The Implied Fictional Narrator

1.1 What Do We Mean by the Term ›Fictional Narrator‹?

What is fiction? What is a fictional narrator? And what does it mean if we accept a narrator in a work of narrative fiction? Much of the debate between proponents of pan-narrator (PN) and optional-narrator (ON) as well as no-narrator (NN) theories can to some degree be resolved by agreeing on what is actually meant when someone uses the term ›fictional narrator‹. Such agreement, however, relies on a common understanding of what it means to call something fictional, and what it means to call something a narrator.

Let us first take a look at what is meant by fictional. In accordance with important prior contributions to the debate about the fictional narrator, my argument is built upon a theory of fiction in terms of make-believe, as in the widely known accounts of Walton (1990) and Currie (1990). In this framework, fictional means that works of representational art, such as books, films and paintings, are used as props in a game of make-believe to »prescribe imaginings« (Walton 1990, 51). The central idea is that recipients generate fictional truths based on these imaginings. What is true in the world of a work is what it prescribes us to imagine. If we agree on this, the issue of the fictional narrator becomes a question of fictional truths. A problem, however, is that to date there is no real consensus regarding which exact fictional truths a given work of fiction ›prescribes‹ (as Walton puts it). Additionally, the scholars behind the make-believe theory of fiction do not agree entirely on the topic of the fictional narrator: while Walton (1990, 367) tends to be »relatively liberal in recognizing narrators«, Currie (2010, 85) appears to be more reluctant and denies that it is »any sort of analytical necessity that narratives have narrators«. What they both agree on, on the other hand, is that it »doesn’t pay to get overly exercised about the question of whether all literary works have narrators or just how many do« (Walton 1990, 365).

Despite the lack of consensus on the fictional narrator in make-believe theory, however, I argue here and elsewhere (Bareis 2008) that the make-believe theory of fiction has some potentially important consequences for narratology and for the narratological category of the narrator in fictional narratives. In particular, it can be argued that if a fictional work prescribes to imagine the fictional truth of a narrator, then, we might say that there is a fictional narrator.

One challenge to this argument is the relation between narrators and the notion of direct and implied fictional truths. Not every novel starts with a sentence like »Call me Ishmael« (as in Melville’s Moby Dick, 1851), conveying the direct fictional truth of prescribing the imagining of its narrator. More than often, fictional truths are generated indirectly, depending on other fictional truths, both direct and indirect ones. Does this mean that fictional narrators can be prescribed by direct as well as indirect fictional truths? At least two views on this question can be noted from the literature. Proponents of the optional narrator theory (ON) would argue that if a fictional narrator is not generated by a direct fictional truth, there might not be a fictional narrator, while proponents of the pan-narrator theory (PN) would claim that fictional truths about narrators are always present and can be both direct and indirect.

However, if we look more closely at the relation between the narrator and direct and indirect fictional truths, it seems that all proponents of a theory of fiction in terms of make-believe agree that a substantial part of games of make-believe with representations involve indirect, implied fictional truths. It is also agreed, even by proponents of ON, that fictional narrators are sometimes generated indirectly. Far from all fictional narratives with generally accepted narrators in fact make the fictional truth of the narrator explicit, for example by including sentences such as »I am the narrator of this narrative«. Especially in cases of third-person narration, there is sometimes no explicit primary fictional truth prescribing the imagining of a narrator, but there are regularly a number of implied truths that are generated in order to understand the narrative properly, of which one is often the implied fictional truth of a narrator distinct from the author. The question at issue then is whether implied fictional narrators are a necessary ingredient of all fictional narratives, or whether there are cases of fictional narratives without any form of fictional narrator, whether explicit or implied.

Based on the arguments above, this article defines the term ›fictional narrator‹ as a direct or implied fictional truth in a work of narrative fiction. Put differently, fictional narrators can be based on both explicit and/or implied fictional truths, or on any combination of either. This definition comes itself with a set of implications. First, we can argue that what we also mean by the fictional truth of a fictional narrator is that the narrator is fictionally the origin of the narrative. In this sense, fictional narrators fictionally ›produce themselves‹ when they produce a narrative in which it is fictionally true that it is narrated. They are fictionally responsible for the narrative they narrate, and the fictional truths their existence is based on are – fictionally speaking – produced by themselves. Therefore, fictional narrators are different from authors; sentences about authors have regular truth conditions.

