In its narratological sense, unreliability is a feature of narratorial discourse. If a narrator misreports, -interprets or -evaluates, or if she/he underreports, -interprets or -evaluates, this narrator is unreliable or untrustworthy.
In literary narratives, narratorial unreliability is usually encoded by the author as a rhetorical device. Only occasionally is this due to the author’s own slips or failings in contrast to non-literary narratives, where narratorial unreliability is more often a result of the author’s own limitations. The concept of unreliability was proposed by Booth ( 1983), who was concerned with intentionally encoded unreliability in fiction. Booth discusses unreliability in relation to the concept of the implied author (Schmid → Implied Author; Shen (2011, 2013) and to that of narrative distance. In Booth’s view, a narrator is “reliable when he speaks for or acts in accordance with the norms of the work (which is to say the implied author’s norms), unreliable when he does not” ( 1983: 158–59). If the reader discovers unreliability as encoded by the implied author for the purpose of generating irony, she/he experiences a narrative distance between the narrator and the implied author, and a secret communion occurs between the latter and the reader behind the narrator’s back (300–09).
While Booth focuses on the narrator’s misreporting and ethical misevaluation, Phelan refines and extends Booth’s distinction of kinds of unreliability (Phelan & Martin 1999; Phelan 2005). Phelan points out that narrators “perform three main roles—reporting, interpreting, and evaluating; sometimes they perform the roles simultaneously and sometimes sequentially” (2005: 50). In light of these three roles, Phelan classifies unreliability by focusing on three axes: the axis of facts; the axis of values or ethics; and the axis of knowledge and perception, the last having received less attention from Booth than the other two axes.
Phelan identifies six types of unreliability which fall into two larger categories: (1) misreporting, misinterpreting (misreading) and misevaluating (misregarding); (2) underreporting, underinterpreting (underreading), and underevaluating (underregarding). The contrast between the “mis-” category and the “under-” category is basically a contrast between being wrong and being insufficient (2005: 34–37; 49–53). Significantly, one type of unreliability, Phelan points out, often interacts with other types. For instance, misreporting may be a result of the narrator’s insufficient knowledge or mistaken values, and therefore it may concur with misinterpreting or misevaluating. But of course, the narrator may be reliable in one way and unreliable in another: e.g. it is very common for the narrator to report the events accurately but misinterpret and/or misevaluate them (see also Lanser 1981: 170–72; Phelan & Martin 1999: 96).
As a significant feature of homodiegetic narration, unreliability has gradually become a key concept in narratological investigations. Critics discussing unreliability in literature fall essentially into two groups, with a certain degree of overlap between them. The first group, which far exceeds the second in number, treats unreliability as a textual property encoded by the implied author for the implied reader to decode; this group adopts a rhetorical approach. By contrast, the second group, which favors a constructivist/cognitivist approach, focuses on the interpretive process and regards unreliability as being dependent on actual readers’ divergent readings for its very existence. The following discussion will deal with the two approaches in a less historical than systematic way.
Most narrative theorists follow Booth’s “canonized” rhetorical definition of fictional unreliability (Nünning 1997a: 85). Chatman (1978) rightly points out that the domain of unreliability is the narrator’s view on the level of discourse, not the personality of the narrator (234), since the narrator’s problematic personality only forms a possible cause of unreliable narration. But Chatman’s preoccupation with the story-discourse distinction has led him to narrow down the concern to the narrator’s erroneous reporting of story facts. When unreliability occurs, the story undermines the narrator’s erroneous discourse through the implied reader’s inference of the true facts (233).
In terms of the narrator’s unreliable reporting of story elements, it is truly a clash that occurs between story and discourse; but as regards the narrator’s mis- or underinterpretation and evaluation of events and characters, it is rather between the narrator’s explicit discourse and the author’s implicit discourse that the clash can be found. Thus, in Bierce’s “Oil of Dog” (1911), the first-person narrator in his boyhood helped his “honest” mother to throw babies into a river which, he explicitly assumes, nature “had thoughtfully provided for the purpose” ( 1946: 800–01). Here the narrator’s evaluation of the facts (“honest”) and his interpretation of the facts (attributing this purpose to nature) is apparently at odds with the implied author’s implicit discourse (see Rimmon-Kenan  2002: 103).
In terms of intentionally encoded fictional unreliability, even along the axis of facts, there is still an implicit clash between the narrator’s discourse and the implied author’s discourse. This calls into question Cohn’s distinction between “unreliable narration” and “discordant narration” (2000: 307), the former only concerning the axis of facts and the latter, by contrast, having to do with the axis of values, a kind that involves a discordance between narrator and author. But as regards the factual unreliability that sets in behind the clash between story facts and discourse presentation, we still have “discordant narration,” since there is also a gap between the “mis-” or “disinformed narrator” and the accurately or adequately informed (implied) author whose norms constitute a standard by which narratorial unreliability can be judged along any axis by the rhetorical critic.
