• Telling Tales

Telling Tales1
Antony Eagle
Draft of December 11, 2006: please do not cite
1.
Works of fiction contain declarative sentences with a completely standard
semantics, continuous with the rest of the natural language in which those works
are written. One can easily imagine that some given group of sentences could
comprise either a novel or a biography, with no semantic alteration. One would
not need to learn anything about meanings in one’s natural language to
understand one, if one already understood the other. However, there must be
some difference between the novel and the biography, as we do not take
sentences in fictions to be attempts to describe reality; nor do we take the
authors of fictions to be asserting the content of those sentences. Moreover,
there must be something we learn when we learn the difference between novels
and biographies, which doesn’t involve the meanings of the sentences but rather
involves learning how to deal successfully with those sentences.
In particular, consider the practice of storytelling: works of fiction,
whether written down or not, are given oral expression by competent speakers of
English, say, to other competent speakers of English. Nevertheless, neither
speaker nor normal hearer take the utterances of the speaker to be attempts to
communicate the truth. Specifically, neither speaker nor hearer take the
utterances of the speaker to be assertions.
The difference between telling a tale and asserting the same sentences as a
biography must, I take it, be pragmatic. The difference must in some way involve
the force with which the semantic content of the sentences which comprise the
fiction are uttered or inscribed. If we are to explain both the continuous
semantics and our non-committal attitude, we must introduce a new kind of act
of uttering or inscribing a sentence with a given semantic content. Understanding
at least that this act exists, even if one is not sure precisely how it works, is the
1
More or less distant ancestors of this material were presented at the ANU, McGill and NYU;
thanks to audiences at those talks for helpful discussion, and to
1
key to understanding the practice of reading (and writing) novels as opposed to
false histories.
This act resembles assertion in many ways, but since assertion is
inconsistent with our non-committal attitude it cannot be assertion. Call it, by
way of a place-holder, quasi-assertion. Works of fiction differ from works of non-
fiction in that the content of a work of fiction is not asserted, but instead
(merely?) quasi-asserted. The result is that the author is not interpreted as
claiming the content to be true, which is intuitively appropriate.
Perhaps some light may be shed on quasi-assertion if we consider some
other paradigm acts of quasi-assertion, notably those that occur in discussions
about fictions. Consider a use of sentence in a class discussion of a work of fiction.
So, in a class discussion of the final stages of Pynchon’s Mason & Dixon, a student
may announce:
(1) Mason at last came to admire Dixon for his bravery. (698)
Intuitively, the student is not asserting (1). Though all of her hearers are well
aware of what the truth conditions of (1) are, no one is tempted to regard an
utterance of (1) in that context as correct iff those conditions obtain. It does not
matter for a correct utterance of (1) whether or not, in fact, Mason came to
admire Dixon—even though, in this case the literal semantic content of (1) may
well be true. But a hearer who interpreted (1) as asserting its literal semantic
content would be misunderstanding the situation in question. Rather, all of the
student’s hearers understand that for an utterance of (1) to be felicitous it has to
be the case that, according to the fiction of Mason & Dixon, (1) is true. If they’ve
read the fiction, and agree that the book is committed to (1), they will acquiesce
in the utterance of (1). It seems straightforwardly plausible that the student’s
utterance of (1) is a quasi-assertion of the semantic content of (1). Their hearers
can grasp the content of what is said, and can recognise it as declarative in form
and thus apt for an assertion, yet none take it to involve any commitment to the
content uttered.
This all suggests that someone who utters (1) in this particular context
does so appropriately exactly in case it is correct (truthful) to assert (2):
(2) In the fiction of Mason & Dixon, Mason at last came to admire Dixon for
his bravery.
2
The natural interpretation of this relationship between the quasi-assertability of
(1) and the assertability of (2) is that they are identical. That is, we arrive at:
Operator A quasi-assertion that S is an assertion that, according to, or in,
some contextually relevant fiction f, S.