Before I turn to the discussion of different arguments, a short note on terminology: Walton’s theory of fiction attempts to explain fiction as a concept common to all representational arts, but not all representational art is also narrative. Whenever I use the term ›fictional narrative‹, I mean all examples of works of fiction that also qualify as narratives. One condition to qualify as fictional narrative is that there is a fictional narrator – either based on direct fictional truths, or on implied ones. This means that there can be works of fiction, such as certain plays or films, that clearly count as fiction in the sense of Walton, but that do not allow for the generation of fictional truths that they are narrated – not even the generation of implied fictional truths. Then, these works of fiction do not count as narratives.

In what follows, I will address a number of arguments that have been put forward both in favour of and against the necessary narrator in fictional narrative. Some of the arguments were originally put forward only for specific sub-classes of fictional narrative, e. g. literary narrative fiction or cinematic narrative fiction. According to the definition above, my aim is to test arguments in favour of and against the fictional narrator for all types of fictional narratives – not necessarily confined to only one media or genre. The arguments discussed are the analytical argument or a priori argument (section 1.2); the anthropomorphic argument, including realist arguments (section 1.3); the evidence argument (section 1.4); and the transmedial argument, including a discussion of unreliable narration in film (section 1.5). In the second part of this contribution, I will address the methodological and theoretical consequences of the preceding discussion for narrative theory and suggest a refined understanding of fictional narrators.

1.2 The Analytical or A Priori Argument

The analytical argument is a recurrent argument in favour of the fictional narrator and a frequent point of departure for the entire debate. It goes like this: (1) Since narratives are speech acts, (2) they presuppose someone who utters them. (3) The utterer of a narrative is called the narrator, and thus (4) there is a narrator for every narrative (cf. Köppe/Stühring 2011, 63).

Despite widespread support, the analytical argument has also received some criticism – in whole or in parts. Köppe and Stühring, for example, consider conclusion (4), that there is a fictional narrator for every fictional narrative, to be false since in their view non-fictional narrators can narrate fictional narratives. This argument strikes me as equivocal, however, since it demands that the term narrator means both ›fictional narrator‹ in cases of fictional narrative with explicit first-person narrators, and ›non-fictional author‹ in fictional narratives with no explicit narrator. Conclusion (4) is also addressed by Kania (2005), whose view is that (4) is acceptable as a conceptual point, but a weak conclusion and in need of further support.

In my proposal in Bareis (2008), I agree with Kania’s view that (4) needs further support, but on different grounds. I cannot see the necessity of understanding narratives exclusively as speech acts. Silent movies can be examples of narratives, as well as pantomimes or shadow plays, and even the description of mainstream cinema narration strikes me as a rather odd example of a typical speech act. Thus, although the a priori understanding of narrative in terms of a speech act is widespread in the literary paradigm in narratology, it is unnecessary in a wider perspective on narrative and fiction. Instead, I suggest the following adjustment of the analytical argument: (1) Since narratives present content in a certain way, (2) there has to be an agency behind both the decisions and the presentation; (3) The agency behind the narrative presentation is called narrator, and narrators differ according to the status of the narrative as either fictional or non-fictional; hence (4) fictional narratives have fictional narrators, and non-fictional narratives have non-fictional narrators.

Again, the fundamental difference between fictional and non-fictional narrative is crucial. In non-fictional narrative, the narrator is usually the same as the author or producer of the narrative. In fictional narrative, this is not true, even in cases where the fictional narrator has the same name as the author. In fiction, narrators sometimes have the name of a real person, and sometimes the identical name of the author of the fictional narrative, but this does not mean that the narrator is non-fictional. Just as Napoleon in a fictional narrative is not the real Napoleon, a real person can fictionally be a narrator (cf. Walton 1990, 356). It follows that narrators of fictional narratives are always fictional, but not necessarily fictitious (in the sense of unreal or made up). If the fictional narrator of a fictional narrative has the same name as the real author of the fictional narrative, then it is fictionally true that the author is the narrator, but it is not non-fictionally true. Non-fictional narrators cannot narrate fictional narratives. As part of a work of fiction, they are necessarily fictional.

