguments that the work’s pretense does not represent fictional objects. 9. Variations on these analyses are popular among noncreationists as well as creationists. Kendall Walton’s work (esp. 1973 and 1990, chs 10–11) is a classic source.
49
3.1
General content
The first kind of pretense-based argument for A2 is this: when we state the content of a work’s pretense we can see it does not really include native fictional objects. Specifically, the content is always that there are things having certain properties, and so is not content of specific fictional objects that they have certain properties. Call this the general content argument.10 Several theorists take for granted such a view of the content of the work’s pretense. [We imagine] that the name ‘Sherlock Holmes’ as used in this story really refers to some man. (Kripke, 2011, 58). [When Iris] Murdoch uses a proper name [in her novel] . . . she pretends to refer . . . [and thereby] pretends that there is an object to be referred to. (John R. Searle, 1975, 71) [In writing a novel] Le Carré is pretending, or making as if, he’s referring to a real man named ‘Jonathan Pine’, and he’s pretending, or making as if, the rest of the sentence is telling us something about this man and hence that it is making a true or false statement about him. (Stephen Schiffer, 1996, 155) It is natural to take the sentences in works of fiction to be involved in pretense of de dicto form — so that the sentences of the Holmes stories do not pretend of Holmes (the fictional character, an abstract artifact) that he was a detective; instead, the pretense has the form: There once was a man, such that he was called “Holmes”, was a detective, was very clever, etc. . . . Such views are intuitively plausible, and seem to better capture the psychology of writing. (Thomasson, 2003b, 212) Those who take part in such a process . . . make believe that there is an individual, typically a concrete one, that (explicitly or implicitly) has a certain set 10. A few sophisticated noncreationists say the work’s pretense in some sense simulates having content about specific things, other than immigrants. They reduce this either to our making believe general propositions (Gregory Currie, 1990, 151–54, 165–68), or to our following the work’s prescriptions for how to pretend or make believe, which themselves do not refer to fictional objects (Walton, 1990, 135–36, ch 10). In chapter one I argued that the first misrepresents how we engage with fiction. On the second claim, see below, note 11.
50
of properties. In the terms used above, they are pretending de dicto, not de re. (Alberto Voltolini, 2006, 74) (These theorists are creationists. And nearly all antirealists hold this sort of view.) The problem is that we cannot take this for granted. The above descriptions of works’ pretenses are intuitively correct, but it is also at least as intuitive that works’ pretenses are about (e.g.) Holmes, the fictional character that Conan Doyle created.11 We need more evidence to show that works’ pretenses have only general content (and perhaps singular content about immigrants), evidence that might be acceptable even to those who deny A2. I consider some purported evidence in §§3.2–3.3. 3.2
Mismatch
The second kind of pretense-based argument for A2 is that the work’s pretense does not represent fictional objects because we do not pretend that fictional objects, which are really abstract artifacts, are (e.g.) people. Call this the mismatch argument. Why think we do not pretend this? Salmon (who does not accept the argument) says that some theorists think fictional objects “are not suited” to be objects of a pretense that they “are living, breathing people” (1998, 301). Thomasson finds it “odd” and “awkward” to say that “works of literature . . . invoke the pretense, of some abstract object, that it is a detective, is a man, solves crimes, etc.” (2003b, 212). And some critics of creationism say that if creationists’ metaphysics were right, then for this reason the work’s pretense would not be about fictional objects. For example, Frederick 11. Likewise, Walton says works prescribe how we should pretend in ways that do not refer to fictional objects, but it is at least as intuitively acceptable that works prescribe that we pretend about (e.g.) Holmes, the fictional character that Conan Doyle created. I will return to this briefly in §4.
51
W. Kroon says “that would make little sense, since fictional objects on this sort of model [i.e., on Kripke’s view] are abstract existent objects” (1992, 523). And R.M. Sainsbury says that if we imagine about any objects, “the objects need to be ones that undeceived writers and readers can use as targets for their imaginations. That means that the nature of the objects should not be resistant to the possession of the properties imagination ascribes” (2009, 113). This argument is pretty obscure. Why think we do not pretend of the fictional character Holmes that he is a detective? These theorists may be trying to counter this sort of argument: 1 We pretend Holmes is a detective. 2 Holmes is an abstract artifact. .·.3 We pretend an abstract artifact is a detective. (Call this the substitution argument.) Since they deny 3 and accept 2, they deny 1. But is this a good reason to deny 1, and so to support A2? Salmon thinks not. He says this confuses de dicto and de re. “That abstract entities are human beings is not something we pretend, but there are abstract entities that we pretend are human beings” (1998, 316 n 45). In other words, 3 is ambiguous between these two readings: 3a We pretend, of an abstract artifact, that it is a detective. 3b We pretend that an abstract artifact is a detective. Read as the de dicto 3b, 3 is false, and so no sound argument has that conclusion. But the argument is unsound on that reading because it is invalid: 1 and 2 do not entail 3b.
52
Instead they entail the de re 3a. Is 3a objectionable? That is not obvious; maybe the argument is a good case for 3a, rather than evidence that 1 must be false. So only on the de re reading is the argument valid, and only on the de dicto reading is the conclusion clearly false; on neither reading are both the case, so on neither reading is the argument evidence against 1. So this worry about the substitution argument is poor support for the mismatch argument. Salmon’s diagnosis of the mismatch argument may be right, but it does not convince Thomasson (ibid.) or Sainsbury (98). So I will consider other explanations. Another possible motivation for the mismatch argument is fictional inerrancy (inerrancy, for short), the view that whatever is fictionally true about fictional objects is really true about them — for example, that Holmes is really a detective. Those sympathetic to inerrancy, confronted with the substitution argument, could easily avoid the conclusion by denying that Holmes is an abstract artifact.12 Officially, this is something creationists deny, because they say Holmes is an abstract artifact, and deny that any abstract artifact is a detective.13 But inerrancy is intuitive, in fact so intuitive that it is easy to slip into without realizing it, even for someone such as Thomasson (or 12. Compare Sainsbury, who discusses what Kit Fine (1982, 97) calls literalism, the view that content reports like ‘Holmes is a detective’ are literally true, not shorthand for reports like ‘Fictionally, Holmes is a detective’. He says that literalist could more easily make sense of our pretending that Holmes is a detective. But literalists such as Terence Parsons (see the next note) and Routley (1980, 573) do not appeal to pretense— if something is really true, why pretend it is? This means they would balk at the substitution argument as stated. But they would be more sympathetic if we changed ‘We pretend’ to something like ‘According to a work’. Then they would avoid the conclusion (and many similar claims; see Routley 1980, 562–63) by denying that Holmes is an abstract artifact. 13. There is another reason to reject unrestricted inerrancy: like naive set comprehension, it is inconsistent: since there are works with inconsistent content, fictional objects appearing in them would really have inconsistent properties. But like set comprehension, inerrancy has consistent, restricted versions. See, for example, Parsons (1980, 175).
53
me) who officially rejects it. (It is as tempting as use/mention confusions are to many theorists. But both are mistakes.) If our only intuitions for the mismatch argument are pretty vague, maybe they lie in this confusion. Sainsbury’s discussion suggests another motivation for the mismatch argument. “Conan Doyle . . . doesn’t regard [the proposition that Holmes wears a deerstalker] as a ludicrous falsehood, ascribing a property of a kind that the object of the ascription simply could not have” (113). So maybe the reason we do not pretend of fictional objects that they are (e.g.) people is that those propositions would be clearly impossible. If this is what Sainsbury has in mind, he is ignoring Salmon and Thomasson’s purported counterexamples. Salmon says we pretend “that Marlon Brando is Don Corleone” (1998, 316 n 45; cf. 301–2), and Thomasson admits that “similar crosscategory ascriptions are made all the time about real people in discussions of theatrical performances, where we may say de re, of an actor, that we are to pretend that he is a cat, or an angel, or even (in Pirandello’s Six Characters in Search of an Author) a fictional character, (in a morality play) [the property of] Charity, or (in a children’s play about mathematics) the number three” (2003b, 212). Do these examples show Sainsbury is mistaken? Not quite. I am not sure we normally pretend anything about the actors in a work (or performance of a work). Their contribution may be like that of set decorators, directors, and authors: to provide us with stimuli that guide us in a pretense that is not about them.14 14. Salmon seems to motivate the analogy partly by an ambiguity. In a familiar sense, actors play roles or parts in works or performances; he says that likewise fictional objects play or occupy roles in works, and are parts of works (1998, 301–2). But in the special dramatic sense, the part an actor plays
54
But there is a better analogy. We do pretend impossible things about immigrant objects. There is fiction about real people, according to which they do impossible things. For example, in reading Graham Priest’s short story Sylvan’s Box (1997, 573– 79) we pretend of the philosopher Richard Sylvan (aka Routley) that he had a box that was both empty and nonempty. But the sort of impossibility Sainsbury mentions involves not unexemplifiable properties, but only properties incompatible with a certain kind of thing. So consider some of the educational puppetry on Sesame Street: in watching this we often pretend that numbers (or numerals) and letters are animate, which they cannot really be. If we reflect on it much, it is hard to keep up the pretense. But pretending unreflectively, we have no trouble. We should expect that pretending about a fictional object is like this. In fact, it might be easier to pretend about fictional objects than about numbers and other nonfictional objects, because fictional objects are made for that purpose (cf. Salmon, 1998, 302). In pretending the number one is animate, or that Socrates is a crime-solving robot dog, we might still think about the objects’ “baggage,” the properties we are used to representing them as having. (Some works make it relatively easy to ignore irrelevant baggage; some do not, and we have trouble engaging with them.) But we would be used to representing is a role (a roleD ), and is a fictional object (normally a fictional character). And we do not say actors occupy rolesD ; they only play them. On the other hand, actors can occupy roles in a nondramatic sense (rolesN ), namely to contribute in a certain way to the performance. For example, one actor may have the roleN of playing a certain character, while another has the roleN of comic relief; actors normally play many rolesN , but rarely multiple rolesD . So the analogy between fictional objects and actors is weak: fictional objects are rolesD , while actors play them; and though both fictional objects and actors have rolesN , I am suggesting that only the former have rolesN of being things we pretend about. (Some noncreationists, such as Currie (1990, 171–80), reduce rolesD — i.e., fictional objects — to rolesN . But this is not Salmon’s view.)
55
Holmes as a detective, while pretending with Conan Doyle’s works; the fictional character has little baggage for us (except maybe for those rare people who are more used to representing him— it — as a cultural phenomenon, one of the biggest influences on 20th century fiction, a piece of intellectual property, etc.). So although it is impossible that Holmes be a detective, we still might pretend he is. There may be other motivations for the mismatch argument, but none of those suggested in the literature are convincing. So the argument fails to support A2. 3.3
Generation
The last argument for A2 is that a work’s pretense is not about (native) fictional objects because it generates or creates them. Call this the generation argument.15 For example, Kripke says “one should not say that, when one is just pretending to refer to a man though really one is not, that that pretence was in and of itself naming a fictional character. That was creating a fictional character” (1973a, 6.19). Thomasson seems to agree (2003b, 212). Voltolini’s view is similar: he says the work’s pretense is a merely necessary condition for generating fictional objects, because they are only generated when someone reflects on the pretense; therefore the pretense does not represent them (2006, e.g., 157). Others agree that what generates a fictional object does not represent it, although they do not say that what generates it is the work’s pretense, in the form reused in ARFOs.16 15. In the literature the term ‘create’ is more common than ‘generate’. But in ordinary use we would usually reserve ‘create’ for an action, not an effect of there being a certain work or pretense; so here I will say that works, pretenses, etc. generate fictional objects. 16. See Searle (1975), van Inwagen (1977, 305–8; 1983, 73, 75), Emt (1992, 154), Schiffer (1996, 156–57; 2003, 52), Salmon (1998, 300, 316 n 43), Thomasson (1999, 13), Jeffrey Goodman (2004,
56
But why would that be impossible? I have seen two explanations. Searle says a fiction-maker, in pretending to refer, “does not really refer to a fictional character because there was no such antecedently existing character” (1975, 71; my emphasis). He and several other theorists write as if this were a matter of time, that the fictionmaker cannot refer to something that does not yet exist.17 I do not see why not. Like Kaplan, I think we can refer to future objects that we can uniquely specify. We can specify things such as Newman-1, and the future products of “a certain closed, developing, deterministic system,” such as an assembly line, and so we can refer to them (Kaplan, 1973, app XI, n 19). Salmon applies this to fictional objects, and even to fiction-makers’ reference to them: “Conan Doyle may have used the name [‘Sherlock Holmes’] for a period even before the character was fully developed. . . .There would soon exist a fictional character to which that use of the name already referred” (1998, 301). Finally, a nontemporal statement of the restriction seems adequate: a representation’s objects must be independent of it.18 So I doubt that time is the problem. Kendall Walton gives another explanation. (He writes in terms of works, rather than works’ pretenses. His explanation does not turn on the difference, so I will translate.) Representation and creation are distinct, but in fiction-making, only one thing seems to be done to fictional objects; therefore works’ pretenses do not both represent and generate their native objects (1990, 127). I add that if creationists are com144–47). And some noncreationists assume this is part of creationism, including Nicholas Wolterstorff (1980, 136), Edward N. Zalta (2000, 143), and Sainsbury (2009, 97). See also note 18. 17. See also Wolterstorff, Thomasson (1999), Zalta, Voltolini, and Sainsbury, cited above. 18. See Kroon (1992, 523), Walton (1990, 127), and again Sainsbury (2009, 95).