A quasi-assertion of S, on this view, is a genuine assertion of a sentence S* with a
different content to S, but such that the content of S is systematically embedded
within S*. This proposal is apparently endorsed by Lewis (1978), who remarks
Thus if I say that Holmes liked to show off, you will take it that I have
asserted an abbreviated version of the true sentence ‘In the Sherlock
Holmes stories, Holmes liked to show off.’ (Lewis, 1978, 262)
Much discussion has gone into elaborating precisely what features the
operator ‘In the fiction f, ...’ should have. Lewis defends the idea that, broadly
speaking,
‘In the fiction f, φ’ is true ... iff φ is true at every possible world in a
certain set, this set being somehow determined by the fiction f.
(Lewis, 1978, 264)
One proposal is that this set should include all the worlds where F is told as
known fact (rather than as fiction); Lewis goes on to propose more sophisticated
versions of this account which needn’t detain us here. Let it suffice that we have
some intuitive understanding of how an operator like ‘in the fiction f, ...’ is
supposed to work (I will return to this issue below). The question that will
concern us more immediately is: does the Operator account provide a successful
analysis of quasi-assertion?
2.
There are a number of difficulties with the Operator account of quasi-assertion. I
shall consider three here.
Invalidity The first I call the invalidity argument; it is due to Richard Joyce
(2005, pp. 292–5). Joyce begins from the idea that there may well be arguments
3
that involve some claims taken from fictions. For instance, we might propose the
following argument:
(1) Mason at last came to admire Dixon for his bravery.
(3a) Bravery is a virtue.
(3b) Therfore, Mason came to admire Dixon for a virtue he possessed.
It is clear that we should take this argument to be valid if it were uttered in our
hearing. But if an utterance of (1) in this context amounts to an assertion of
‘According to the fiction of Mason & Dixon, Mason at last came to admire Dixon
for his bravery’, it is not clear why we should take this argument to be valid. For
the argument would seem to have the form in which the second premise and the
conclusion were tacitly prefixed, but in which the first premise was not so
prefixed. And since ‘in the fiction f, φ’ is not a factive operator, we cannot
conclude that this argument as it stands is valid.
The easy fix is to maintain that (3a) must also be prefixed; what is true of
the virtues actually is also true according to the fiction of Mason & Dixon, despite
it never being explicitly mentioned. But an utterance of (3a), once made, is not
tied to this particular argument. It may equally be pressed into service in a wholly
factual argument, and in that case there is no independent motivation for
prefixing it with the tacit operator. One can easily imagine a free-ranging
conversation about the virtues in which the speaker of (3a), at the time of
utterance, has not yet decided whether he will continue with the argument given
in (1, 3a–b), or continue to make a different and wholly factual point about the
virtue of bravery. As such, we should not interpret that utterance as necessarily
tacitly prefixed; nor should we interpret it as unprefixed, for fear of invalidating
(1, 3a–b) should that be the series of utterances he eventually makes. (1) stands
ready to be used in either fictional or non-fictional contexts, and neither
interpreting it as straightforwardly factual or fictional does justice to that
capability.
The problem is worse, actually. Because Mason and Dixon are historical
characters as well as characters in a fiction, each sentence in (1, 3a–b) is capable of
being used in both a fictional and a factual context without alteration in semantic
content. As at least some of these utterances are quasi-asserted, the only stable
option for the Operator account is to interpret ‘every claim that issues from [the
speaker’s] mouth as brimming with unspoken prefixes’ (Joyce, 2005, p. 293). And
4
this is grossly implausible; any methodologically plausible account of the meaning
of a sentence should keep to a minimum the semantic constituents that are
present neither explicitly or implicitly in the syntax (Stanley, 2000). Even if there
are sometimes unarticulated constituents, the Operator account claims that they
are tremendously common, without motivating this methodological infelicity.
Concern The second objection I call the argument from concern; I take it
from some remarks in a slightly different context by Stephen Yablo (2001, p. 76).