1.3 Anthropomorphic / Realist Arguments

A group of arguments put forward against the necessary fictional narrator are based on an anthropomorphic understanding of the narrator. This line of argumentation often entails the application of a reality-based epistemology with regard to narrators. In particular, so-called omniscient narrators have been at the heart of the epistemological attacks. The main point of criticism is that it is untenable to ascribe omniscience to a fictional narrator, as Gaut (2004, 247) has argued: »The omniscient narrator is presumably a human being (the fictional teller is not usually an extra-terrestrial or God). How could a mere human being gain access to all this knowledge, often the most intimate thoughts of people which they do not tell to anyone else?«

At first sight, the anthropomorphic understanding represented by Gaut’s question seems legitimate, but I would argue that it fails on the premise that it neglects the fictionality of the text. It is not mandatory that a narrator be conceived of as a human being and that therefore realistic demands apply to her, him, or it. This is demonstrated repeatedly in the history of the novel: there are numerous non-anthropomorphic narrators at least in the western canon of literature, and non-human narrators as well as human narrators frequently ignore the laws of logic and the margins of what they could possibly narrate in epistemological terms. As constituents of fiction, fictional narrators are not limited to any form, shape, or size. They are neither forced to follow a reality-based epistemology or other basic laws of logic. Since they are fictional, they can do and know practically everything, even things they logically cannot know at all. They are not necessarily epistemologically restricted, unless the author chooses to create them thus. It can therefore be fictionally true of a narrator that she is omniscient (even though this is actually not fictionally true about most so-called omniscient narrators). Note that the fictionality premise also dismisses arguments against the implied fictional narrator, which criticise the lack of realist behaviour or the frequent self-contradictions of implied narrators. A typical (but in this case probably ironic) example of the anthropomorphisation and the application of real-life epistemology to fictional narrators is Carroll’s suggestion »to save him [the narrator, J.A.B.] the embarrassment« of being contradictory »by denying his existence« (2006, 177). Even when we take Carroll’s argument to have been made tongue-in-cheek, his suggestion would of course result in his throwing the baby out with the bath water.

Despite these problems, the application of real-world epistemology to fictional narratives is a strong and common phenomenon. Another telling example is Currie’s earlier reasoning about narrators in novels and films:

With literature it is often natural to imagine that what one is reading is a true account of certain events witnessed by someone, who then went to the trouble of setting it all down for us in writing; […] But what are we to imagine that would be analogous to this in the filmic case – that the person in the know has gone to the trouble of recreating it all for us on camera, spending millions of dollars, employing famous actors and a vast army of technicians? (Currie 1995, 21 sq.)

However, from the perspective of make-believe theory, this line of reasoning does not make sense. The points raised are close to Walton’s notion of silly questions (cf. Walton 1990, 174–183). We are not meant to ask these questions, since they undermine the very notion of fiction, the participation in a game of make-believe. In fact, there are plenty of literary examples where it remains entirely unclear how one could imagine a situation for the composition of the text.

The western tradition of fictional narrative literature is full of impossible, implausible and improbable narration. To be able to explain cases like these is the very task a theory for fictional narrative is meant to accomplish. In my view, one of the keys to the question is to regard it as obvious that fictional truths (and thus the implied narrator) are generated by means of implication, but that this does not mean that realistic implications are always at work. There are other ways to describe the implication of fictional truths. It follows, as I have argued earlier (cf. Bareis 2015), that discussions about the principles governing the generation of fictional truths need to be broadened from the Reality Principle to include other principles of generation, such as the Mutual Belief Principle, the Principle of Minimal Departure or the less known Principle of Genre Convention. This criticism seems to be Wilson’s point (2007, 85) in reply to Gaut’s position, when he argues that the Reality Principle (or the Realist Heuristic, as Wilson calls it) is a »defeasible principle«, and in cases of omniscient narration it is simply overruled by a combination of the Genre Convention Principle and something that could be called the Principle of Make-Believe, which tells you not to ask silly questions that ruin your game of make believe.