57
mitted to works’ pretenses generating fictional objects, they should conclude that works’ pretenses do not represent their native objects. But why does Walton say that works’ pretenses (or works) do not both represent and generate their native objects? He says “it is not clear that once we say that Moby Dick [or its pretense] creates Ahab there is any point in going on to say that it denotes him. The addition seems gratuitous. . . .There do not seem to be two distinct functions, creating and denoting, which require independent analyses. What would it be like for a work [or its pretense] to create its characters without denoting them?” (ibid.). This is a mistake: if a work’s pretense did both generate and represent its natives, it would not follow that it could do either of those separately; distinct functions can be inseparable. And that seems to be the point of Robert Howell’s response to Walton: “If, by extending your finger, you bring into existence an object at the moment your finger is fully extended, it does not follow that your finger cannot also at that moment be pointing at that object” (1997, 434 n 39). For an actual example, consider my uttering the phrase ‘this utterance’ (or ‘consider this very utterance’; and assume I utter it as an English expression in a context where it is the most salient utterance). By uttering that phrase, I both create an utterance and represent it. I could not have represented it without creating it, because there would have been nothing to represent. And given the meanings of those expressions in English and the contextual evaluation of the occurrence of ‘this’, my creating the utterance automatically results in my representing it (by producing a representation of it, namely itself). Of course this is an unusual case of representation;
58
most representations are not tied in this way to the generation of their represented objects. But this sort of case shows generation need not be entirely separate from representation. And maybe this is how a work’s pretense both generates and represents fictional objects, namely by being the sort of representation that will (a) generate objects, and (b) automatically represent whichever objects it happens to generate. So the fact (assuming it is one) that works’ pretenses generate fictional objects does not show they do not represent fictional objects, and so does not support A2. 4 Further work I have argued that the published cases for A2— and so the arguments that certain ARFOs do not represent fictional objects— are unconvincing. But this does not resolve whether any ARFOs do represent fictional objects. I will say a bit more on what must be done to answer that question. First, it is misguided to discuss the question in terms of “a work’s pretense,” or even specifically a fiction-maker’s or consumer’s pretense. As I argued in chapter one, whether a fiction-maker or consumer pretends anything is irrelevant to whether a work generates a fictional object. That depends on how the work guides consumers’ imaginative simulations (i.e., on what simulations it is made to guide us to have), whether or not anyone does engage with the work, and whether or not any fictionmaker pretended anything. Furthermore, the sort of pretense most of the above theorists (except Walton) have in mind is propositional; but I argued in chapter one that a
59
work’s propositional content underdetermines whether and how it generates fictional objects. Pretense is not essential to generating fictional objects, so it may not be central to explaining representation of them. Second, we need an account of how anything can represent fictional objects. Can this only be done by description (or by a representation whose reference is fixed by a description)? That seems to be the view of Voltolini (2006, 139–40) and noncreationists such as Gregory Currie (1990, 171–180) and Edward N. Zalta (2003). Or can we represent fictional objects nondescriptionally, for example mentally— i.e., can we behold them? If so, we do not do so perceptually, or in other ways familiar in the literature on beholding. Probably we would do it much like how we behold other kinds of abstract artifacts; but it is not obvious how we do that. With some sort of account of how anything can represent fictional objects, we could start answering the questions I started with in §1. Do works represent their native fictional objects by guiding us to simulate beholding things— i.e., in the very guidance that generates those objects? As far as I can tell, this is logically possible for the reasons I discussed in §3.3. But is it the case? If works do not represent fictional objects in this way, is there some other way they do it? Is there some other way that consumers’ mental states in engagement represent fictional objects? If not, how do we move on to represent them outside engagement, as we often do? Which of these representations compose or are reused in which ARFOs? For example, in reporting a work’s content, do I reengage in the simulation it prescribes? And what would that
60
entail about whether the report (or I) represents fictional objects? I have not shown how to answer these questions, but I have at least shown a bit of their proper context. That is a start.
61
Chapter 3 No Identity without an Entity 1 Introduction Peter Geach’s puzzle of intentional identity is to analyze this sentence: G Hob thinks a witch has blighted Bob’s mare, and Nob wonders whether she (the same witch) killed Cob’s sow. (1967, 147) Specifically, we must analyze it as reporting intentional identity— roughly, as reporting that while Hob and Nob in some sense have attitudes about the same witch, there is no witch (in fact no entity) they have attitudes about. It seems G can be read that way, but it is surprisingly hard to explain how. The puzzle has drawn dozens of responses, as varied as remedies for hiccups, and often as odd. The variety comes partly from disagreement on what the puzzle is even about. Several delicate phenomena are involved, including a certain kind of anaphora and intentional identity, and it is not obvious which is the active ingredient. Not even Geach is sure. He presents the puzzle as part of a program to analyze certain problematic “intentional” constructions, but his class of cases varies from paper to paper, as do his approximate and impressionistic descriptions of the problem. In other words, the problem is as much philosophical as semantic or logical: his questions needs clarifying before we can seriously evaluate any answers. This makes it easy to read into his papers— and into the puzzle— one’s own philosophical, logical, or linguistic programs, as many commentators have done. I could do that myself, but instead I will
62
focus on the exegetical and philosophical problem of understanding Geach’s program. In the end this will show what the puzzle is about— namely, descriptions of representations allegedly about fictional (and similar sorts of) objects— and it will greatly narrow our options for solving it. 2 Puzzle pieces G illustrates much more than intentional identity, and thus more than Geach’s program concerns. Thus I will start by setting aside two irrelevant features. First, G seems to contain a problematic kind of anaphora.1 It seems we can read the sentence as a conjunction, and on such a reading, the occurrence of ‘a witch’ does not c-command the occurrence of ‘she’. This means that in G’s syntactic tree, the first branching node dominating ‘a witch’— perhaps the node for ‘Hob thinks’ or for ‘and’— does not dominate ‘she’ (Tanya Reinhart, 1976, 31–32). In logical terms, the scope of the occurrence of the quantifier phrase ‘a witch’ is narrower than the scope of the conjunction, and so does not include the occurrence of ‘she’ (Geach, 1967, 147, 150). Thus on this reading the occurrence of ‘she’ does not function as a variable bound by the occurrence of ‘a witch’. But it does not merely stand in for a repetition of its antecedent, as some anaphoric pronoun-occurrences do. No, it acts more like a bound occurrence, even though it seems to have escaped the scope of its antecedent 1. The occurrence of ‘she’ in G is anaphoric on the occurrence of ‘a witch’ (its antecedent), as opposed to being a demonstrative (deictic) pronoun. (Since Geach calls ‘the same witch’ a “gloss” on the occurrence of ‘she’ (1967, 150), I assume the puzzle is to analyze G without the parenthesis, merely reading the occurrence of ‘she’ as anaphoric on the occurrence of ‘a witch’.) Some commentators (e.g., Nicholas Asher, 1987, §6.3) extend the term ‘anaphora’ to include historical chains of referrings linked by speakers’ intentions to corefer with earlier links. Here I use it strictly in the sense of a syntactic (though often unexpressed) connection between occurrences of linguistic expressions.
63
expression’s occurrence. Call it a fugitive. Fugitive anaphoric pronoun-occurrences have been studied extensively, but there is still no consensus on their semantics. For example, Geach attempts to analyze them as pronouns of laziness, standing in for “a repetitious phrase somehow reconstructable out of the antecedent” (1975, 194).2 And Gareth Evans argues fugitives are E-type pronouns: among other things, “they are singular terms whose reference is fixed by description” (1977, 492).3 And there are many other theories, some of which I will mention in §5. But note well: the problem of fugitive anaphora is independent of intentional identity, contrary to some commentators (who perhaps confuse the problem of intentional identity with Geach’s puzzle about G).4 Most of Geach’s examples of intentional identity reports are anaphora-free; for example, ‘Smith and Brown admire the same poet’ (1964a, 137–38) and ‘Hob and Nob meant to refer to the same person’ (1967, 150).5 And fugitive anaphora occur without intentional identity: for example, there are fugitive occurrences of ‘he’ and ‘it’ in ‘If a man owns a donkey, he beats it’. Since anaphora is not the subject of Geach’s program, I will mostly focus on nonanaphoric cases. 2. He originally used the term ‘pronoun of laziness’ only for pronoun-occurrences substituting for repetitions of their antecedent (1980, 151–52). On the extension, see his 1964b (97–98), 1967 (150), and 1975 (194–95). 3. See also Evans (1980). It is hard to separate Evans’s definition of E-type anaphora from his theses about it. And it is hard to understand either of them, since they depend on misreadings of Saul Kripke (1980, specifically his phrase ‘reference fixed by a description’) and Geach (see the latter’s 1978, and 1986, 534–35). Furthermore, though the term ‘E-type’ is common in the literature on fugitive anaphora, its use rarely follows Evans’s. I happily avoid it. 4. The commentators include Esa Saarinen (1978, 157), Walter Edelberg (1986), Asher (1987, 126), Nino B. Cocchiarella (1989, 30), Jeffrey King (1993, 63–65), Ahti Pietarinen (2001, 148), and Robert van Rooy (2006, 128). 5. By my count, only four of his thirteen examples (see the end of §3) have anaphoric pronouns.
64
G’s second irrelevant feature is the problematic general term ‘witch’. Like the terms ‘Zeus’, ‘unicorn’, and ‘phlogiston’, ‘witch’ originated in myth (i.e., widespread false belief). As with terms originating in fiction, such as ‘Spock’, ‘hobbit’, and ‘flubber’, mythical terms may raise semantic problems. For example, many theorists think fictional and mythical proper names, at least on their original use, are not really names of anything, and that this affects the semantics of sentences containing them. (I discussed the fiction case in chapter two.) As Geach puts it, “predications with ‘Fafner’ as subject are only pseudo predications, and are neither true nor false, being based on a presupposition that is not fulfilled,” namely that someone “is or was able to identify and reidentify an object under” that name (1980, 203–4).6 Like many theorists, Geach considers mythical general terms such as ‘dragon’ unproblematic: they simply have empty extensions (ibid.). But if Saul Kripke is right that many general terms (such as natural kind terms) are effectively proper names of kinds (1980, 127), then mythical general terms may be as defective as mythical proper names. He considers the term ‘unicorn’, which he assumes was introduced as a species name under the mistaken belief that someone had identified a species, specifically one superficially like a horse with one horn on its head. Since no species was named, ‘unicorn’ was not really given a semantic value. Thus “statements about unicorns, like statements about Sherlock Holmes, . . . don’t really express propositions” (2011, 67–71; cf. 1980, 156–58). We need not accept that conclusion, but this at least shows the semantics of such terms is 6. See also his 1980 (15, 186–88, 217), 1969 (156–65), and 1976 (313).
65
problematic. That includes ‘witch’, which may have been introduced to name a kind of person, under the mistaken belief that someone had identified such a kind, perhaps: person who performs a certain kind of supernatural deed.7 Since there are no supernatural deeds, no kind of deed or agent was identified and named; therefore such an introduction of ‘witch’ would have been defective.8 But note well: the problem of mythical (and fictional, etc.) general terms is independent of the problem of intentional identity (and of fugitive anaphora). Such terms occur in many other contexts, and are absent from many intentional identity reports, such as the two I gave above. With these distractions set aside (for now), we can discuss Geach’s program. 3 Geach’s program Geach’s program, discussed mostly in four papers (1964a, 1967, 1969, and 1976), concerns the semantics, and especially the logic, of certain “intentional” constructions in natural language. What is an intentional construction? In logic and semantics, ‘intentional’— or more often, ‘intensional’9 — is a technical term, and one that dif7. The current use of ‘witch’ is hard to define precisely, and its etymology is obscure, so I am not sure it is problematic in exactly the way ‘phlogiston’ is. 8. To many theorists, these problematic terms look like a job for Fregean semantics. Gottlob Frege says the general term ‘mōly’ (used for a magic herb in the Odyssey) has a sense but no referent (1892b, 133). So in the context of an attitude report, it does have a referent (namely its customary sense); thus a sentence containing it (in such a context) can express a proposition and determine a truth-value. ‘Witch’ might be in such a context in G (its occurring after ‘Hob thinks’ is defeasible evidence). But Frege’s reference-shifting semantics for such contexts, and especially his application of that semantics to nonreferring terms, are controversial (in fact I think they are false), so I will not assume them. See §3 for Geach’s views on them. 9. Geach does not use the term/spelling ‘intensional’, whose derivation he finds dubious (1980, 181). Some theorists use the two spellings to distinguish two notions, often using ‘intensional’ for one of the notions I discuss in the next paragraph.
66
ferent theorists use, define, and explain quite differently. (Though disturbingly few of them recognize the equivocation.) Geach is not sure how to define it either, but he characterizes it in two ways (e.g., 1969, 158). First, he gives examples of intentional operators. Propositional attitude verbs such as ‘believe’ are intentional, and in ‘Eve believes Jeeves leaves’ the expression ‘Jeeves leaves’ is in an intentional context (i.e., in the scope of an intentional operator). Some of the operators take sentential (or infinitival) arguments; call these intentional sentential operators. Others take as arguments singular terms and determiner phrases: for example, ‘look for’, ‘dream of’, and some noncontiguous phrases with a sentential operator and a predicate, such as ‘believe — to be a scoundrel’; call these intentional predicates. Second, he says intentional operators are nonextensional: where Ψ is an intentional operator and α an appropriate argument, the extension of ⌜Ψα⌝ is not generally a function of α’s extension. For example, if I believe the first woman in space was American, I do not therefore believe that Valentina Tereshkova was American, even though ‘the first woman in space’ and ‘Valentina Tereshkova’ are coextensive. (Although Geach once says nonextensionality is sufficient for intentionality (ibid.), mostly he seems to count it as only necessary.)10 The literature on intentionality often goes further than Geach, so I should say how he does not characterize it. First, many philosophers, following Brentano, use ‘intentional’ as a synonym for ‘representational’. It is not clear whether Geach accepts this. 10. See his comparisons with modality at 1963 (140), 1964a (138), 1972 (94, 96), and 1967 (149–50).