Since the gap between implied author and extradiegetic or heterodiegetic narrator is usually limited, with some exceptions (see e.g. Cohn 1999, 2000; Yacobi 2001; Pettersson 2005), narratologists have mainly dealt with unreliability in homodiegetic narration. In this kind of narration, however, the text only contains the first-person narrator’s account, and insofar as the decoding process is concerned, the “implied author’s norms” can only be a matter of the reader’s inference and judgment (see Booth  1983: 239–40). As Phelan (2005: 48) points out, flesh-and-blood readers can only try “to enter the authorial audience” with or without success.
Hansen (2007: 241–44) offers a taxonomy of four types of unreliability. The first is intranarrational, occurring within a single narrator’s discourse. The second is internarrational, where one narrator’s unreliability is “unveiled by its contrasts” with other narrators’ versions. The third is intertextual unreliability, “based on manifest character types” such as naïfs and madmen. But we find a narrator naïve or mad primarily through the deviant features of the narrator’s own discourse in light of world knowledge and genre expectations instead of through a comparison of this narrator with similar narrators in other texts. The last type is extratextual unreliability, which depends on “the knowledge the reader brings to the text” for its very existence. Because the criteria here involve a shift from text to reader (raising the question of incompatible criteria—see below), Hansen’s (2007) classification of the fourth type does not fit with his classification of the previous types, since readers with different reading strategies, conceptual frames or in different contexts may interpret the same intranarrational or internarrational phenomena quite differently.
Given the difficulties in arriving at the implied author’s norms, Rimmon-Kenan ( 2002: 7–8) draws attention to various textual features that may indicate the narrator’s unreliability: (a) contradiction between the narrator’s views and the real facts; (b) a gap between the true outcome of the action and the narrator’s erroneous earlier report; (c) consistent clash between other characters’ views and the narrator’s; and (d) internal contradictions, double-edged images and the like in the narrator’s own language. Wall (1994) highlights the first-person narrator’s peculiar verbal tics or “mind-style” (Fowler 1977; see also Shen 2005a) which “form discursive indicators of preoccupations” that “might be one of the most readily available signals that the narrator is unreliable” (Wall 1994: 20). However, as different types of texts tend to foreground different features of narratorial unreliability, Wall’s emphasis is applicable to certain texts but not necessarily to others.
As for the cause of the narrator’s unreliability, Chatman (1978: 233) mentions that it may stem from various factors such as cupidity (Jason Compson), cretinism (Benjy), gullibility (Dowling, the narrator of The Good Soldier), perplexity and lack of information (Marlow in Lord Jim), and innocence (Huck Finn). Riggan (1981) devotes a book-length study to unreliable narrators as picaros, madmen, naïfs or clowns, pointing to the relation between a deviant or deranged mind and unreliability in recounting one’s own experiences. Rimmon-Kenan ( 2002: 101–02) identifies three main sources of unreliability: the narrator’s limited knowledge; his personal involvement; and his problematic value-scheme. Fludernik (1999: 76–7) draws attention to the different causes underlying the same type of unreliability; e.g. the factual type may arise either from “deliberate lying” or from “the narrator’s insufficient access to the complete data,” or it may form “symptoms of a pathological scenario.”
Olson (2003) differentiates between “fallible” and “untrustworthy” narration, the former attributable to external circumstances and the latter caused by the narrator’s disposition. The two types of unreliability may elicit quite different responses from readers, who are inclined to justify the former according to the circumstances involved while being more skeptical and critical towards the latter. Olson’s differentiation is valuable, but the distinction would be more memorable if she used different terms such as “circumstantially unreliable” for the former type and “dispositionally unreliable” for the latter. In fact, Booth, upon whose theory Olson bases her distinction, uses “untrustworthy,” “fallible” and “unreliable” interchangeably ( 1983: 158). While Booth makes a point of including the “circumstantial” kind, asserting that unreliability is “more often a matter of what James calls inconscience” (159), Schwarz excludes the “circumstantial,” arguing that “Stevens is more an imperceptive than unreliable narrator; he is historically deaf to his implications rather than untruthful” (1997: 197). We need to bear in mind, however, that (un)reliability essentially concerns whether the narratorial discourse is able to report, interpret or evaluate events and characters correctly or sufficiently. No matter how honest a narrator is, so long as her/his discourse fails to meet these standards, the narration will remain unreliable. Just as a person’s view may change in the course of real life, the degree of a narrator’s (un)reliability may vary at different stages of the narration (see Phelan 2005, 2007; McCormick 2009).