1.4 The Evidence Argument

Currie (2010, 76) has argued that the question is not »whether there sometimes or often are internal narrators«, but whether we should accept an internal narrator »when there is no positive evidence for the existence of one«. Most proponents of ON seem to share this claim. Currie gives no clear answer to what he means by positive evidence, but let us assume that he means directly generated fictional truths, since it is »granted on all hands that the work itself often makes it clear, or at least strongly suggests, that there is such a narrator« (2010, 76). Currie’s choice of words seems to imply that »the work itself« refers to direct fictional truths or, at least, easily generated implied fictional truths.

At first sight, this seems to be a legitimate demand. There is no reason to assume a narrator if there is no evidence in the work itself. Likewise, if there is evidence of a narrator, it should follow that there is one. The question then becomes: what counts as evidence? Direct fictional truths generating a narrator seem to be an obvious choice for hard evidence, while covert narrators ›who‹ can only be identified by implied fictional truths are more likely to fail the evidence test. I would argue, however, that defining what counts as hard evidence for a fictional narrator can be highly problematic – both in cases where it seems straightforward that there is a narrator and in cases where it is less straightforward. Thus, I regard the evidence claim as insufficient for determining the presence or absence of fictional narrators.

A telling example of the frailty of seemingly straightforward evidence for a narrator is Gustave Flaubert’s novel Madame Bovary, one of the classical masterpieces in realistic nineteenth century literature. From an epistemological point of view, the novel is an implausibility. It starts out as a first-person narration by a school friend of Charles Bovary, but there is very little mention of the narrating school friend throughout most of the narration, and as the novel proceeds, the first-person narrator effaces himself completely and the narrative develops into a typical piece of third-person narration, with very little evidence for the narrator’s existence. Most certainly, the narrator at the end of the book cannot be the school friend from the beginning, since the former possesses knowledge that the school friend could not possibly have. The case of Madame Bovary thus shows that the seemingly hard evidence for the existence of a narrator in the beginning of the novel does not hold for the rest of the book. Therefore, not even direct fictional truths about fictional narrators are secure evidence for fictional narrators.

It is clear, then, that despite its intuitive appeal, the evidence argument runs into problems when it comes to definitions. Counterexamples are easily found. A further, and perhaps graver problem is that the evidence argument results in a division of the class of narratives into two different types, with far-reaching consequences. On the one hand, it posits the existence of fictional narratives with fictional narrators, who (fictionally) are the producers of the narrative, and on the other hand, it produces seemingly similar narratives without a fictional producer of the narrative, i. e. texts without the decisive evidence of an ›I‹ or any other type of evidence in the text. This has also been pointed out by Margolin (2011, 50): »The ›presence-alone‹ principle, while intuitively sensible, bisects the field of narrative into two opposing and in principle mutually exclusive sub-sets, with the borderline between them being in reality inherently fuzzy«. For narratologists, this division constitutes a major drawback. It is also counter-intuitive to the practice of (at least literary) narratology, which for decades has made use of narrators in cases of heterodiegetic, extradiegetic narration without a traceable teller-figure.

But there is a further reason to question the evidence claim. Even if the fictional truth of a narrator is accepted only in those certain cases where there is sufficient evidence, this means that a fictional truth is only accepted (with all the consequences it entails) when it is direct and seemingly indisputable. Further, even if we should agree on some sort of evidence that is explicitly expressed, the evidence would still remain fictional. It therefore seems more reasonable to opt for a solution that is not evidence-based. Two possible solutions seem to follow logically from the discussion so far. Either one needs to dismiss the concept of the fictional narrator altogether to avoid the unsatisfactory fuzzy line of demarcation between narrators with more or less fictional evidence. Or, and this is my suggestion, one needs to accept fictional narrators in all fictional narratives, even in cases of low or no explicit evidence. Importantly, this acceptance of fictional narrators needs to be grounded on conceptual and methodological grounds rather than so-called evidence. In other words, it rests on a solid definition of fictionality.

In order to discuss this topic further, it is necessary to question the understanding and definition of core concepts in narratology. However, regardless of where one chooses to draw the line between narratives with or without narrators, the evidence argument relies on the acceptance of fictional truths as the basis of an analytical description of a narrative, since claims of evidence of fictional narrators necessarily rely on the generation and acceptance of fictional truths. The consequences of this will be discussed in part 2 of this article.