67
Two of his examples of intentional operators (‘is obliged’ and ‘owes’) may not be representational: maybe I can be obliged to do something without anyone or anything representing this. (See n 11.) Second, as I suggested above, he does not use ‘intentional’ as a synonym for ‘nonextensional’. Third, though he sometimes (1967, 147) uses W.V.O. Quine’s term ‘opaque’— which Quine uses partly to indicate that quantifying into a context is impossible (1966, 145–50, 152–59)— Geach thinks we can quantify into intentional contexts. Fourth, he never says intentional operators are Carnap-intensional, meaning for an operator Ψ that the extension of ⌜Ψα⌝ is a function of α’s Carnapian intension, that is, the function from possible worlds to α’s extension with respect to those worlds (or some similar modal function). (Contrast Jaakko Hintikka (1969), whom some of Geach’s commentators follow; see §5.)11 Finally, he often says or implies that intentional operators are Frege-intensional, meaning for an operator Ψ that the extension of ⌜Ψα⌝ is a function of α’s sense, that is, its mode of presentation (MOP) of its extension.12 But he once says some are Shakespearean: where β is a name, Ψ is Shakespearean if and only if the extension of ⌜Ψ β⌝ is a func11. Though Geach does not note it, two of his examples are Carnap-intensional: ‘is obliged’ (1980, 94, 189) and ‘owe’ (1964a, 130, passim; 1967, 148–49). And although he uses Buridan’s example ‘I owe you a horse’, this lacks the problematic reading I discuss below, on which (a) there is no particular horse I owe you, but (b) it is still sensible to purport to identify a horse that I owe you.(The same goes for ‘Tom has obliged himself to marry a sister of Bill’s’.) I suggest these deontic operators are modal instead of intentional, as Geach himself says elsewhere (1958, 1). (He might assume ‘owe’ is equivalent to ‘promise to give’ (1964a, 134, 136), which does have the problematic reading, and is intentional.) 12. See Geach’s 1964a (130–32, 137), 1967 (149), and 1976 (313–18). Frege calls such contexts ungerade, meaning indirect, as in indirect speech (1892a, 37). It is uncontroversial that indirect speech contexts are intentional; what is controversial is Frege’s analysis of indirect speech, in terms of reference-shifting operators. That is why I call context that fit Frege’s analysis Frege-intensional, rather than ungerade.
68
tion of β’s extension, regardless of its sense (1980, 188–89; cf. 1976, 319). So I will not assume intentionality entails Frege-intensionality. So which intentional constructions is Geach studying? He does not precisely define the class; that is part of what he is trying to work out. But he has a few families of examples. The main family of them includes ‘Peter is looking for a detective story’, ‘Peter dreams of a girl’, and ‘Peter believes a politician is a scoundrel’. In general, they have the following form, where α is a noun phrase, Ψ is an intentional predicate, and φ is a general term (common noun phrase): 0φ α Ψ a φ According to Geach, the problem with these constructions is that they have a certain ambiguity that is very hard to analyze. One reading is this: 0∃φ There is a φ such that α Ψ it. He says the problematic reading of 0φ does not entail 0∃φ . So for example, he is interested in a certain reading of ‘Peter dreams of a girl’ that does not entail ‘There is a girl such that Peter dreams of her’. For lack of familiar but neutral terms for these readings, call a reading like 0∃φ of a sentence of the form 0φ the attested reading. Call the alleged problematic reading that Geach is looking for the unattested reading. He suggests a two ways to distinguish the two readings in English. One is a wordorder convention: ‘There is a girl Peter dreams of’ (or ‘There is a girl such that Peter dreams of her’) places the indefinite description ‘a girl’ before the intentional operator ‘to dream of’, and so conventionally expresses the attested reading; ‘Peter dreams of
69
a girl’ places it after, and so conventionally expresses the unattested reading (1964a, 131; 1969, 159–61, 163–64). But the second has the form of 0φ , which Geach says is ambiguous; this does not help. The other proposed way to distinguish them is to state the attested reading as ‘Peter dreams of a particular girl’ (or instead of ‘particular’, something such as ‘definite’, ‘specifiable’, or ‘identifiable’), and the unattested as ‘Peter just dreams of a girl’.13 But ‘just’ merely excludes the salient alternative, which in some (but not all) contexts is the attested reading (cf. 1969, 159); it does not add anything to help us analyze the unattested reading. He seems to recognize this, calling this last disambiguation “groping”— “No clue is to be found here to the real logical difference” (1967, 149), and “Heaven only knows why such an explanation works” (1969, 158). Are the attested and unattested readings inconsistent? Geach thinks not; in fact he thinks the unattested reading is existence-neutral.14 For example, the unattested reading of ‘Peter dreams of a girl’ could be true if there is a girl Peter dreams of, but it could also be true if there is no girl he dreams of. He does not mean it would be true because there is a dream-girl Peter dreams of (or an imaginary girl); he does not believe in such things. Instead, he says the unattested reading is compatible with there being nothing Peter dreams of, girl or not.15 More generally, the unattested reading 13. See his 1964a (130, 133, 134, 136–7), 1967 (149–51), and 1969 (157–59). 14. See his 1964a (137–38), 1969 (160), 1976 (317), and compare his 1967 (147). 15. See his 1964a (130), 1967 (149), 1969 (156–60, and cf. 162). Besides ‘dream’ and ‘imaginary’ he also mentions ‘possible’, ‘indefinite’, ‘false’ (as in ‘false god’; ‘mythical’ is clearer), ‘nonexistent’, and ‘fictional’. ‘Indefinite’ is opposed to ‘definite’, which in the relevant use is likely a scope-indicator (see below). It is much less clear the other terms (or their supposed opposites ‘real’ and ‘actual’) are scope indicators on their relevant uses. They seem to be adjectives of a sort including ‘alleged’ and ‘fake’,
70
of 0φ is compatible with the truth or falsity of 0∃φ , as well as the truth or falsity of this weakened existential: 0∃ There is something such α Ψ it.16 (It is not clear why he thinks the unattested reading is existence-neutral. I will discuss this in §6.) Is the distinction between the attested and unattested readings a matter of the scope of an intentional operator relative to an indefinite description? For example, ‘Peter believes a politician is a scoundrel’ might be read with ‘a politician’ taking wider scope than ‘believes’: a politician is such that Peter believes he/she is a scoundrel (i.e., there is a politician whom Peter believes to be a scoundrel). Or it may take narrower scope: Peter believes that there is a politician who is a scoundrel. In parts of his 1964a Geach does seem to treat the distinction as a matter of scope. But his proposed analyses do not reduce it to scope (see §5); later he contrasts the attested/unattested distinction with well known scope distinctions (1967, 149–50); and eventually he clearly distinguishes unattested and narrow-scope readings (1976, 315–17). He has at least two reasons to distinguish them. First, many of Geach’s cases lack a plausible paraphrase in terms of an existential quantifier in the scope of a sentential operator; for example, ‘admire’, ‘dream of’, ‘recognize’ (all in his 1964a), and ‘worship’ have no correwhere for such an adjective χ, ⌜α is a χ φ⌝ does not entail ⌜α is a φ⌝. Geach may think that realism about fictional (or mythical, etc.) objects depends mostly on misreading scope indicators as ordinary predicates (see the above references). It does not, as subsequent realists such as Kripke (1973a) and Peter van Inwagen (1977) show. 16. See his 1964a (132–33, 135–36), 1967 (147, 151), 1969 (160, 165), and compare his 1976 (313– 14).
71
sponding sentential operators (David Kaplan, 1986, 266–68). (Compare ‘want’: on one reading, ‘I want a sloop’ does have a plausible paraphrase: ‘I want that there be a sloop that I have’; cf. Quine 1966, 184; and Geach 1980, 133.) Of course this might seem like a false negative; maybe we simply have not looked hard enough. But the second reason is more conclusive: the narrow-scope and unattested readings relate differently to identification. If the sentence ‘Peter believes a politician is a scoundrel’ is true only on its narrow-scope reading, then it is not sensible to ask “which politician?” or to identify one — e.g., “namely, the president”— because there is no particular politician Peter believes is a scoundrel (Geach, 1980, 91, 120). But Geach thinks that if ‘Hannibal worships a god’ is true only on its unattested reading— so that there is no god Hannibal worships— then it is sensible to ask (or to purport to tell) which god (1969, 161).17 And it is senseless to say ‘Hannibal worships a god, but no god in particular’.18 More generally, he says “we very often take ourselves to know, when we hear the discourse of others, that they are meaning to refer to some one person or thing— and that, without ourselves being able to identify this person or thing, without our even being certain that there really is such a person or thing to identify” (1967, 17. In his 1964a (133–34), Geach likely conflates the senselessness of identifying an object with there (in some sense) being no object. (Several of his commentators do the same.) But elsewhere he admits that in unattested cases we can still purport to identify an object of the intentional state (1980, 186– 88; 1969, 161; 1976, 314–17). Compare G.E.M. Anscombe’s distinction between giving an intentional state’s “intentional object” and giving its “material object” (1965, 9–10, and cf. 5): the latter is possible only for the attested reading, while the former is possible for the attested or unattested readings— but of course not for the narrow-scope reading. (Geach alludes to her discussion: 1967, 147.) 18. Compare Kripke’s infelicitous case ‘The Greeks worshipped a god, any old god’ (1973a, 3.12– 14). (On ‘any old’, see below, n 19.) He notes that such cases have a reading as ⌜For all x, if x is a φ, then α Ψ x⌝ (cf. Geach, 1980, 91). But this is not the unattested reading; I call it the ‘I love a parade’ reading.
72
151). In other words, “one purports to identify” the object of the intentional state (1969, 161). But when the narrow-scope reading is the only true reading, it is senseless to even purport to identify an object of the intentional state. So that reading is not the alleged unattested one; call it the undirected reading.19 I will say that both the attested and unattested readings are directed, meaning they are not undirected: we can at least purport to identify objects of such intentional states. Besides the main family of examples, which have the form 0φ , Geach discusses two other families. One replaces the indefinite description with a proper name; for example, ‘Smith believes this ancient hill-fort was built by Arthur’, rather than ‘Smith believes this ancient hill-fort was built by a man’ (1969, 162–63). In general, they have this form, where β is a proper name: 0β α Ψ β Like 0φ , Geach says an instance of 0β may be true even if its corresponding instance of 0∃φ is false, and even if the corresponding instance of 0∃ is false. For example, Smith can believe this ancient hill-fort was built by Arthur, even if there was no such person as Arthur, and even if there is nothing which Smith believes built this an19. Nathan Salmon uses the term ‘undirected’ for a kind of use of definite or indefinite descriptions (2002, 106); here I use it for a certain semantics of intentional predicates with descriptions as arguments, and by extension for the kind of intentional state thus represented. Like some other theorists (e.g., Mark Richard, 1998, 269), I used to call this reading “nonspecific” (or oppose it to a “specific” reading); but that is misleading. An undirected intentional state may be more or less specific. For example, suppose I want a sloop, although there is no particular sloop I want. This may be compatible with further specifying what I want: I do not want just any old sloop; specifically, I want a fast one. (Compare Anscombe’s claim that the possible “indeterminacy” of a state’s object shows the state is intentional (1965, 4). Her examples include cases of undirected states (4, 6, 12) as well as a directed — in fact attested — nonspecific state (12, no 9). But whether indeterminacy is undirectedness or nonspecificity, it is not a mark of the problematic kind of intentionality Geach is most concerned with.)
73
cient hill-fort. On the other hand, he does not say that an instance of 0β entails that the corresponding instances of 0∃φ and 0∃ are false; it is existence-neutral. And like the unattested reading of ‘Hannibal worships a god’, the problematic reading of ‘Hannibal worships Baal’ is not merely a narrow-scope (undirected) reading. For one thing, names have no scope (Geach, 1963, 139, 144); for another, the sentence purports to identify an object of the intentional state, which shows the state is directed, even if it is not attested. Can we not conclude ‘Baal is worshipped by Hannibal’, and then ‘Something is worshipped by Hannibal’?20 Geach says no: we nonbelievers use the name ‘Baal’ as a quasiname. We use quasinames in intentional contexts without committing to there being anything they name in those contexts, or to their even being defined outside intentional contexts (1969, 161–63). In effect, we use them to purport to identify objects of intentional states, whether or not we think there really are objects of those states. The last family of cases is where we find intentional identity. So far I have given Geach’s simple examples, which each report one intentional state. But there are compound reports, combining simple reports about different agents, different state-types, or merely different instances of the same state-type. Consider a case like ‘Smith and Brown admire the same poet’. Geach says this is ambiguous in the same way as ‘Pe20. Geach thinks proper names express “nominal essences,” categories of things that they are “names for,” whether or not they really refer to anything (1980, 67–71, 203–4). For example, ‘Baal’ expresses the nominal essence god (it is a god-name) and ‘Arthur’ expresses person (it is a person-name). He might then say that ‘Baal is worshipped by Hannibal’ entails ‘A god is worshipped by Hannibal’, and in general that if β expresses the nominal essence denoted by φ, that ⌜β is such that αΨ it⌝ entails not only 0∃ , but 0∃φ . But this view of names is independent of his program. He can still state 0β ’s existence-neutrality in terms of 0∃ rather than 0∃φ . (I think names do not express nominal essences; see my discussion of Amie Thomasson’s semantics for names in chapter two.).