In fictions we can be genuinely appalled or encouraged by descriptions of the
situations in which the characters find themselves. The fact of this emotive
response must play an important role in explaining why we engage with fictions at
all, either as creators or consumers.
Yet if the Operator account of quasi-assertion is correct, what we
understand by a sentence in a fiction is a prefixed version of that sentence. So
when we are appalled by the mistreatment of debtors in Little Dorrit, the
Operator theory should maintain that what we are in fact appalled by is
(something like) the claim ‘According to the fiction of Little Dorrit, debtors are
mistreated’. But this claim is not appalling; to be concerned at the fate of debtors
is quite different to being concerned that Dickens describes their fate. If the
claim about what is contained in Little Dorrit inspires any emotional reaction at
all, it is perhaps gratitude that Dickens chose to raise the issue to prominence.
The foregoing worry, that the Operator theory fails to get the topic of our
concerns correct, also carries over to our sense of what fictional utterances are
about. The sentences of Little Dorrit are, we think, about the life and trials of
Little Dorrit, not really at all about the novel Little Dorrit. This is made obvious
when we reflect that the sentences in the novel are not about the novel itself.
Those sentences cannot be interpreted as the Operator theory suggests, unless we
think that the sentences which compose Little Dorrit are self-reflexively
mentioning the novel Little Dorrit in the content they manage to convey (see also
Joyce, 2005, p. 291). In telling a tale, we describe the claims that characterise the
content of the tale. We do not describe the tale itself; we need not mention its
length, complexity, or distinctive narrative voice, and though the content of the
tale is surely some kind of fact about it, that content is not introduced as a
feature of the tale when the tale is told.
Questions, etc. The Operator theory, strictly speaking, is an account of
quasi-assertion. But one would hope that a similar kind of act characterises the
5
way in which questions, with their normal semantic content, nevertheless occur
in a non-genuinely-interrogative sense within fictions (‘quasi-questioning’).
Prefixing a question with a sentential operator like ‘According to the fiction f’
does not, in general, even yield a grammatical sentence. And there seems little
prospect of something even vaguely similar to an operator analysis that can give
an adequate account of fictional questions, or commands. A good account of the
language of tale-telling should, it seems, be at least structurally similar across
different categories of grammatical sentence occurring within that fiction, and
the Operator theory cannot meet this demand.
3.
The Operator account claims that all quasi-assertions of S are assertions of ‘In
the fiction f, S’. But this cannot be an adequate account of how competent users
of the language manage to understand the content of a fiction, since the Operator
theory gives different content to the sentences in a fiction; namely, prefixed
content. A sentence within a tale, and the same sentence used to express a
genuine commitment of the speaker, have different semantic contents. The
Operator view is what Predelli (1997) calls a ‘replacement view’: quasi-asserting
that p is asserting, by means of an elliptical utterance of ‘p’, a different content
that replaces the content of p.
Replacement views are highly implausible, particularly if we focus on the
way in which sentences from a tale might be used both inside and outside of that
tale with what seems, prima facie, to be the same semantic content and the same
assertoric force. Imagine, for example, that (1) is uttered by A, and in response B
utters the following (incorrect) claim:
(4) Mason never respected Dixon.
B’s utterance of (4) in this context is, I suggest, perfectly understandable: B is
objecting to A’s characterisation of the events in the novel. We easily understand
that A is using the sentence in precisely the same way it is used in the original
novel. (1) does not commit A to the content expressed, and B’s utterance neither
commits B to (4) nor commits B to thinking that A is intending to assert (1) as
literally true. By far the most plausible way to understand the dialogue between A
and B is that the sentence (1) has a content continuous with that same sentence as
it appears in the novel, and that B is using (4) in a way that disagrees with that
very content, in disputing A’s claims about the novel. As the problem of concern
6
showed, the content of the sentence in the novel is not replacement content; so
A’s use of (1), and B’s use of (4), do not involve replacement content either.
4.