1.5 The Transmedial Argument: Unreliable Narration Without a Narrator?

The transmedial argument has been forwarded as an argument both for and against a fictional narrator. In some transmedial accounts the presence of a narrator (or a ›narrating instance‹) is considered obligatory. In these accounts, the point of departure is often the analytical (a priori) argument: If according to the analytical argument, all narratives have narrators, then this must be true for narratives in other media than literature, even when these are not realised as verbal narration.

An argument for the opposite view – against the presence of a narrator in all narration – has been forwarded by proponents of ON. They reject the existence of ubiquitous fictional narrators in media other than that of fictional narrative literature, i. e. verbal narration, which means that its existence is denied in other means of fictional representation, such as depictive representations in images, or mixed media with no explicit teller figure. Similarly, opponents of PN question the usefulness of importing narratological concepts developed for literary narrative into other non-verbal domains. Thomson-Jones (2007, 76), for example, has argued that the insistence on a narrator in film may have as its »ulterior motive […] the desire to fit film into a literary paradigm«, a procedure that is bound to neglect the differences between literature and film. For Thomson-Jones, the question of whether a film can have a narrator or not depends on the specific film; it can be answered with either yes or no, and the decision is not based solely on the fact that the film has a narrative structure. In the case of Thomson-Jones, the argument turns out to be a modified version of Currie’s evidence claim: If there is no sign of a narrator in a fictional narrative, including film, then there is no fictional narrator. Accordingly, the brief look at arguments of a transmedial nature do not sharpen the definition of the fictional narrator. The same arguments I have already outlined apply: evidence claims are used that entail fuzzy lines of demarcation between narratorless and narrator-mediated narratives. Even cases of seemingly clear narrator evidence can turn out to be contradictory. And again, the generation of a fictional narrator has to be based on fictional truths, thereby accepting all the consequences of fictionality for the design of narratology.

But there are other arguments of a transmedial nature that are of relevance for the definition of the fictional narrator. My example is the case of unreliable narration in films. In fact, the theoretical description of unreliable narration is a major point in my argumentation for the presence of a fictional narrator in all fictional works: unreliable narration becomes much harder to define if one does not allow for a fictional narrator, and this applies to both literary narratives and filmic narratives. As with the definition of the fictional narrator, however, we first need to define what is meant by the narratological category of unreliable narration, which is a fairly complex task.

Ever since Wayne C. Booth introduced the notion in his classical study The Rhetoric of Fiction (1961), there has been a long-standing theoretical debate about how to define unreliable narration. If we take a brief look at common sources and definitions, however, most contributors agree that unreliable narration is an effect that can occur in various narrative media and is mostly characterized as the simultaneous textual production of contradictory interpretations in an intricate interplay between the (implied) author, the narrator, and the reader. The usual view is that the unreliable narrator’s account is contradicted by the intentions of the (implied) author, and that the reader has to identify and then naturalize these contradictions in order to interpret unreliable narration adequately.

The majority of contemporary theorists of unreliable narration base their description of unreliable narration on the category of the narrator, identifying different types of unreliable narrators. Thus, James Phelan (2005, 34–37; 49–53) proposes the categories of mis- or underreporting, mis- or underreading, and mis- or underevaluating, with a strong focus on anthropomorphised narrators. If we accept this, it should follow that for those who accept the possibility of an absence of a narrator, there can be no unreliable narration. Support for this view can be found in the fact that unreliable narration is a frequent phenomenon in film and that narratological descriptions of films are not in agreement on whether narrators are obligatory. After all, cinema is regarded by many critics as an instance of narrative without narrators and thus serves as one of the strongest arguments against ubiquitous narrators.

In some accounts, however, unreliable narration does not depend on the presence of an unreliable narrator. For example Currie (1995) has argued that there can be unreliability without a narrator, and explains narrative unreliability in fiction film narratives by attributing complex intentions to the implied author. Instead of splitting up contradictory interpretations into the fallible narrators’ account and what is ›really‹ meant by the implied author, which (roughly speaking) is the general understanding in the original design by Booth, Currie’s solution attributes both the unreliable account as well as the reliable version to the implied author and his/her intentions alone. He thus sees the implied authorʼs complex intentions in analogy to real-life conversational irony. Since both the reliable and unreliable reading can be attributed to the complex intentions of the implied author alone, Currie argues that there is no necessity for the existence of a narrator even in cases of unreliable narration.