74
ter dreams of a girl’. One of its readings is ‘There is a poet that Smith and Brown admire’; on this reading Geach says the sentence expresses “real” or “actual identity,” i.e., the real identity of the objects of Smith’s and Brown’s admiration. Of course, it does not literally do so.21 Geach might better say that it entails ‘There is a poet that Smith admires, and there is a poet that Brown admires, and the first poet is the second’, which does include an identity statement. Putting it that way also shows the connection between the real identity reading of a compound report and the attested reading of a simple report: a compound report on its real identity reading is equivalent to a conjunction of simple reports on their attested readings, plus another conjunct stating that certain objects of all the relevant intentional states are identical. On the other hand, Geach says ‘Smith and Brown admire the same poet’ has a reading on which it expresses “intentional identity” (1964a, 137–38). Though he never defines that term, in the context of its program its meaning is clear enough: it is an analogue for compound reports of the unattested reading of simple reports, in the same way that the real identity reading corresponds to the attested reading. So suppose Smith and Brown both fall for the “Ern Malley” literary hoax,22 in which case Geach would say that ‘Smith admires a poet’ and ‘Brown admires a poet’ are both true only on their unattested readings. So we can at most purport to identify objects of their admiration, namely Ern Malley. And in fact we can purport to say that the object 21. Thanks to Salmon for this point. He also noted to me that the sentence does not identify (or even purport to identify) any poet whom Smith and Brown admire. 22. Two Australian writers fooled publishers and critics into thinking there was a poet named ‘Ern Malley’. Geach alludes to this (1964a, 137; 1969, 164–64).
75
of Smith’s admiration is the object of Brown’s admiration. That is what we purport to assert when we assert the compound report on its intentional identity reading— that, plus what is expressed by the two simple reports on their unattested readings.23 In general, the compound reports that have real identity and intentional identity readings often have something like this form, where proφ is a pronominal phrase such as ⌜ the same φ⌝ or a pronoun of appropriate gender: I1 α 1 Ψ1 a φ, and α 2 Ψ2 proφ , and . . . α n Ψn proφ . G has this form, where n = 2: α 1 is ‘Hob’, α 2 is ‘Nob’, Ψ1 is ‘thinks — has blighted Bob’s mare’, Ψ2 is ‘wonders whether — killed Cob’s sow’, φ is ‘witch’, and proφ is ‘she’ (glossed as ‘the same witch’). Where the intentional predicate is the same, the reports can be put in this form: I2 α 1 , . . . and α n Ψ proφ . ‘Smith and Brown admire the same poet’ has this form.24 Where the agent is the same, the reports can be put in this form: I3 α Ψ1 a φ, Ψ2 proφ , and . . . Ψn proφ . Geach gives a couple of examples with a single agent, though he needlessly repeats the subject term: e.g., ‘I saw a man on the stair yesterday at time t1 , and I saw him 23. See his 1964a (137–38), 1967 (151), 1969 (161, 164–65), and 1976 (318). 24. So do these examples: ‘Hob and Nob meant to refer to the same person’ (1967, 150), and ‘Hob or Nob may have more than one definite description “of that person” that he might produce’ (ibid., 152 n 1). (Here and in a few other quotes, I change Geach’s single quotation-marks to doubles, and vice versa, to fit the usual American philosophical use.) Note well: though these reports have only one occurrence of an intentional predicate, they report multiple intentional states. The predication is not collective (as in ‘we elected the president’), but distributive (Geach, 1980, 97) — in this case, that means the report is equivalent to one of form I1. (This will be relevant to §5.2.)
76
(the same man) on the stair again today at time t2 ’ (1967, 146–47).25 He also shows more variations, by quantifying over agents or intentional states (i.e., state tokens): ‘A man . . . dreams of the same girl night after night’ (1969, 161), and ‘Many critics admired one and the same poet’ (ibid., 164–65).26 Finally, he sometimes substitutes a proper name for the indefinite description or for proφ , or adds it as a “namely-rider”: ‘He dreams of Petronella every night’ and ‘Hannibal and Hasdrubal did worship the same God, namely Baal’ (1969, 161).27 So that is Geach’s program: to analyze these families of intentional constructions, specifically on certain problematic readings that purport to be about certain things, whether or not there really are such things. This sets the context for his puzzle. 4 The puzzle, as Geach understands it I now present Geach’s puzzle, in detail, as he understands it. To repeat, it concerns this compound report: G Hob thinks a witch has blighted Bob’s mare, and Nob wonders whether she (the same witch) killed Cob’s sow. In his paper (1967) he describes a case, and some possible variations, to clarify G’s intentional identity reading. To help exposition I will treat him as stipulating constraints 25. The other example is ‘Nob assumes that just one witch blighted Bob’s mare, and Nob wonders whether she (that same witch) killed Cob’s sow’ (1967, 151–52). These intrasubjective cases belie some of Geach’s commentators, who assume intentional identity can only hold between distinct agents, or that the intrasubjective case is unproblematic (e.g., Asher 1987, 152–53; and Pietarinen 2001, 161– 62). See Edelberg (2006) on this point. 26. A more complex example, with verb-phrase anaphora, is ‘Hob and Bob might “refer to her” with the same description in mind, and likewise Bob and Nob, but not Hob and Nob’ (1967, 152 n 1). 27. His other two are ‘Wewena, Chuckery, and Cousheda all worship the same God, Mumbo Jumbo’ (1969, 162) and ‘Smith believes a certain man to have reigned over Britain, and Jones believes the very same man to have built this hill-fort’ (1969, 163).
77
on G’s interpretation. Some of them are not explicit or clear in his paper, but as I will explain, he commits to them anyway. (In this section, references are to his 1967, except where noted.) He begins with these stipulations: S1 G states intentional identity. S2 Villagers of Gotham, including Hob and Nob, think there is a witch around, and someone reports on their attitudes by asserting G. (147–48) S3 The reporter could consistently assert G without committing to there being witches. (148) S2 and S3 mention the reporter’s attitudes and assertions, adding a superfluous layer of representation.28 To simplify, for S2 and S3 (and some implicit principles about assertion and belief), substitute these: S2′ G is true. S3′ There is no witch such that Hob thinks she has blighted Bob’s mare, and Nob wonders whether she killed Cob’s sow. S3′ simply denies G’s real identity reading. It must then be true only on its intentional identity reading; the point of the other stipulations is to clarify what that is. If G’s real identity reading is false, then Geach would add that so is the weaker statement that something is a common object of both Hob’s and Nob’s attitudes (see above, n 16). (He also thinks that G’s intentional identity reading is existence-neutral (147). But 28. Some proposed solutions rely on such an extra layer of representation; see §5.
78
nothing in the example depends on this, so I will not bother to make a stipulation of it.) And he would add that Hob’s and Nob’s attitudes are directed (151): we can (purport to) identify their shared object, for example by an ad hoc proper name such as ‘Maggoty Meg’ (1976, 314–15). He also assumes the occurrence of ‘she’ in G is anaphoric on the occurrence of ‘a witch’— that it is not a demonstrative (deictic) pronoun. So add these: S4 There is nothing such that Hob thinks it has blighted Bob’s mare, and Nob wonders whether it killed Cob’s sow. S5 Hob’s thought and Nob’s wondering are directed.29 S6 In G, the occurrence of ‘she’ is anaphoric on the occurrence of ‘a witch’. Within these constraints, he tries to analyze G’s intentional identity reading by giving several paraphrases. Since he rejects them all, I will add stipulations to rule them out. His first paraphrase is the attested, real identity reading (148)— call it G1. S3′ is G1’s negation, so that paraphrase fails. He then considers a variation on G1: G2 As regards somebody, Hob thinks that she is a witch and has blighted Bob’s mare, and Nob wonders whether she killed Cob’s sow. (ibid.) 29. Several commentators read G as attributing only an undirected attitude to Hob, Nob, or both; see §5. The undirected reading of G’s first conjunct — taken alone — is a natural one. But in the context of Geach’s program, it is clear this is not the relevant reading. (Of course, Hob or Nob might have attitudes of both types; a few commentators analyze G that way: see n 37.) In one of the most cited responses to Geach’s paper, Edelberg (1986) argues that besides handling cases such as G, an analysis of intentional identity must also handle what he calls “asymmetric intentional identity” (12–17). But his cases of the latter are not intentional identity at all: he stipulates that the agents’ attitudes are undirected (“neither [agent] has anyone in mind”). His “new puzzle of intentional identity” may be interesting, but he has not shown that one must solve it to solve Geach’s puzzle. Compare van Rooy’s criticism (2006, 142–43).
79
It is odd that he even considers G2, because G does not even entail it: unlike G, G2 represents personhood (with the quantifier, ‘somebody’; i.e., ‘some person’). Although he seems to assume witches are persons (ibid.), G itself does not say this. Furthermore, he thinks ⌜α and β are the same witch⌝ does not entail ⌜α and β are the same person⌝, even assuming witches are persons (1980, 174–76). Hob and Nob might have attitudes about distinct holders of the “office” of Gotham Witch— the same witch, but distinct persons. (I disagree. But this is a reason for Geach to ignore G2.) It is also odd how he rejects G2. He says it “would imply that Hob and Nob had some one person in mind as a suspected witch; whereas it might be the case, to the knowledge of our reporter, that Hob and Nob merely thought there was a witch around and their suspicions had not yet settled on a particular person” (ibid.). But the two claims he denies, ‘Hob and Nob had some one person in mind’ and ‘Hob and Nob’s suspicions had settled on a particular person’ can be read as stating intentional identity (in form I2) rather than real identity, just as well as G can. And on their intentional identity reading it is no problem if G entails them. But if he means to say that G2 is a real identity reading, then there is no point in considering it separately from G1. He also says that instead of G2, what might be true is ‘Hob and Nob merely thought there was a witch around’. But that differs from G2 mainly in that its most natural reading is undirected, which is not what we want from a paraphrase of G. Finally, assuming there was a woman attacking livestock, Hob and Nob could have had attitudes about her without having “one person in mind as a suspected witch,”
80
i.e., without “their suspicions [having] settled on a particular person,” if those claims would entail that they knew who she was (cf. J.G. Barense, 1969). In general, an agent can have an attested intentional state about something without knowing what thing it is (e.g., knowing who it is).30 And I see no reason this would not extend to unattested states: even if there is no witch Hob and Nob had in mind, because there is no witch at all, we can still purport to identify the object of their attitudes as Maggoty Meg. Then in some sense they have Meg in mind (this is another intentional identity report), whether or not they know who (or what) Meg is— that is, whether or not they have a suspect. In any case, Geach need not get into these issues, because he has a much better reason to reject G2: S4 alone rules it out. He next considers paraphrasing the occurrence of ‘she’ as a pronoun of laziness, specifically as standing in for ‘the witch who blighted Bob’s mare’: G3 Hob thinks a witch has blighted Bob’s mare, and Nob wonders whether the witch who blighted Bob’s mare killed Cob’s sow. (150) He rejects it, “for our reporter might be justified in asserting” G as an intentional identity report “if he had heard Hob say ‘The witch has blighted Bob’s mare’ and heard Nob say ‘Maybe the witch killed Cob’s sow’, even if Hob had not thought or said anything about Cob’s sow nor Nob about Bob’s mare” (ibid.). So add this stipulation: S7 Hob had not thought or said anything about Cob’s sow nor Nob about Bob’s mare. 30. See Salmon (1988, 205–6, 213 n 17) for this distinction and some references.
81
S7 is consistent with G being true, but rules out G3. He then raises an independent problem for paraphrases like G3, which use a definite description in Nob’s attitude report (151–52). He assumes the definite description is analyzable; taking it in isolation, he accepts the Russellian analysis of ‘the witch who blighted Bob’s mare killed Cob’s sow’ as ‘just one witch blighted Bob’s mare and she killed Cob’s sow’. But he doubts that G3’s second conjunct entails ‘Nob wonders whether (the following is the case:) just one witch blighted Bob’s mare, and she killed Cob’s sow’. He does not say why, but it is likely because that attitude report is undirected, contrary to S5. Instead, he suggests analyzing G3’s second conjunct as ‘Nob assumes that just one witch blighted Bob’s mare, and Nob wonders whether she (that same witch) killed Cob’s sow’. Why this? Again, he does not say. But it is more or less a Strawsonian presuppositional analysis, a common alternative to the Russellian.31 In any case, he rejects that analysis too: given the stipulations so far, it can only be read as an intentional identity report. Therefore, since the best analysis of G3 “would introduce intentional identity over again,” G3 “is quite useless as a way of getting rid of intentional identity” (152). Of course, this objection depends on the Strawsonian analysis being right— or more generally, it assumes that G3’s second conjunct is best analyzed as a compound attitude report with at least one attitude directed. And it assumes that the intentional identity report introduced is just as hard to analyze as G (which some commentators would deny; see n 25). That is a lot to as31. Thanks to Salmon for suggesting this reading.
82
sume. More conservatively, Geach could object that such paraphrases of G3 either make Nob’s attitude report undirected, or they violate S7 in trying to connect Hob’s and Nob’s attitudes. The Russellian analysis has both flaws; the Strawsonian avoids the first, but not the second. Finally, he considers a general paraphrase strategy, for some common noun phrase F: G4 Hob thinks that the (one and only) witch that is F has blighted Bob’s mare, and Nob wonders whether the witch that is F killed Cob’s sow. (152) He says this has the previous problem: analyzing the ‘Nob wonders’ clause seems to produce another report of intentional identity. Again, I suggest a more conservative objection: G4 makes the attitudes undirected, violating S5. Aside from that, his response to G4 is odd. He says that “on the face of it,” and waiving some unidentified difficulties,32 if G4 is true for any F then G is true. But he doubts the converse, i.e., that G entails G4’s existential closure. He repeats that G could be true even if Hob said “The witch has blighted Bob’s mare” and Nob said “Maybe the witch killed Cob’s sow,” so long as they both use ‘the witch’ meaning to refer to the same person. About this case he asks: Now is it in truth necessary, if Hob and Nob are to mean to refer to the same person as “the witch,” that they should both have some one definite description actually in mind, or even, one producible from each of them by a suitable technique of questioning? This appears to me to stand or fall with the corresponding theory, held by Russell and by Frege . . . , that any ordinary proper name is used equivocally if it does not go proxy for some one definite description; and in 32. They may be the alleged difficulties I mention in note 42.