Interestingly, A’s utterance of (1) could just as easily have been followed by an
utterance of (3a), and the dialogue would have taken a different turn, to a
discussion of the virtues by way of literary illustration. Or A could have gone on
to recite further sentences from the novel, and the conversation would have
become A’s telling the tale of Mason & Dixon. There seem to be, then, at least
three ways in which (1) could legitimately be uttered, with the very same content
(in addition to a false assertion of (1) as known fact). (1), uttered discourse
initially, can easily go on to play any of these roles. Yet (1) has a content that can
be publically recognised and utilised by other conversational participants, so the
content of (1) must be uniform in these roles. Moreover, it is implausible that
following (1) with these various other sentences somehow retrospectively alters
the force with which (1) was uttered, so the force of (1) must be uniform in these
contexts. Obviously nothing in what I’ve said is particular to (1). Generally, for
almost any declarative fictional sentence S, a discourse-initial quasi-assertive
utterance of S stands equally ready to be used in any of the following ways: (i) as
part of telling a tale; (ii) as part of talking about a tale; or (iii) as illustrative of
some factual claim, (though not itself claimed as true).
This is fairly decisive against any account that gives an operator analysis of
sentences like (1) when used for claims about fictions. Lewis’ original theory
(1978) is not a version of the Operator theory, because he declines to give an
analysis of all quasi-assertions. He does claim that (1), when used in a classroom
discussion of the novel, abbreviates (2). This claim I dub the About-Operator
theory (OA). But the potential for (1) to be neutral between all the types of quasi-
assertion (both within and without the telling of tales), which (2) clearly is not,
shows that the division of quasi-assertions required by OA cannot succeed.
There are two responses that a defender of OA could make. The first is to
appeal to the intentions of the speaker of (1) in establishing which sense that
sentence has on a particular use. Yet it is not plausible that speaker’s intentions
can do this. If (1) is uttered discourse-initially by a speaker who hasn’t yet decided
what they will go on to say, or a speaker whose intentions for the dialogue are
hijacked by their conversational partners, that speaker nevertheless manages to
utter (1) with the same sense as an unimpaired speaker. But then (1) has sense
7
independently of the intentions of the speaker. Speakers do often intend to go on
and elaborate their utterances in various ways, and may be frustrated or annoyed
if those intentions are thwarted by other participants. Yet this annoyance is not
in any obvious way irritation at the other speakers for misunderstanding or
misusing the utterance; it is merely the commonplace annoyance at not being
allowed to speak one’s mind. As long as the speaker has the minimal aim to make
an utterance of (1) with reference in some way to a fiction, any further use of that
sentence is acceptable both semantically and pragmatically. This is good evidence
for the thesis that there is indeed a single speech act of quasi-assertion which can
play a common role in each of these ways of continuing a conversation begun
with (1).
The second response is to appeal to speech act pluralism. In defending
semantic minimalism, Cappelen and Lepore (2005) argue that every sentence has
a minimal semantic content that is context invariant, but that when a given
sentence is uttered a multitude of different speech acts are performed. In the
case of quasi-assertion, this amounts to the claim that (1) has a core semantic
content, and that when uttered (at least) three different speech acts occur:
uttering it within the fiction, uttering it illustratively, and uttering it about the
fiction. To defend OA, this third speech act must amount to an assertion of (1)
within the scope of the ‘in the fiction f’ operator. However, if (1) occurs in a
fiction, it is not plausible to think that it contributes any speech act with this
content: for fictions do not reflexively discuss themselves. So that is not a speech
act that every use of (1) can make, and there is no reason apart from special
pleading to suppose that (1) can make that speech act under the circumstances
required by OA.
These objections provide sufficient reason to reject both the general
Operator account and the specific OA account of quasi-assertion. Taking into
consideration all we have learned from the preceding discussion, we need an
alternative theory that gives a sentence S an appropriate content for each of its
potential roles in a discourse; that gives a discourse-initial utterance of S the
potential to go on to be used in any of these ways; that is not susceptible to the
problems of invalidity or concern; and that extends neatly to non-declarative
sentences. In the remainder of this paper, I will try to give such a theory.