But if Currie’s view holds true, it would have the unfortunate consequence that the field of narrative would once again be divided into two sub-classes: narrative unreliability type 1 would be the equivalent of irony in non-fictional discourse in cases of fictional narratives without evidence of a fictional narrator, and narrative unreliability type 2 would pertain in cases of intriguing interplay between fictional narrators and (implied) authors in a large number of unreliably narrated fictional narratives. I would therefore like to maintain a substantial difference between real-life conversational irony on the one hand and the multifaceted device of unreliable narration in fictional narrative and its manifold manifestations on the other. If one allows for an equivalent of the fictional narrator in film, the problem does not arise.

Unreliable narration without a narrator is problematic in yet another respect: If there is only one agency responsible for the intentions of the text in cases of unreliable narration, these intentions are more than just complex; they are contradictory. Such contradictions are not as easy to construe in the reading process as it may seem at first sight, since we will need to dismiss one of the readings: either the original, reliable reading is superseded by the now superior unreliable reading, or the reliable reading prevails and the unreliable reading is dismissed. Köppe and Kindt try to avoid this very problem by suggesting that there can be »seemingly, or prima facie, authorized imaginings and actually authorized ones« (2011, 89; original emphasis). One of course immediately wonders: authorised by whom? Who is in charge of authorisation? What does it mean for an imagining to be authorised, and what is a prima facie authorised imagining? When there exists a temporary authorisation, how can we be sure that an imagining actually is authorised? What are the necessary standards? In Köppe and Kindtʼs account of unreliable narration without a narrator, it is »[t]he narration« itself or »the work W« (2011, 90) that authorises imaginings of states of affairs. However, in most other accounts of unreliable narration, it is the narrator who is in charge of the narration, and therefore it is much more obvious who is responsible for the unreliability. In the narratorless version, both readings exist more or less simultaneously.

Thus, without a narrator, attributions of unreliability become very difficult. It is therefore no surprise that Currie in his article on the topic ends up with a special focus on cases of ambiguity, which are typically narratives with two more or less equally plausible interpretations and no clear-cut and typical reading regarding narratorial reliability. These cases are sometimes labelled ›mimetically indecisive‹ (cf. Martínez/Scheffel 1999, 103) cases of unreliability or the like, but they are certainly not prototypical of unreliable narration.

The most prominent cases labelled as unreliable narration in film are instances of what in literary theory has been called fallible filters: here, the twisted world view is not attributed to the narrator, but to the filter or focaliser in the narrative. In Ron Howard’s movie A Beautiful Mind, it is the protagonist’s schizophrenic perspective that is presented to the viewer, though without the film explicitly signalling that what we see is what the protagonist John Nash is perceiving, including everything he sees in the hallucinations that he is suffering from. In other words, as viewers we are not alerted by filmic conventions that what we see does not correspond to filmic reality, that it is merely the subjective ›reality‹ of the protagonist as he experiences it. Or at least, to begin with, we are not made aware of the subjective nature of the images, or, if so, only by extremely subtle hints that are easily disregarded or not noticed the first time one watches the film. In examples like this, it would be wrong to consider the fallible filter to be the narrator of the film. It is the narrator who is still responsible for the entire narrative and all its aspects, even for the choice of a fallible filter.

In this context, the filmic example works approximately like certain cases of focalisation in literary narration. It is not K who narrates The Trial, but an anonymous, non-anthropomorphic narrator who focalises through K, and it is not Aschenbach who narrates Death in Venice, but likewise an anonymous, non-anthropomorphic narrator who chooses to focalise the narration through Aschenbach as the fallible filter. The same scenario can be argued to apply in Ron Howard’s movie A Beautiful Mind. Nash is not the narrator, but merely the character and filter through which parts of the narrative are focalised.