83
spite of these great names, such a theory seems to me extremely ill-founded and implausible. (152–53) Is this a trick question? If they use the same description (‘the witch’), then of course they have the same description in mind. And what could that have to do with descriptivism about proper names? I suggest Geach does not think ‘the witch’ is the description Hob and Nob “have in mind” when they use ‘the witch’, at least not in the special sense that each (purportedly) identifies the object of his attitude as the thing satisfying that description. So how do they (purport to) identify it? Assuming they do so by description, it is a sensible question whether Hob and Nob must both have exactly one and the same description in mind. And this is analogous to the question whether two people using the same proper name (univocally) must both have exactly one and the same description in mind as fixing its referent. The answer is no: neither is necessary. Why not? He says Hob and Nob might have multiple descriptions in mind, and more importantly, that they need not share any (or all) of their descriptions. His counterexample is that Hob and Nob might each share a description “of that person” with Bob, but not with one another (152 n 1). (Note that this counterexample is equivalent to two intentional identity reports plus the denial of a third; see above, n 26.) It is not clear how deep Geach thinks this cuts. He might intend that Hob and Nob need not share any way of (purportedly) identifying the objects of their attitudes. This is how Nathan Salmon reads him (2002, 110), and in §5.3 I will argue this is what he ought to say. But he does not say it here, and his analyses of intentional identity in other papers (see §5.3)
84
suggest he would deny it. So, finally, I add this conservative stipulation to rule out G4: S8 G can be true even if there is no shared definite description by which both Hob and Nob purport to identify an object of their respective attitudes. 5 Proposed solutions I will now summarize all published proposed solutions to the puzzle, or at least all those I know of, grouped by Geach’s reasons to reject them. In some cases I agree with him; in others I think he rejects a solution too hastily. But I will save my criticisms for §6. 5.1
Solutions diverging from Geach’s stipulations
Most commentators diverge from Geach’s understanding of the puzzle, by rejecting or ignoring one of the stipulations. Some commentators doubt G has an intentional identity reading;33 Geach would say they are missing the point and violating S1. Some of them express their doubt by denying that G is true (i.e., denying S2′ ). Several others say the reading Geach intends is nonliteral, not purely a consequence of G’s semantics.34 He would deny this; he thinks G, like many other intentional identity reports, is literally true. (I am not so sure; see §6.) 33. See D.C. Dennett (1968, 337–38), Michael McKinsey (1986, 162, 174 n 10), and King (1993, 76). Compare Geach’s own (temporary) skepticism (1969, 164–65). Robin Cooper (2005, 344–47) claims to analyze intentional identity reports, but treats them simply as real identity reports. Taken as a solution to the puzzle, his analyses violate not only S1, but S3, S3′ , S5 and S7; I will ignore them below. 34. See King (1993, 79), Craige Roberts (1996, 16–17), van Rooy and Thomas Ede Zimmerman (1996, 134–35), Mark Crimmins (1998, 40), Henk Zeevat (2000, 304–5), Alan Berger (2002, §6.5), and van Rooy (2006, 140–43).
85
Some commentators take G to express not simply a relation between Hob and Nob, but someone’s attitude about them— specifically, a single attitude that commits that person to real identity.35 (This is at best a nonliteral reading, as Alan Berger admits.) Geach would respond that this violates S3, and misses the point: an intentional identity report purports to identify objects of multiple intentional states, and to entail that they are the same object; it does not purport to report anyone’s single attested intentional state about two other agents’ unattested states. Terence Parsons says there is a witch Hob’s and Nob’s attitudes are about— denying S3′ — although he says it is a “nonexistent” witch (1974, 577–78). As I noted in 3, Geach denies that problematic intentional states such as Hob’s and Nob’s are really just attested states about special objects (see n 15). He thinks there is no object that both Hob’s and Nob’s attitudes are about (S4). For that reason he would reject Parsons’s analysis plus several others that do not take the supposed object to be (really or actually) a witch.36 (I am not sure he should do so; see §6.) 35. Howard Burdick (1982, 209) does not identify the representer. Tyler Burge’s first paraphrase (1983, 96) is ambiguous: G* According to the community’s hysterical beliefs, there is a witch wreaking havoc, and as regards that witch, Hob thinks she has blighted, . . . and Nob thinks she has killed . . . [sic] On one reading it is merely a more complex (and still unanalyzed) intentional identity report, adding to G a third attitude report, of the community’s (collective?) belief. On the other, it is a report of a collective community belief in a witch (and acceptance of a real identity report concerning Hob and Nob), violating S3 in spirit. (Furthermore, on either reading, community belief probably requires individual belief, and in this case intentional identity: each villager of Gotham believes there is a witch (the same witch) such that Hob . . . and Nob . . . . So on either reading, G* paraphrases G as a more complex and unanalyzed intentional identity report. Cf. Burge’s second paraphrase, below in §5.3.) Edelberg (1992, 589–95; see also his 1995) and Berger (2002, ch 6) analyze G only from someone’s perspective. Edelberg does not identify the representer for this case, while Berger picks Nob — or rather, discusses a belief Nob would have if he were ideally rational and better informed. Berger says this is the most we can get (135), but Geach would disagree (as do I). 36. See Hector-Neri Castañeda (1974, 47–48), Cocchiarella (1989, 31–32), Salmon (2002, 116–17 and nn 26–27), and Graham Priest (2005, 65 n 12).
86
Many commentators read one or both attitude reports as undirected, contrary to S5.37 Geach would say they miss the point of his program and puzzle, namely to analyze a kind of reading other than the undirected (or attested) kind. And several of them analyze Nob’s supposed undirected attitude as about Cob’s sow, or Hob’s as about Bob’s mare, contrary to S7.38 Geach would also reject several solutions because they assume there is a definite description by which both Hob and Nob (purport to) identify objects of their respective attitudes, contrary to S8.39 5.2
Solutions that otherwise miss the puzzle’s point
Geach would reject a few other solutions for special reasons. Friederike Moltmann (2006, 249–50) analyzes a purportedly analogous case as reporting a collective attitude, rather than distinct attitudes of distinct agents (see above, n 24). Of course, this only works for intentional identity reports of a form equivalent to I2 (⌜α 1 , . . . and α n Ψ proφ⌝). G does not have such a form. (In fact it is Geach’s only example of a compound report of form I1 that is not equivalent to a report of I3 or I2.) Geach would call her approach a nonstarter. 37. See Dennett (1968, 337–38), Robert Kraut (1983, 277–79, 288–91), McKinsey (1986, 162–64, 168–69), Stephen Neale (1990, 221), King (1993, 74, 76), William G. Lycan (1994, 127), van Rooy and Zimmerman (1996, 133–34), Pietarinen (1999, 15–17; 2001, 148), Zeevat (2000, 304–5), Berger (2002, 140–41, for Hob only), Richard Breheny (2003, 8), and van Rooy (2006, 129). Some others analyze G as stating Hob has both undirected and directed thoughts (e.g., Burge 1983, 97; Cocchiarella 1989, 31–32; and Salmon 2002, 122 n 27); this is consistent with S5. 38. See McKinsey (1986, 162–64, 168–69), King (1993, 74, 76), Lycan (1994, 126–27), and van Rooy and Zimmerman (1996, 133–34). L. Jonathan Cohen (1968, 330, 332) also violates S7 in his analysis of an analogous case, by giving truth conditions for one agent’s attitude that include those of the other agent’s. 39. See Geach (1964a, 137–38), Dennett (1968, 337–38), McKinsey (1986, 162–64, 168–69), King (1993, 74, 76–79), van Rooy and Zimmerman (1996, 133–34), Zeevat (2000, 304–5), Breheny (2003, 8), and Takeo Kurafuji (2007, 435). It is not clear whether Neale accepts this; he paraphrases Nob’s attitude report using a description such as ‘the local witch’ (1990, 221), but he does not say whether Hob also (purportedly) identifies the object of his thought by that description.
87
He would reject other solutions for incompleteness. Many commentators analyze ⌜α Ψ a φ⌝ after Hintikka’s (1969) analysis of ⌜α Ψ γ⌝ (where γ is a singular term): he
replaces Ψ with a modal operator (e.g., ⌜α believes⌝ becomes ⌜in all possible worlds compatible with what α believes⌝), and assigns to γ a function from worlds to individuals (a “world-line”). (This is to treat intentional operators as Carnap-intensional, contrary to Geach and good sense.) Besides assigning the occurrence of ‘she’ such a function, they assign one to the occurrence of ‘a witch’. They say these two functions are not defined at the actual world; but for some (or all) worlds at which they are defined, they have the same value. Most of these commentators also say the functions are inconstant, having different values at different worlds, as a way to represent their undirectedness, contrary to S5.40 But even the directed analyses must explain what sort of attitude this Hintikkan formal model represents. Do Hob and Nob have attested attitudes toward a nonactual object? Then the analysis violates S4. Or do the functions stand in for certain ways of identifying objects (e.g., MOPs or concepts)? Then the analysis is metarepresentational (see §5.3). Maybe other interpretations are possible; but without some interpretation this approach simply translates the problem of analyzing intentional identity reports, and more generally unattested reports, to the (newer and less intuitive) problem of explaining attitudes involving world-lines undefined at the actual world. Barry Hartley Slater’s account (1988, 98) is incomplete in a more idiosyncratic 40. See Kraut, Lycan, van Rooy and Zimmerman, Pietarinen, Zeevat, and van Rooy (references in n 37). The exceptions are Saarinen (1978, 195, 201–3), Roberts (1996, 16–17), and Ephraim Glick (2011).
88
way. He analyzes pronouns anaphoric on indefinite descriptions using Hilbert’s ‘ε’ operator. I will define ‘ε’ for a relatively simple case (cf. Slater, 2005). For extensional predicates φ and ψ containing no free variables, ⌜εα φα⌝ is a singular term such that ⌜∃x φx⌝ ⊢ ⌜φεα φα⌝ and ⌜∀x(φx ↔ ψx)⌝ ⊢ ⌜εα φα = εα ψα⌝. When φ is satisfied, ⌜εα φα⌝ denotes a certain satisfier of φ, the one picked out by a certain choice function — for example, Hilbert’s choice function for the domain of natural numbers, which maps each predicate to the least number satisfying it (if any). (So on Hilbert’s interpretation, we can read ⌜εα φα⌝ as ⌜the least thing α such that φα⌝.) Stated in mostly English, this is Slater’s paraphrase of G: G5 Hob thinks that there is a witch who has blighted Bob’s mare, and Nob wonders whether εx(x is a witch who has blighted Bob’s mare) killed Cob’s sow. G5 has several problems. First, it ascribes to Hob only an undirected attitude, contrary to S5. Perhaps Slater should add, between the two reports, ‘and Hob thinks, of εx(x is a witch who has blighted Bob’s mare), that she is a witch who has blighted Bob’s mare’. Second, ε-terms have scope, which G5 leaves ambiguous: on one reading Nob’s attitude is toward the ε-term’s extension (disambiguated with λ: ⌜λx(Nob Ψ x)εα φα⌝), and on the other reading the attitude report is undirected. Since he acknowledges S7 (93), I assume he means the former. Third, our domain is not the natural numbers, so we cannot use Hilbert’s choice function, and Slater does not give another. However, this is a special case, where nothing satisfies the predicate, and so the choice function could not pick a satisfier anyway. Slater says that when nothing sat-
89
isfies φ, ⌜εα φα⌝ denotes the object that, by stipulation, all empty descriptions denote, perhaps the empty set (call this the garbage object).41 I assume this requires further interpretation, but as far as I know he gives none. That is the fourth and worst problem. At face value, G5 entails Nob wonders about the garbage object, which is not only very unlikely but makes the attitude attested, violating S4. If this is not Slater’s meaning, then he has at best only restated Geach’s problem of analyzing unattested and intentional identity reports (with indefinite descriptions as arguments of the intentional predicates), now as the problem of analyzing sentences such ⌜α Ψ εx φx⌝ when φ is unsatisfied. The restatement does not solve Geach’s puzzle, and I doubt Geach would see any benefit in it, because I see none myself. 5.3
Metarepresentational solutions
All the remaining solutions are metarepresentational. They read G as ascribing to Hob and Nob not only full propositional attitudes, but ways of (purportedly) identifying the objects of their attitudes. With a few exceptions, they are consistent with Geach’s stipulations, deal with the right sort of case, and are not obviously incomplete. Still, I will argue he has reasons to reject all metarepresentational solutions, including his own. The first family of metarepresentational solutions say Hob and Nob share a way of (purportedly) identifying the objects of their attitudes. Geach gives such an analysis of ‘Smith and Brown both admire the same poet’: 41. There are of course other options for treating nondenoting descriptions: see, e.g., Rudolf Carnap (1947, §8). As Slater notes (2005, §1), Hilbert and Bernays effectively treat them as unevaluable due to a failed presupposition.