8
5.
Operator theories take utterances about fictions as paradigmatic, and extend an
initially plausible account of those utterances to quasi-assertion in general. While
that proposal failed, a similar strategy with a different paradigm may still succeed.
In particular, if we focus on quasi-assertions used in the telling of tales, we may
get a general account.
The most natural proposal here is that telling a tale involves pretending that
what one tells is true. The teller of a tale, when uttering a declarative sentence,
is pretending ... to make an assertion, or acting as if she were making
an assertion, or going through the motions of making an assertion, or
imitating the making of an assertion... [She] is engaging in a
nondeceptive pseudoperformance which constitutes pretending to
recount to us a series of events. (Searle, 1975, 324–5)
Using this to characterise non-literary fictions is problematic (Walton, 1990, 81–
5). Walton suggests that fictions in general are to be characterised as objects that
aid the practice of make-believe; sentences in a literary fiction are ‘props’
facilitating making-believe that the content of those sentences holds, and thus
play the same role as props in other types of representational fictions. Walton
and Searle do not radically disagree; both accept that fictional discourse can only
be understood by understanding pretence and by understanding language. Searle’s
account of telling tales as pretended assertion seems roughly right as measuring
the joint contribution of fiction and language in this case. We then arrive at
Make-believe A quasi-assertion that S is an utterance of S in the course of
making-believe that the content of S obtains (Walton, 1990, 400; Currie,
1990).
This account seems quite intuitive for sentences used in the telling of a
tale itself; the teller of the tale is pretending to narrate events which didn’t
actually occur. As the creator of the tale, he is in a position to make it the case
that when he says S, S is part of the content of the tale. For the audience, then,
adequately engaging in that fictional tale apparently also involves pretending that
S is true, and that what the teller does is assert S. Teller and hearer must engage
in a joint pretence that S, whose existence and content is determined by
9
utterances of the teller (things are generally similar, though more complex, in
cases of collective storytelling by people who are both hearers and tellers).
To understand the idea of making-believe that S, it is helpful to think
firstly about supposition. If someone begins a conversation by saying ‘Suppose, for
the moment, that φ’, her hearers add φ to the background presuppositions of the
conversation. While everything that is said is on the presupposition that φ, and
the conversation as a whole is sensible only if it is understood as committed in
some way to φ, we do not take it that the any of the participants is committed to
φ. Perhaps the best model of this suppositional context is counterfactual: if one
supposes that φ, it is then appropriate to utter only those sentences S such that S
obtains in all the nearest φ-worlds. This doesn’t commit one to the actual truth of
S, though it is not incompatible with it.
I doubt that pretending can be understood directly in terms of supposing:
to do so is to conceive of tale telling as a kind of counterfactual history, and that
misses much of the point of telling tales. Nevertheless there is a similar
imperatival operator to ‘suppose that’ in the case of pretence: the ‘imagine that’
operator. This may be explicitly uttered at the beginning of a tale to introduce
the fictional context, or it may be implicitly introduced either by the conventions
of publishing or storytelling. (‘Once upon a time’ functions idiomatically as
‘imagine that’.)
What one is called upon to imagine is a situation in which the sentences
soon to be uttered are all true, or, perhaps better, a situation qualitatively
indistinguishable from a situation in which all those sentences were true. This
second articulation recognises that many fictional characters are apparently
essentially fictional, and thus exist in no possible world, even though non-actual
individuals with all the properties they are supposed to have do exist in some
possible world, and that possible world can give the content of the imagining even
if that is not what makes the sentences in the fiction true. Even if imagining and
supposing are two different kinds of attitudes, nevertheless the counterfactual
model of supposing seems to carry over smoothly to imagining: if one imagines
that φ, in the course of that imagining it is appropriate to utter only sentences S
that obtain in all the nearest φ-worlds.