In A Beautiful Mind, the unreliability is resolved towards the end of the narrative, and the remainder of the film is narrated in a straight-forward, objective manner. This could be compared to Ambrose Bierce’s short story An Occurrence at Owl Creek Bridge, in which the bulk of the story consists of the depiction of a hanged man’s last-second escape, until the very last sentence of the story shatters the reader’s conceptual frame by clarifying that the escape from death never actually happened; the narrated escape represented merely the man’s delusional imaginary experience in his final moments before extinction.

It should be mentioned that there are good reasons to question if the cases above should actually count as clear-cut examples of unreliable narration, since they contain a high amount of reliability. In A Beautiful Mind the unreliability is eventually resolved concerning the reader’s understanding about how Nash experiences his reality and with regard to the filmʼs allowing the viewer to learn what is really fictionally true in the world of the film. Likewise, in the two literary cases I have mentioned, the reader is provided with the truth at the end of the texts. Examples like these are therefore probably best labelled as partially unreliable.

But there are cases of fallible filters where the unreliability remains hidden throughout the entire narrative, as for instance in the novella Der fernste Ort by Daniel Kehlmann (2001), which is modelled on the Bierce story: both texts narrate the hallucinations of their protagonists at the moment of death. While Bierce chose to resolve the mystery at the end of the narrative by letting his narrator furnish direct fictional truths, thus clarifying what has really happened, Kehlmann’s narrator choses to keep the perspective of the fallible filter intact throughout the entire story. Due to the rather subtle markers of unreliability throughout the novella, this resulted in the unreliability remaining undetected by most critics at the time the text was originally published.

In all of these cases, it is not the filter ›who‹ is unreliable, but the narrator (who in this specific case is entirely unmarked, in the sense of degree zero narration): the narrator focalises through the filter but fails to inform us that the filter is fallible. The unreliability lies completely on the side of the narrator, since the twisted worldview of the filter is presented reliably according to how the filter actually perceives the world. Again, it is not the filter but always the narrator who underreports, withholding the crucial information that the presented worldview is the worldview of the fallible filter. To whom should we attribute the unreliability in these cases, if not to the narrator?

In Currie’s (1995) account, the main quality of unreliable narration (including the discussed cases of partial unreliability) threatens to disappear from view: if unreliability is attributed to the very same entity and its complex intentions, its understanding is much closer to ordinary self-irony, instead of the much more effective ironising of a third person. It is the splitting into different entities (either the narrator and the filter, or the narrator and the (implied) author) that makes unreliable narration such a powerful tool. Thus, my argument is that unreliable narration strongly implies the existence of a narrator even in those cases where there is no ›hard evidence‹ of a narrator in the text.

The existence of unreliable narration in film is not the only valid reason for positing the existence of a narrator in filmic narratives; totally reliable filmic narrative is also in my view best explained in terms of having a narrator. I am not able to give a thorough account of this claim at this point, but I want to raise at least one question in favour of understanding fictional filmic narrative in terms of a narrator.

This argument is also based on the emphasis of the fictionality of fiction films. The argument most often cited against the view of mandatory narrators in film is probably the famous passage from Bordwell: »To give every film a narrator or implied author is to indulge in an anthropomorphic fiction« (1985, 62). I would like to widen his claim here: to give any fictional narrative a narrator, be it a film or a novel, a graphic novel or any other form of fictional narrative, is to indulge in fiction – which sometimes happens to be anthropomorphic, and sometimes not. Even in cases of direct fictional truths, prescribing that one should imagine that there is a narrator is equivalent to encouraging the reader or viewer to indulge in fiction – the difference lies solely in whether one holds direct fictional truths to be more valid than other fictional truths. Since the recognition of a narrator always implies the generation of fictional truth, I would argue that to posit a fictional narrator always means that a game of make-believe is being played with the work in question serving as a prop. Currie claims that there has to be evidence for a narrator. Köppe and Kindt go one step further: they try to solve the problem by making the claim that the literary work has to authorise the imagining of a narrator, otherwise there would not be a narrator. At the same time, they allow for the possibility of ›seemingly authorised‹ fictional truths, which can turn out to be false. Therefore, the core of the problem remains unsolved: what does it mean for an imagining or a fictional truth to be authorised? What is true in a narrative work of fiction?