90
M1 For some w, w is a definite description, and Smith and Brown both admire-as-apoet-someone-conceived-under-the-ratio-evoked-by w. (1964a, 138) (‘Ratio’ is Buridan’s term, roughly equivalent to Gottlob Frege’s ‘Sinn’. ‘Lookingfor- . . . -evoked-by’ is an unanalyzable relative term (137). And cf. Alonzo Church 1956, 8 n 20.) To extend this to G, we might use two unanalyzable relative terms, maybe ‘think to be a witch who has blighted Bob’s mare someone conceived under the ratio evoked by’ and ‘wonder whether she (is a witch who) killed Cob’s sow someone . . . ’. Aside from being very awkward, this wrongly packs logically significant structure into unanalyzable predicates; the analysis would not entail, as it should, that Hob thinks Bob’s mare has been blighted, among other things. But even if we dealt with that problem, there would be another: the analysis would violate S8. And no doubt this is exactly the sort of analysis Geach was thinking of in stipulating S8.42 In his 1976, he no longer shyly quantifies over definite descriptions as senseproxies, but quantifies over senses themselves (1964a, 137; 1976, 313, 315). Mostly he quantifies over senses of proper names, which he calls aspects (1976, 313). (He 42. This one, or another one he maybe meant to give. M1 does not follow the pattern he set with his analyses of simple unattested reports (137). Following the pattern, his analysis should be: M* Smith and Brown both admire-someone-conceived-under-the-ratio-evoked-by ‘poet’. He describes the second arguments of these unanalyzable relative terms as definite descriptions (despite writing them as common noun phrases such as ‘poet’ and ‘detective story’), so a M*-like analysis of G would likely also violate S8. Though S8 may be his reason to reject such analyses in his 1967, he gives a different reason in his 1964a, 138. His attested analysis requires that the intentional state’s object really satisfy the description occurring after the intentional predicate, but not that it be represented as satisfying it. Geach says this is wrong (I am not sure it is), and then seems to reject all his analyses on that basis, even those of unattested, real identity, and intentional identity reports, none of which have that feature (cf. 1969, 164). This seems to be a mistake. (Also, it is not obvious that the compound report (purports to) say Smith and Brown admire anyone as a poet, instead of that the person they admire is a poet; cf. my Shel Silverstein case, below.)
91
understands them as “uses,” including possible uses that merely could be given to a name.) As in his 1967 (152–53), he rejects the descriptivist view that each aspect can be expressed by a single definite description, the description used to (purportedly) identify the referent (1976, 315). To report an attitude while quantifying over aspects, he uses brackets by analogy with Quine’s quasiquotation: within the brackets, Greek-letter variables range over aspects, while other expressions’ extensions are their senses (316; since I have already used Greek, I will make these German).43 So he analyzes the unattested reading of ‘Jones believes a detective can F’ as: M2 For some a, a is a detective-aspect and Jones believes [a can F]. (317) The most likely analogue for G is this: G6 For some a, a is a witch-aspect, and Hob thinks [a has blighted Bob’s mare], and Nob wonders whether [a killed Cob’s sow].44 Some commentators give similar analyses. Michael Pendlebury’s analysis (1982, 352–54) is a notational variant of G6 (which, ironically, he rejects for its “extravagant and unnecessary ontological commitments;” 348–49). (Cf. Mark Crimmins (1998, 40); although he says G is only pretend-true, his conditions for its pretend-truth are another notational variant of G6.) Likewise, Matts Dallhöf (1995, 146) requires that Hob’s and Nob’s attitudes involve the same “individual concept,” by which he means a way of (purportedly) identifying an individual (38). And according to Takeo Kura43. Compare Kaplan’s capital-M meaning marks (1969, 214). Thanks to Salmon for this reference. 44. Salmon (2002, 110) reads Geach as suggesting an analysis ascribing distinct but related aspects, at least partly on the grounds of Geach’s support of S8. As I explained, my reading of that passage is more conservative, and so I am not sure Salmon is right. In any case, I will soon discuss analyses like the one he ascribes to Geach.
92
fuji (1999, 203–4; 2007, 435), G is true only if Hob and Nob share the same “guise” or “acquaintance relation,” which is also something like a way of (purportedly) identifying something. (For brevity, but without loss of generality, call these “ways”— aspects, individual concepts, etc.— MOPs.) All these metarepresentational solutions are implausible, because it is unlikely that intentional identity requires shared MOPs.45 When we assert a compound report, intending the intentional identity reading, we purport to say that the relevant intentional states have the same object. And when we really say that intentional states have the same object — when we intend the real identity reading— we need not commit to the states’ representing it by the same MOP. For example, I might say that Smith admires a man for illustrating The Giving Tree, and Brown admires the same man for writing the song “A Boy Named Sue,” without my committing to Smith and Brown sharing any MOP of Shel Silverstein. By analogy, if I intend the intentional identity reading of ‘Smith and Brown admire the same poet’, in Geach’s Ern Malley case, I might leave open whether they (purport to) identify Ern Malley differently, e.g., as the author of different poems. As Geach explains intentional identity, it requires purporting to say two intentional states share an object; but we can do that without purporting to say they share a MOP. So he should reject these same-MOP analyses. Other metarepresentational analyses allow Hob and Nob to have distinct (but related) MOPs. Tyler Burge’s sketch of a formal analysis (1983, 97–98) refers to or 45. Roberts (1996, 237) and Salmon (2002, 110) also raise this problem. Richard’s ‘Each girl fears a unicorn’ case (1998, 272) might also be a counterexample, since the girls have different MOPs; but it is not clear whether he intends that all the relevant girls (purportedly) “fear the same fictional unicorn.”
93
quantifies over Hob’s and Nob’s “applications” (acts of de re mental representation). The two applications belong to the same “quasi-anaphoric chain” (91–94); formally he represents this by giving terms the same numerical subscript, which “requires that any referent of the terms be the same” (98).46 Many who analyze G in terms of dynamic semantics (typically in the form of Discourse Representation Theory) say something similar. They say Hob and Nob have attitudes as toward “discourse referents,” most often understood as representations.47 Nearly all of those theorists say the discourse referents must be linked by communication— in the case of Hob and Nob, only indirectly, perhaps by their reading the same newspaper report of a witch sighting.48 And Walter Edelberg (1992, 572–76, 582–84) says Hob and Nob have “ideas” (in a technical sense) that are related either causally-historically or by “rough similarity of explanatory role.” (This is his nonperspectival account; for his perspectival one, see n 35 above.) Generally, these theorists seem to think of Hob and Nob as meeting a necessary but insufficient condition for having attitudes about the same thing, such as 46. Compare Breheny (2003, 8), who says there is a “notion network” such that Hob and Nob both believe there is something at the end of it. Unlike Burge’s analysis, this makes the attitudes undirected. 47. See Hans Kamp (1985, 257–59), Asher (1987, 151–55), and Samuel J. Cumming (2007, 73, 78– 79). (See also Zeevat (1987, esp. 197–99), and van Rooy (2000, 165–67, 171–73), though their solutions likely make the attitudes undirected.) Dynamic semantics, among other things, analyzes an occurrence of an indefinite description not as an existential quantifier phrase but as “introducing a discourse referent.” Depending on the theorist, this seems to mean either that it expresses a certain kind of “intentional object” (understood either as a singular representation or the represented object itself), or that it stands in for a variable bound by an existential quantifier whose scope includes the whole discourse. Only the representational version of the first interpretation is relevant here; Geach would reject analyses of G according to the other interpretations because they make the attitudes attested. See Moltmann (2006) for an overview of forms of dynamic semantics and interpretations of ‘discourse referent’. 48. Zeevat (1987, esp. 197–99) does not identify this or any other link; note that his 2000 (289, 298, 304–5, 307) is more forthcoming, but likely to the point of violating S2′ , S5, and S8.
94
having MOPs with the same source.49 But these metarepresentational solutions will not work either. The problem is that even if intentional identity would require that Hob and Nob have MOPs linked by a historical, communicative, or conceptual-role relation, G does not literally express such a relation (cf. Salmon, 2002, 110–11). Of course representations such as Hob’s and Nob’s attitudes and MOPs have histories and roles, but nothing need represent such things. The two attitude reports in G do not express anything about communicative links or conceptual roles; nor does the conjunction of reports express such a relation. Nor will such a relation appear in any plausible semantics of anaphora, or fugitive anaphora, or fugitive anaphora in intentional contexts. As Geach explains intentional identity, we purport to say that two intentional states have the same object; but we can do that without purporting to say or imply that the states have any other connection. So Geach should reject these multi-MOP analyses too. Geach has reasons to reject all the published solutions I know of. Backed into this corner, we might say that G must have a literally true reading, and so any good semantics must allow for it. But G is a hard case, and hard cases make bad semantics. The failure of proposed solutions (by Geach’s lights) does not show we need a semantic revolution; it suggests Geach is thinking of things in the wrong way. 49. Edelberg’s explanatory role criterion is the only one not requiring a common source. It is likely too weak, as van Rooy and Zimmerman (1996, 126–27) and van Rooy (2000, 272–73) note (and cf. Salmon, 2002, 111).
95
6 Assessment of Geach’s program Two of Geach’s assumptions are particularly shaky. In the puzzle, these come out in stipulations S2′ and S4, that G is true and that there is nothing Hob’s and Nob’s attitudes are about. In effect, I argued in §5.3 that G could not literally state any connection between Hob’s and Nob’s intentional states, besides sharing an object. This is inconsistent with S2′ and S4, so at least one of them is illegitimate, as is the assumption motivating it. In fact, there are reasons to doubt both assumptions. First, consider the assumption that whatever the unattested reading is, it is a literal reading; in other words, the attested/unattested distinction is one of semantics or logic, in that the unattested reading can be true while the attested is false. (And likewise for the intentional/real identity distinction.) In general, we should not assume that any distinction between two ways of using or understanding an expression must be a semantic distinction (cf. Salmon, 1991). Compare Keith Donnellan’s assumption that the distinction between referential and attributive uses of descriptions is semantic. Though this seemed obvious to him, Kripke (1979) and others disagree and argue that the distinction is pragmatic. So we should reconsider Geach’s assumption. Second, consider the assumption that the unattested and intentional identity readings do not entail corresponding attested and real identity readings, not even ones stating relations to such objects as dream girls and mythical gods. In effect, Geach dismisses out of hand realism about mythical, fictional, imaginary, and other such objects. But since he wrote, realism (especially about fictional objects) has become more
96
popular and sophisticated, too much so to ignore. And in fact there are difficult data under Geach’s nose. If I say Smith admires a poet, and I purport to identify that poet as Ern Malley, then I at least purport to commit to there being a poet whom Smith admires— that is, commit to the attested reading. Likewise, if I say Smith admires Malley, I should conclude (or purport to, as much as I purported to assert the former) that Malley is admired by Smith. (Geach would say that since I know of the Malley hoax, I am using ‘Malley’ as a quasiname, in which case it is undefined outside an intentional context, and so my second report is untrue. This is, at the least, counterintuitive and controversial.) There are also tough cases for Geach’s view that intentional predicates are nonextensional. Suppose Smith falls for the Malley hoax, and so I say Smith admires Malley. Malley is Australia’s most famous mythical poet, so it follows— or at least I may purport to validly infer— that Smith admires Australia’s most famous mythical poet. Compare Kripke’s example: “suppose the Greeks worshipped Zeus, and Zeus is the tenth god mentioned by Livy. Then the Greeks did worship the tenth god mentioned by Livy” (1973a, 3.15). So we should reconsider Geach’s second assumption too. If we loosen up on those two issues, the aim of Geach’s program is to analyze certain intentional constructions such as ‘Hannibal worships a god’, which seem to be (a) true and (b) consistent with denying that there is any object of the intentional state. Once we drop Geach’s other assumptions about these constructions, it is clear what they are: reports of the content of fiction, myth, hallucination, and the like. Compare
97
the case I discussed in chapter two: the novel A Study in Scarlet is about Sherlock Holmes, but it is somehow sensible to deny that there is such a thing as Holmes, and so deny that the novel is about “him.” This problem with reports of fiction content is unsolved; in fact it is awfully subtle and complex, involving all sorts of metaphysical and representational issues. But it is a problem we understand, because it is grounded in familiar, nontechnical matters. On the other hand, Geach explains his program in (admittedly) obscure technical terms, and assumes it concerns an elusive distinction only a skilled logician could identify. So rather than pursue Geach’s program as he understands it— in which case we reach a dead end— I loosen his assumptions and reduce it to the more familiar problem. Given that reduction, we should approach the puzzle by rejecting either S2′ or S4. Just one commentator rejects only S2′ : Crimmins says G is not literally true, but otherwise sticks to Geach’s stipulations. But for G to be pretend-true on his analysis, Hob and Nob must share a MOP, and I argued this is a mistake. Some other metarepresentational analyses do not assume this, and though they were meant as literal analyses, we might take them nonliterally. But they have other problems. Burge’s formal sketch is unclear and underspecified, as he admits.50 Hans Kamp, Nicholas Asher, and Samuel J. Cumming’s similar analyses (see n 47) are embedded in discourse representation theory, and so make the controversial, and dubious, assumption that separate sentences in the same discourse do not have independent truth conditions. And while 50. See Salmon (2002, 120–21, n 14) for some specific criticisms. And as far as I can tell Burge’s formal semantics (1983, 106–10) does not provide truth-conditions (or even pragmatic correctnessconditions) for his analysis, because it does not cover “quasi-anaphora.”