While imagining is distinct from supposing, it is still the case that it
behaves presuppositionally: the content of the imagining is added to the
10
conversational background in such a way that it can be cancelled by further
utterances in the conversation. It seems presuppositions induced by imaginings
are very easily cancelled. That explains both why people hearing a tale aren’t
deceived, and why it is often inappropriate to explicitly cancel an imagined
presupposition: to object to a tale on the grounds of untruth is to cancel a
presupposition that no one was going to adopt, and to therefore spoil the
pleasures of imagination for no good reason.
Pretending and imagining are not the same thing; the teller of the tale
pretends, while the audience imagines. But we shall not go too far wrong if we
understand ‘making-believe that φ’ as the common attitude to φ shared between
audiences who imagine φ and tellers who pretend φ. Both are entertaining φ
without regard to its actual truth (and entertaining φ without regard to whether
referring terms within it denote anything), and presumably for purposes for which
the actual truth value of φ (or successful reference of its terms) is irrelevant. The
sentence is true, and its terms refer, from the perspective of the pretence, and
that is enough (Kroon, 1994, 217).
6.
The Make-believe theory is plausible as an account of quasi-assertions by the
tellers of tales. But what of the other uses we suggested that (1) could be put to?
The Make-believe theory will maintain that when someone talks about the
content of a tale, they also make-believe that the tale is told as fact. In a certain
sense, criticism and appreciation of a tale is done by imaginatively entering into
the presupposition that the tale told is true, and then speaking under that
presupposition. A critic joins with the teller in the elaboration of a tale. Yet
unlike the teller, the critic has no power to make his utterances part of the
content of the tale.
The Make-believe theory analyses a remark by A about a tale told by C as
an attempt by A to engage in the very same make-believe that C engages in. That
is, A imagines what she takes C to intend to induce in his hearers by his pretence;
and A then articulates the content of her imagination. A may not succeed in
imagining what C pretends, which is why it cannot be a condition on A’s engaging
in the same make-believe that A and C suppose exactly the same things. But there
must be substantial overlap; in particular, A should, if pressed, be willing to make-
11
believe any sentence S that is used by C in articulating the content of his
pretence.
A utters (1) correctly iff it is the case that (1) obtains in Pynchon’s
pretence, and A is entertaining that same pretence. Both of these conditions are
met; (1) appears in the novel and so must be part of Pynchon’s making-believe
that the events of Mason & Dixon obtain, and the student A is attempting to come
to grips with that very fiction by imagining what obtains in it. In uttering (1)
correctly, A needn’t be uttering (1) truly; the correctness conditions on
imaginings do not involve the truth of what is imagined. It is easy enough to slide
from ‘A says (1) correctly’ to ‘A says (1) truly’, and then have to give a theory under
which what A says in saying (1) is in fact something true; the Operator theory
makes this very slide, and then has to give implausible replacement content to A’s
utterance. Better, then, to settle with the idea that A’s utterance has correctness
conditions given by how well and accurately she captures the content of the
pretence she is attempting to engage with.
Predelli (1997) supports a similar view by considering some otherwise
puzzling utterances. Imagine a history teacher who, in an attempt to grab the
attention of his students, begins his lesson with (Predelli, 1997, 72):
(5) It is May 1940. Germany outflanks the Maginot line. Now, nothing
stands between Hitler’s troops and Paris.
This sentence is perfectly acceptable in this context. The lesson being given after
the fact, the context of utterance is not appropriate to evaluate (5), for the ‘now’
which occurs in (5) cannot refer to the time of utterance. An appropriate
semantics must take the context of evaluation to be some context which contains
May 1940 as its temporal coordinate, and so one distinct from the context of
utterance. Of course (5) isn’t true simpliciter: if it were true simpliciter, it would
be true in the context of utterance. Yet it is appropriate in some sense, as it is
true from the point of view of the time period under discussion.
While (5) isn’t fictional, it is dramatic, and what impact it has it gains
largely from the immediacy of the imagining that it demands from its hearers. As
such it seems a good model for what is happening in the cases of genuine make-
believe: the utterance of (1) might be strictly actually false. Yet as the
counterfactual account of the content of imagination suggests, the appropriate
context against which to evaluate an utterance of (1) is the world of the fiction,
12
not the actual world. What makes both (1) and (5) acceptable is the existence of a
familiar and widespread convention about the appropriateness conditions of
utterances in the context of pretence.
The view here defended is that quasi-asserting S is in all cases making-
believe that S. The difference between the role of author and critic is whether
that make-believe is creative or merely elaborating on a previous creative act, and
while this difference is intuitive I have no precise analysis to give of it. Making-
believe that S, in turn, to be understood in terms of presupposing S for the
purposes of a conversational context, though that presupposition is very simply
cancellable. This has the consequence that the relevant context of evaluation for
the correctness of an quasi-assertion is whether it obtains in a counterfactual
situation in which the content of the presupposition obtains. Whether it obtains
actually is neither here nor there.
While this view is similar to the views of Predelli (1997) and particularly
Walton (1990, 396–405), it does not share with them the idea that what A says
when she utters (1) about Mason & Disxon is in some way true, and thus that (1)
needs either to be paraphrased in some complicated way (Walton, 401–2) or we
have to adopt the non-standard contexts of evaluation as giving (1) semantic
content (Predelli). Rather, the semantic content of (1) is as it seems; if the actual
world is compatible with that content, (1) is true, otherwise it is false. But to
understand how (1) is used to comment on the fiction we don’t need to
understand it as asserted, but just as correctly used to elaborate the content of
the fiction under discussion. Thus I do not have to contort myself in attempts to
make A’s utterance anything other than a correct pretence of a kind that Walton
and Predelli both accept exists in any case.
In part, I think Walton’s desire to paraphrase (1) so that it can be true in
the mouths of critics of a fiction is explained by the continued appeal of the idea
that, somehow, by (1) A means something much like (2). (I suppose that (2) is in
fact true, but I don’t think that has much to do with whether or not (1) is true.)
Walton doesn’t subscribe to the Operator theory, of course, and so doesn’t think
that (2) paraphrases (1); but he does think that some paraphrase of (1) that makes
it come out true must be given. I see no reason why the defender of a pretence
account should do this, and I see nothing but difficulty in saying these utterances
are in any special sense true, rather than merely correctly uttered.
13
7.
The pretence view has been heavily criticised by Thomasson (1999, 97–9). She
discusses sentences like
(6) James Bond is the most famous successful spy in the world.
This sentence is not correctly uttered while pretending that the Bond fiction
obtains, because as Bond is (fictionally) a successful spy he must not be imagined to
be famous in the world of the fiction. If (6) seems true, that cannot be because it
is correct, and must instead be because it is in fact true. Certainly Bond is well
known; and no real successful spy is well known, for much the same reason as
Bond isn’t fictionally well known. So there is good reason to suppose that (6) is
actually true, which is a problem for those like Walton who maintain that all
apparent reference to fictional characters is to be understood in terms of
pretence.
Nothing in what I have said so far commits me to this view. All I have
maintained is that all quasi-assertion is to be understood in terms of pretence,
and I’ve said nothing either way about genuine assertions of sentences like (6)
which involve reference to fictional characters. But since I do in fact think that
one reason for adopting a pretence view is that it gives a way of giving literal
meaning to sentences in fictions without taking those sentences seriously, I am
reluctant to accept Thomasson’s preferred reading of (6), namely that it refers to
an abstract fictional entity JB and says something true of that thing.