98
Edelberg’s nonperspectival account requires that Hob’s and Nob’s “ideas” be related, he does not motivate any particular sort of historical relation— in fact he leaves open whether the relation is merely similarity of “explanatory role,” which seems too weak (see n 49). However, these commentators’ informal discussions motivate a better nonliteral paraphrase: G7 There is a historical/communicative chain C, and there are MOPs a and b belonging to C, such that Hob thinks [a is a witch who has blighted Bob’s mare], and Nob wonders whether [b killed Cob’s sow]. If Hob and Nob believe the same myth— and so, purportedly, believe in the same mythical witch— then I suspect something like G7 will be true. A chain (or tree) of communication will connect Hob’s and Nob’s MOPs to a single origin, perhaps to one person thinking she saw a witch. On the other hand, if their MOPs do not share such a history, then intuitively they do not (even purportedly) represent the same thing. The general idea is to analyze the unattested and intentional identity readings as involving a pretense that a myth (or work of fiction, etc.) is true. Within that pretense we assert a certain intentional report; we might even pretend to assert it on its attested or real identity reading, making its pretended semantics straightforward. Outside that pretense, facts about the histories of certain MOPs (such as G7 reports) determine our assertion’s pretend-truth-value. This is Kendall Walton’s strategy for analyzing reports of fiction content (1990, esp. 398–405; cf. 426–30). And that strategy has
99
problems. For example, in asserting that Smith admires Malley, or that Hannibal worshipped Baal (assuming Geach is right about Hannibal’s religion), I do not seem to be pretending. If I were, then I could stop pretending, and disavow my pretense; but it is not sensible to retract my claim that Hannibal worshipped Baal— that he did seems to be plain fact rather than myth or pretense.51 I will not fully assess this strategy here, but I conclude it is not yet quite satisfying. The other approach to analyzing G begins by (really) rejecting S4, and so giving G some kind of real identity reading. But if G is literally true on its real identity reading, then there is a witch that Hob’s and Nob’s attitudes are about. If this does not follow from the use of ‘a witch’ in the first attitude report,52 it does from the use of ‘the same witch’ in the gloss of ‘she’: the reporter purports to say not only that Hob’s and Nob’s attitudes concern the same object, but that they concern the same witch. So if there are no witches, then G is false. (Likewise, if we drop the gloss ‘the same witch’, the reporter still purports to say that Hob’s and Nob’s attitudes concern the same female. But we assumed there is no female whom their attitudes are about, so again G is false.) So must we also reject S2′ and read G nonliterally? Some commentators do not; they admit that there are witches.53 Parsons says there are “nonexistent” witches, 51. Compare van Inwagen’s arguments against Walton’s views on fiction van Inwagen (e.g., his 2003, 136–38). 52. This is a complex issue; see, e.g., Erin Eaker (2009). 53. Castañeda and Priest (see n 36) say the object is a merely possible witch. But G would entail that the object is actually a witch, which Priest denies, and Castañeda does not discuss. So they do not defend an analysis of G as literally true.
100
and that Hob’s and Nob’s attitudes concern one of these. Likewise he says Sherlock Holmes is a nonexistent detective. A major problem with his view is that he thinks these objects are there independent of any human activity; for example, the witch was around independently of anyone in Gotham believing in witches. But intuitively if Hob and Nob think of anything, it is something created in the local witch-mania— just like Conan Doyle wrote stories about a fictional detective whom he created, not one that would have been around in any case.54 In a related strategy, Nino B. Cocchiarella and Salmon say ‘witch’ is ambiguous: on one reading, G is false because there are no witches1 , but on another, G is true because Hob’s and Nob’s attitudes concern a witch2 . What is a witch2 ? Cocchiarella calls it an “intensional witch.” Such “intensional objects” are “real, abstract individuals that are somehow correlated with predicable concepts,” (18). His explanation of these objects is obscure, but as far as I can tell, since he often says ‘the correlate of a concept’, he thinks there is only one correlate of any given concept— one correlate of witch, one of witch who blighted Bob’s mare, etc. One major problem is that this seems to require that Hob and Nob have attitudes about the same (intensional) witch because they share a concept (cf. 17); but I argued they need not do so. Salmon’s ambiguity view is more promising. In a forthcoming paper he says ‘witch’ can mean mythical witch— a mythical object falsely believed to be a witch.55 54. See Kit Fine (1984, esp. 131–33) for this and other criticisms. 55. His original analyses were real identity readings, but with ‘witch’ replaced by ‘mythical witch’ or ‘witch or mythical witch’. Without the ambiguity explanation, this seems like a bit of a cheat, because G does not seem literally to say anything about mythical witches. (David Braun raises this objection in an unpublished paper, to which Salmon’s paper responds. I knew of the problem independently.)
101
Compare Kripke’s claim that the general term ‘god’ is ambiguous: “It may be used in such a way that only a pagan believer of the right kind would recognize the existence of the gods on Mount Olympus. But usually we use it otherwise— for example, when we ask, ‘How many Greek gods were there?’, ‘Can you name any of the Greek gods?’, and the like” (2011, 64). Dictionaries also give some (defeasible) evidence that we acknowledge such distinctions. So this appeal to ambiguity is not ad hoc. But even if this works for G, it will not work for many other cases. ‘Poet’ does not have a special sense meaning mythical poet, so this does not help us analyze the true intentional identity reading of ‘Smith and Brown admire the same poet’ (compare my discussion of problematic general terms in §2). So Salmon’s response still misses something. I suggest we reject both of Geach’s assumptions. Unattested and intentional identity readings reduce to attested and real identity readings, and many seemingly true reports are not literally true. For example, in the Malley case the report ‘Smith and Brown admire the same poet’ really states Smith’s and Brown’s intentional relations to something, though that thing is not really a poet. In asserting that report, I speak nonliterally, but not as much as a Walton-style pretense analysis would predict. Instead, I really say there is something they both admire; the only thing I merely purport to say is that this thing is a poet. Why purport (or pretend) to say this, if it is literally false? In a nonphilosophical conversation in the context of common knowledge about the Ern Malley hoax, it is the most natural and effective way to communicate
102
what I intend to. It is more natural because we so often slip into talking as if myths, works of fiction, etc. were true. As I explained in chapter one, I think this involves a kind of mental simulation, one that is second nature to us (if not first nature, given its connections to other forms of play). Unlike slipping into simulation or pretense, talking clearly and correctly about myth or fiction is essentially an intellectual, reflective activity, which some people never learn. And saying the false report is more effective than saying ‘Smith and Brown admire the same mythical poet’, because the latter might wrongly emphasize Smith and Brown’s mistaken beliefs over their admiration, or wrongly suggest that they do know it is a myth but admire Malley anyway. (They might admire him in the sense in which one can admire Sherlock Holmes, even though no fictional character has ever really done anything admirable. This is another complex phenomenon, but I will not analyze it here.) So G can have its real identity reading: there is a witch whom Hob thinks (is a witch who) has blighted Bob’s mare, and of whom Nob wonders whether she killed Cob’s sow. What keeps the reporter from committing to (real, nonmythical) witches is not an ambiguity in ‘witch’, or a Walton-style pretense, but only the natural and communicatively effective pretense that this mythical witch is a witch. It is the littlest of white lies, because in a normal conversation, among relatively unreflective people, with common knowledge that there are no witches, the truth would likely sound stranger than the fiction (i.e., the myth). It would take more work to defend this analysis of G, and the corresponding theories of mythical objects and the means of representing them— especially in com-
103
paring it to alternatives like the Waltonian antirealist view and other myth-analogues of views on fiction. That will wait for another paper. Dozens of theorists have written dozens of papers trying to analyze Geach’s logical distinctions. Ironically, this turns out to be another case of intentional identity: there are no such logical distinctions. My goal was not to solve the puzzle, but to identify the target of all this attention — and not only purport to identify it, by the misleading name ‘intentional identity’, but really identify it. And so, in effect, I have reduced Geach’s program to the more familiar and better understood program of analyzing certain discourse about fiction. This has nothing to do with anaphora, or quantified modal logic, or many of the other issues Geach’s commentators focus on. Instead, G raises almost exactly the same problems as a sentence about fiction, such as ‘A Study in Scarlet and The Hound of the Baskervilles are about the same detective’. We have not solved those problems yet, but we know where they live.
104
Bibliography Anscombe, G.E.M. (1965). “The Intentionality of Sensation: A Grammatical Feature.” In Metaphysics and the Philosophy of Mind: Collected Papers, Vol. II, pp 3–20 (1981). Oxford: Blackwell. Asher, Nicholas (1987). “A Typology for Attitude Verbs and Their Anaphoric Properties.” Linguistics and Philosophy 10: 125–97. Barense, J.G. (1969). “Identity in Indirect Discourse.” The Journal of Philosophy 66, no 12: 381–82. Beaney, Michael, editor (1997). The Frege Reader. Oxford: Blackwell. Berger, Alan (2002). Terms and Truth: Reference Direct and Anaphoric. MIT Press. Braun, David (2005). “Empty Names, Fictional Names, Mythical Names.” Noûs 39, no 4: 596–631. Breheny, Richard (2003). “Folk Ideas About Reference and Specific Indefinites.” Direct Reference and Specificity Workshop, ESSLLI. http://www.ilg.uni-stuttgart.de/Nereus/events/03DirRefSpec/pdf/ breheny.handout.pdf.
Brock, Stuart (2010). “The Creationist Fiction: The Case against Creationism about Fictional Characters.” The Philosophical Review 119, no 3: 337–64. Burdick, Howard (1982). “A Logical Form for the Propositional Attitudes.” Synthese 52: 185–230. Burge, Tyler (1983). “Russell’s Problem and Intentional Identity.” In Agent, Language, and the Structure of the World, edited by James B. Tomberlin, pp 79–110. Hackett Publishing Company. — (2009). “Five Theses on De Re States and Attitudes.” In The Philosophy of David Kaplan, edited by Joseph Almog and Paolo Leonardi, pp 246–316. New York: Oxford University Press. — (2010). Origins of Objectivity. Oxford: Oxford University Press. Carnap, Rudolf (1947). Meaning and Necessity: A Study in Semantics and Modal Logic. Chicago: University of Chicago. Castañeda, Hector-Neri (1974). “Thinking and the Structure of the World.” Philosophia 4, no 1: 3–40. Church, Alonzo (1956). Introduction to Mathematical Logic, v 1. Princeton: Princeton University Press. Cocchiarella, Nino B. (1989). “Conceptualism, Realism, and Intensional Logic.” Topoi 8: 15–34. Cohen, L. Jonathan (1968). “Geach’s Problem about Intentional Identity.” The Journal of Philosophy 65, no 11: 329–35. Cooper, Robin (2005). “Austinian Truth, Attitudes and Type Theory.” Research on Language and Computation 3: 333–62. Crimmins, Mark (1998). “Hesperus and Phosphorus: Sense, Pretense, and
105
Reference.” The Philosophical Review 107, no 1: 1–47. Cumming, Samuel J. (2007). Proper Nouns. Ph.D. thesis, Rutgers. Currie, Gregory (1990). The Nature of Fiction. New York: Cambridge University Press. — (2010). Narratives and Narrators: A Philosophy of Stories. Oxford: Oxford University Press. Dallhöf, Matts (1995). On the Semantics of Propositional Attitude Reports. Göteborg University. David, Peter (1994). But I Digress. Iola, Wis.: Krause Publications. Dennett, D.C. (1968). “Geach on Intentional Identity.” Journal of Philosophy 65, no 11: 335–341. Deutsch, Harry (1991). “The Creation Problem.” Topoi 10: 209–25. Dummett, Michael, Gilbert Harman, Saul A. Kripke, David Lewis, Barbara Partee, and Willard Van Orman Quine (1974). “Second General Discussion Session.” Synthese 27: 509–521. Eaker, Erin L. (2009). “A New Starting Place for the Semantics of Belief Sentences.” In The Philosophy of David Kaplan, edited by Joseph Almog and Paolo Leonardi, pp 208–232. New York: Oxford University Press. Edelberg, Walter (1986). “A New Puzzle About Intentional Identity.” Journal of Philosophical Logic 15: 1–25. — (1992). “Intentional Identity and the Attitudes.” Linguistics and Philosophy 15: 561–596. — (1995). “A Perspectivalist Theory of the Attitudes.” Noûs 29, no 3: 316–342. — (2006). “Intrasubjective Intentional Identity.” Journal of Philosophy 103: 481–502. Emt, Jeanette (1992). “On the Nature of Fictional Entities.” In Understanding the Arts: Contemporary Scandinavian Aesthetics, edited by Jeanette Emt and Göran Hermerén, v 3 of Studies in Aesthetics, pp 149–76. Lund: Lund University Press. Evans, Gareth (1977). “Pronouns, Quantifiers, and Relative Clauses.” Canadian Journal of Philosophy 7, no 3: 467–536. — (1980). “Pronouns.” Linguistic Inquiry 11, no 2: 337–62. — (1982). The Varieties of Reference. Oxford: Clarendon Press. Edited by John McDowell. Everett, Anthony and Thomas Hofweber, editors (2000). Empty Names, Fiction and the Puzzles of Nonexistence. Stanford: CSLI Publications. Fine, Kit (1982). “The Problem of Non-existents: I. Internalism.” Topoi 1: 97–140. — (1984). “Critical Review of Parsons’ Non-existent Objects.” Philosophical Studies 45: 95–142. Frege, Gottlob (1892a). “On Sinn and Bedeutung.” Zeitschrift für Philosophie und philosophische Kritik 100: 25–50. Reprinted and translated in Beaney (1997), pp
106
151–71. — (1892b). “Comments on Sinn and Bedeutung.” In Beaney (1997), pp 172–80. García-Carpintero, Manuel (2007). “Fiction-Making as a Gricean Illocutionary Type.” The Journal of Aesthetics and Art Criticism 65, no 2: 203–16. Geach, Peter Thomas (1958). “History of a Fallacy.” In Geach (1972), pp 1–13. — (1963). “Quantification Theory and the Problem of Identifying Objects of Reference.” In Geach (1972), pp 139–46. Revised 1969. — (1964a). “A Medieval Discussion of Intentionality.” In Geach (1972), pp 129–38. — (1964b). “Referring Expressions Again.” In Geach (1972), pp 97–102. — (1967). “Intentional Identity.” In Geach (1972), pp 146–53. Originally published, without a footnote, in The Journal of Philosophy 64, no. 20: 627–32. — (1969). “The Perils of Pauline.” In Geach (1972), pp 153–65. — (1972). Logic Matters. University of California Press. — (1975). “Back-reference.” Philosophia 5, no 3: 192–206. — (1976). “Two Kinds of Intentionality.” Monist 59: 306–20. — (1978). “Evans on Quantifiers.” Canadian Journal of Philosophy 8, no 2: 375–79. — (1980). Reference and Generality: An Examination of Some Medieval and Modern Theories. Ithaca: Cornell, third edition. First edition 1962. — (1986). “Review of Gareth Evans, The Varieties of Reference.” Philosophical Books 61, no 238: 534–38. Glick, Ephraim (2011). “A Modal Approach to Intentional Identity.” Noûs Published online. Goldman, Alvin (2006). Simulating Minds: The Philosophy, Psychology, and Neuroscience of Mindreading. Oxford: Oxford University Press. Goodman, Jeffrey (2003). “Where is Sherlock Holmes?” The Southern Journal of Philosophy XLI: 183–97. — (2004). “A Defense of Creationism in Fiction.” Grazer Philosophische Studien 67: 131–55. — (2005). “Defending Author-Essentialism.” Philosophy and Literature 29: 200–8. Hintikka, Jaakko (1969). “Semantics for propositional attitudes.” In Models for Modalities: Selected Essays, pp 87–111. Dordrecht: Reidel. Howell, Robert (1997). “Review of Walton (1990).” Synthese 109: 413–34. Ingarden, Roman (1931/1973a). The Literary Work of Art: An Investigation on the Borderlines of Ontology, Logic, and Theory of Literature, With an Appendix on the Functions of Language in the Theater. Ivanston, Ill.: Northwestern University Press. Translation by George G. Grabowicz of Das literarische Kunstwerk, 3rd edition (1965).