Thomasson’s objections to Walton’s view rest largely on the implausibility
of his paraphrases and the methodological offence of offering ‘different analyses
of sentences of the same type ... just on the basis of the types of object they are
purportedly about’ (99). Such objections hold no weight against the present view:
I do not ascribe content to sentences occurring within fictions any differently to
the content of sentences of the same type occurring outside the fictions, as my
account of quasi-assertion is pragmatic, not semantic. In large part the present
view is motivated by giving continuous truth conditions to fictional and non-
fictional discourse without paraphrase.2
2
In fact, I think Thomasson falls foul of her own methodological scruples. Unless she maintains
that every empty name, whether fictional, mythical, or prehistorical; invented or hypothesised,
actually refers to some abstract artifact, she will give different analyses for sentences of the same
form as (6); for example
14
 This remains unconvincing unless I can give a broadly make-believe-based
account of (6). I propose to do so by suggesting that when people use fictional
names in ordinary discourse, they are then making-believe that those names refer.
Others who pick up on those empty names do so by engaging in the same make-
believe. This make-believe is of a special sort: it is making-believe that there are
fictional entities. Making-believe that there is a person, Bond, who is a spy, will not
get (6) to be acceptable. What we need to do is make-believe that there are
fictional entities—perhaps even make-believe that those entities are abstract
artifacts as Thomasson maintains—and that one of those things is the fictional
character Bond, and that thing is famous. This, in turn, is appropriate to make-
believe, intuitively, just in case the exploits of the fictional character are widely
publicised; and certainly the Bond fictions which contain details of those exploits
are themselves famous. (Brock (2002) defends a similar theory to fictional
characters from within an Operator theory; for obvious reasons I prefer the
make-believe approach here.)
Thomasson, indeed, has to accept some element of make-believe in her
artifactual theory: if Bond is abstract, while he might be famous, he is not the
kind of thing that could be a spy. So while the fictional entity might exist, we still
need to pretend of that thing that it is a spy to understand the ordinary discourse:
mixed coordinations like ‘Bond is famous and a spy’ are ordinarily acceptable and
yet false on the artifactual theory without make-believe. But if the ordinary
discourse involves make-believe in any case, there seems no need to confuse
matters by also having a non-pretended existence claim. We should also note that
make-believe is the only plausible way to understand historical novels, where the
characters really did exist and what is pretended is that those historical figures
did things other than they in fact did: the artifactual theory plays no role because
these historical characters were not created by the author.
In any case, we are rather far afield from our original topic of quasi-
assertion now. I hope that it is clear that Thomasson’s objections to the make-
believe theory are not devastating for my position, largely because I hope like her
to give a uniform semantics for discourse whether asserted or quasi-asserted.
Something still needs to be said about the semantics of sentences like (6). But I
(7) The Alvarez asteroid is the most famous candidate to explain the extinction of the
dinosaurs.
Suppose that (7) is false; then ‘Alvarez asteroid’ is an empty name. This theory was certainly not
proposed as a fiction; and it seems implausible to think that the name refers to an abstract
artifact. So there must be some account given of this empty name which differs from the account
Thomasson gives of the similar (6).
15
hope that the discussion so far at least makes it plausible that there is hope for a
general account of fictional discourse which is indifferent to the existence of
fictional objects; and that sentences like (1) and (6), at least when quasi-asserted,
can have their normal content without commitment to that content either
actually obtaining or not.
8.
Having defended the make-believe theory from Thomasson’s challenge, we
should return to the original objections to the Operator theory, and evaluate
whether the make-believe theory does any better.
Optimism seems warranted. We can easily resist Joyce's validity objection: the
semantic content of a quasi-assertion that S is still S, and hence the purely
semantic relations of consequence and entailment remain undisturbed, regardless
of whether we quasi-assert or genuinely assert. In that case, neither of the drastic
measures Joyce thinks that the operator fictionalist must appeal to are needed.
Of course there remains as issue of how we can use these arguments; valid
arguments with made-believe premises may end up with merely make-believable
conclusions. But that seems as it should be, as long as the non-merely make-
believe premises, like (3a), can be pressed into non-make-believe arguments.
As Walton (1990, 392) emphasises, the way in which a critic who quasi-
asserts a clai