107
— (1937/1973b). The Cognition of the Literary Work of Art. Evanston, Ill.: Northwestern University Press. Translation by Ruth Ann Crowley and Kenneth R. Olson of Vom Erkennen des literarischen Kunstwerks (1968). — (1962/1989). Ontology of the Work of Art: The Musical Work, the Picture, the Architectural Work, the Film. Athens: Ohio University Press. Translation by Raymond Meyer and John T. Goldthwait of Untersuchungen zur Ontologie der Kunst (1962). Kamp, Hans (1984/1985). “Context, Thought and Communication.” Proceedings of the Aristotelian Society 85: 239–61. Kaplan, David (1969). “Quantifying In.” In Words and Objections: Essays on the Work of W.V. Quine, edited by Donald Davidson and Jaakko Hintikka, pp 206–242. Reidel. — (1973). “Bob and Carol and Ted and Alice.” In Approaches to Natural Language, edited by Jaakko Hintikka, Julius Moravcsik, and Patrick Suppes, pp 490–518. Boston: Reidel. — (1986). “Opacity.” In The Philosophy of W.V. Quine, edited by Lewis Edwin Hahn and Paul Arthur Schilpp, The Library of Living Philosophers. Chicago: Open Court, second edition. — (1990). “Words.” Proceedings of the Aristotelian Society, Supplementary Volumes 64: 93–119. King, Jeffrey C. (1993). “Intentional Identity Generalized.” Journal of Philosophical Logic 22: 61–93. Kraut, Robert (1983). “There Are No De Dicto Attitudes.” Synthese 54, no 2: 275–94. Kripke, Saul A. (1973a). Reference and Existence: the John Locke Lectures for 1973. Unpublished. — (1973b). “Empty Reference.” The Shearman Lectures. Unpublished. — (1979). “Speaker’s Reference and Semantic Reference.” In Philosophical Troubles: Collected Papers, Volume I, pp 99–124. Oxford University Press. — (1980). Naming and Necessity. Harvard University Press. — (2011). “Vacuous Names and Fictional Entities.” In Philosophical Troubles: Collected Papers, Volume I, pp 52–74. Oxford: Oxford University Press. Presented in 1973. Kroon, Frederick W. (1992). “Was Meinong Only Pretending?” Philosophy and Phenomenological Research 52, no 3: 499–527. Kurafuji, Takeo (1999). Japanese Pronouns in Dynamic Semantics: The Null/Overt Contrast. Ph.D. thesis, Rutgers. — (2007). “A Demonstrative Analysis of Anaphora in Hob-Nob Sentences.” In New Frontiers in Artificial Intelligence, edited by Akito Sakurai, Kôiti Hasida, and Katsumi Nitta, v 3609 of Lecture Notes in Computer Science, pp 427–37. Berlin: Springer. 108
Lamarque, Peter (1991). “Review of Walton (1990).” The Journal of Aesthetics and Art Criticism 49, no 2: 161–66. — (2003). “How to Create a Fictional Character.” In The Creation of Art, edited by Berys Gaut and Paisley Livingston, pp 33–52. New York: Cambridge University Press. Lycan, William G. (1994). Modality and Meaning. Kluwer Academic Publishers. McKinsey, Michael (1986). “Mental Anaphora.” Synthese 66: 159–75. Moltmann, Friederike (2006). “Unbound Anaphoric Pronouns: E-Type, Dynamic, and Structured-Propositions Approaches.” Synthese 153, no 2: 199–260. Moore, George Edward (1933). “Imaginary Objects.” Proceedings of the Aristotelian Society, Supplementary Volumes 12: 55–70. Neale, Stephen (1990). Descriptions. MIT Press. Parsons, Terence (1974). “A Prolegomenon to Meinongian Semantics.” The Journal of Philosophy 71, no 16: 561–80. — (1975). “A Meinongian Analysis of Fictional Objects.” Grazer Philosophische Studien 1: 73–86. — (1980). Nonexistent Objects. New Haven: Yale University Press. Pendlebury, Michael (1982). “Hob, Nob, and Hecate: The Problem of Quantifying Out.” The Journal of Philosophy 60, no 4: 346–354. Pietarinen, Ahti (1999). “Independence-Friendly Logics and Some Non-Firstorderisable Concepts.” In Proceedings of the 9th European-Japanese Conference on Information Modelling and Knowledge Bases. — (2001). “Intentional Identity Revisited.” Nordic Journal of Philosophical Logic 6, no 2: 147–188. Predelli, Stefano (2002). “‘Holmes’ and Holmes: A Millian Analysis of Names from Fiction.” Dialectica 56, no 3: 261–79. Priest, Graham (1997). “Sylvan’s Box: A Short Story and Ten Morals.” Notre Dame Journal of Formal Logic 38: 573–82. — (2005). Towards Non-Being: The Logic and Metaphysics of Intentionality. Oxford University Press. Quine, Willard Van Orman (1966). “Quantifiers and Propositional Attitudes.” In The Ways of Paradox and Other Essays, pp 183–94. New York: Random House. Original 1956. Reinhart, Tanya (1976). The Syntactic Domain of Anaphora. Ph.D. thesis, MIT. Richard, Mark (1998). “Commitment.” Philosophical Perspectives 12: 255–81. Roberts, Craige (1996). “Anaphora in Intensional Contexts.” In The Handbook of Contemporary Semantic Theory, edited by Shalom Lappin, pp 215–46. Routley, Richard (1980). Exploring Meinong’s Jungle and Beyond: An Investigation of Noneism and the Theory of Items. Camberra: Australian National University. Russell, Bertrand (1905). “On Denoting.” Mind New Series 14: 479–93.
109
Saarinen, Esa (1978). “Intentional Identity Interpreted: A Case Study of the Relations Among Quantifiers, Pronouns, and Propositional Attitudes.” Linguistics and Philosophy 2: 151–223. Sainsbury, R.M. (2009). Fiction and Fictionalism. London: Routledge. Salmon, Nathan (1986). Frege’s Puzzle. Cambridge, Mass.: Ridgeview Publishing Company. — (1987/1988). “How to Measure the Standard Meter.” Proceedings of the Aristotelian Society 88: 193–217. — (1991). “The Pragmatic Fallacy.” Philosophical Studies 63, no 1: 83–97. — (1993). “Analyticity and Apriority.” Philosophical Perspectives 7: Language and Logic: 125–33. — (1998). “Nonexistence.” Noûs 32, no 3: 277–319. — (2002). “Mythical Objects.” In Meaning and Truth: Investigations in Philosophical Semantics, edited by Joseph Keim Campbell, Michael O’Rourke, and David Shier, pp 105–23. Seven Bridges Press. — (2004). “Two Conceptions of Semantics.” In Semantics versus Pragmatics, edited by Zoltan Szabo, pp 317–28. Cornell University Press. — (2011). “Fiction, Myth, and Reality.” In Saul Kripke, edited by Alan Berger, pp 49–77. New York: Cambridge University Press. Schiffer, Stephen (1996). “Language-Created Language-Independent Entities.” Philosophical Topics 24, no 1: 149–67. — (2003). The Things We Mean. New York: Oxford University Press. Searle, John R. (1975). “The Logical Status of Fictional Discourse.” In Expression and Meaning: Studies in the Theory of Speech Acts, pp 58–75. (1979) Cambridge University Press. Shanton, Karen and Alvin Goldman (2010). “Simulation Theory.” Wiley Interdisciplinary Reviews: Cognitive Science 1, no 4: 527–38. Slater, Barry Hartley (1988). “Intensional Identities.” Logique et Analyse 31, no 121: 93–107. — (2005). “Epsilon Calculi.” In The Internet Encyclopedia of Philosophy. http://www.iep.utm.edu/ep-calc. Thomasson, Amie (1996a). “Fiction and Intentionality.” Philosophy and Phenomenological Research 56, no 2: 277–98. — (1996b). “Fiction, Modality and Dependent Abstracta.” Philosophical Studies 84: 295–320. — (1999). Fiction and Metaphysics. Cambridge University Press. — (2003a). “Fictional Characters and Literary Practices.” British Journal of Aesthetics 43, no 2: 138–57. — (2003b). “Speaking of Fictional Characters.” Dialectica 57, no 2: 205–23. — (2007). Ordinary Objects. Oxford University Press.
110
— (2010). “Fiction, Existence, and Indeterminacy.” In Fictions and Models: New Essays, edited by John Woods, pp 109–48. Munich: Philosophia Verlag. van Inwagen, Peter (1977). “Creatures of Fiction.” American Philosophical Quarterly 14, no 4: 299–308. — (1983). “Fiction and Metaphysics.” Philosophy and Literature 7, no 1: 67–77. — (1985). “Pretence and Paraphrase.” In The Reasons of Art: Artworks and the Transformations of Philosophy, edited by Peter J. McCormick, pp 414–22. Ottowa: University of Ottowa Press. — (2000). “Quantification and Fictional Discourse.” In Everett and Hofweber (2000), pp 235–47. — (2003). “Existence, Ontological Commitment, and Fictional Entities.” In The Oxford Handbook of Metaphysics, edited by Michael J. Loux and Dean Zimmerman, pp 131–57. Oxford: Oxford University Press. van Rooy, Robert (2000). “Anaphoric Relations Across Attitude Contexts.” In Reference and Anaphoric Relations, edited by Klaus von Heusinger and Urs Egli, pp 157–81. Dordrecht: Kluwer. — (2006). Attitudes in Changing Contexts. Dordrecht: Springer. van Rooy, Robert and Thomas Ede Zimmerman (1996). “An Externalist Account of Intentional Identity.” In Proceedings of the Konstanz Workshop “Reference and Anaphorical Relations”, pp 1233–36. Voltolini, Alberto (1994). “Ficta versus Possibilia.” Grazer Philosophische Studien 48: 75–104. — (2003). “How Fictional Works are Related to Fictional Entities.” Dialectica 57, no 2: 225–38. — (2006). How Ficta Follow Fiction: A Syncretistic Account of Fictional Entities. New York: Springer. — (2009). “The Seven Consequences of Creationism.” Metaphysica 10, no 1: 27–48. — (2010a). “Against Against Fictional Realism.” Grazer Philosophische Studien 80: 47–63. — (2010b). “How Creationism Supports Kripke’s Vichianism on Fiction.” In Truth in Fiction, edited by Franck Lihoreau, pp 93–106. Frankfurt: Ontos Verlag. Walton, Kendall L. (1973). “Pictures and Make-Believe.” The Philosophical Review 82, no 3: 283–319. — (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, Mass.: Harvard University Press. — (1997). “Spelunking, Simulation, and Slime: On Being Moved by Fiction.” In Emotion and the Arts, edited by Mette Hjort and Sue Laver, pp 37–49. New York: Oxford University Press.
111
Wilson, George (1997). “Le Grand Imagier Steps Out: The Primitive Basis of Film Narration.” Philosophical Topics 25, no 1: 295–318. Wolterstorff, Nicholas (1980). Works and Worlds of Art. Clarendon/Oxford University Press. Yagisawa, Takashi (2001). “Against Creationism in Fiction.” Philosophical Perspectives 15: 153–72. Zalta, Edward N. (2000). “The Road Between Pretense Theory and Abstract Object Theory.” In Everett and Hofweber (2000), pp 117–48. — (2003). “Referring to Fictional Characters.” Dialectica 57, no 2: 243–54. Written 1984/1987. — (2006). “Deriving and Validating Kripkean Claims Using the Theory of Abstract Objects.” Noûs 40, no 4: 591–622. Zeevat, Henk (1987). “A Treatment of Belief Sentences in Discourse Representation Theory.” In Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, edited by Jeroen Groenendijk, Dick de Jongh, and Martin Stokhof, pp 189–215. Dordrecht: Foris. — (2000). “Demonstratives in Discourse.” Journal of Semantics 16: 279–313.
112