While most narrative theorists concentrate on the ironic effects caused by unreliability, Phelan (2007) draws a distinction between “estranging unreliability” and “bonding unreliability” in order to account, in a more comprehensive and balanced way, for the effects of the technique on the audience’s intellectual, affective, and ethical relationship to the narrator. The estranging type increases the distance between the narrator and the authorial audience, while the bonding type, conversely, reduces that distance. Since most previous work on unreliability focuses on the estranging type, Phelan concentrates on bonding unreliability, of which he identifies six subtypes: (1) “literally unreliable but metaphorically reliable”; (2) “playful comparison between implied author and narrator”; (3) “naïve defamiliarization”; (4) “sincere but misguided self-deprecation”; (5) “partial progress towards the norm”; (6) “bonding through optimistic comparison”.
The constructivist approach has been pioneered by Yacobi (1981, 2001, 2005), who directs attention to how readers resolve textual incongruities with five integrating mechanisms: (1) the genetic; (2) the generic; (3) the existential; (4) the functional; (5) the perspectival. The “genetic” mechanism attributes fictive oddities and inconsistencies to the author’s production of the text, regarding them as the author’s mistakes, among other things. The “generic” principle appeals to generic conventions of plot organization such as the progressive complication and the happy ending of comedy. The “existential” principle refers incongruities to the fictive world, typically to canons of probability that deviate from those of reality, as in fairy tales or in Kafka’s “Metamorphosis.” The “functional” mechanism attributes textual incongruities to the work’s creative ends that require such oddities. And the “perspectival” principle ascribes textual incongruities to the narrator’s unreliable observation and evaluation as symptoms of narrator/author discord (see McCormick 2009 for a good application of these mechanisms).
Significantly, Yacobi’s mechanisms involve substantially different strategies—alternative rather than complementary ways of integrating discrepancies. Thus, mechanisms (1) and (5) are diametrically opposed to each other while mechanisms (1) and (2) and (1) and (4) are incompatible with each other (only the perspectival or the generic goes with the functional, the former being a specific case of the latter). These competing or contradictory mechanisms, however, may function differently for readers with different world/literary knowledge or social identity or in different cultural/historical contexts.
To understand the relation between the two approaches concerned (i.e. rhetorical vs. constructivist/cognitivist), it is important to distinguish when and how they conflict and when and how they do not. In terms of critical coverage, there is no conflict, but rather complementarity. The rhetorical approach tries to reveal how the implied reader (a critic who tries to enter into that reading position) deals with one type of textual incongruity—the gap between narrator and implied author—while Yacobi’s constructivist approach tries to show how different actual readers deal with textual incongruities in general. However, in terms of yardstick, there is a conflict between the two approaches. For Yacobi, who uses the reader’s own “organizing activity” as the guiding principle (1981: 119), all five mechanisms are equally valid (e.g. regarding the narrator’s problematic claim as the author’s own mistake is as valid as treating it as a signal of the narrator’s unreliability against authorial norms). It should be noted that many cognitivist narratologists do not share this position. Rather, they are concerned with generic readers who are equipped with the same “narrative competence” (Prince  2003: 61–2) and who share stereotypic assumptions, frames, scripts, schemata, or mental models in comprehending narrative in a “generic context” of reception (see Shen 2005b: 155–64).
From Yacobi’s constructivist angle, narratorial unreliability—concerning the perspectival mechanism—is just “a reading-hypothesis” that, “like any conjecture, is open to adjustment, inversion, or even replacement by another hypothesis altogether […] What is deemed ‘reliable’ in one context, including reading-context, as well as authorial and generic framework, may turn out to be unreliable in another” (2005: 110). This forms a notable contrast with the rhetorical approach, which treats the gap between narrator and implied author as being encoded in the text prior to interpretation. If an actual reader can decode the gap in the way intended—and signaled—by the implied author, she/he has successfully entered the position of the implied reader, and the reading is then an “authorial” reading versus a misreading.
Interestingly, when constructivist and cognitivist critics, including Yacobi, proceed with analysis of narratorial unreliability, they themselves often take recourse to the methods of the rhetorical approach. In Yacobi’s ground-breaking essay for the reader-oriented approach (1981), for instance, we see an implicit shift to the rhetorical stance. She starts by criticizing the rhetorical approach for placing unreliability in the narrator and/or the author rather than in the reader’s organizing activity (119–20). Then she draws on a scheme proposed by MacKay (1972) for differentiating information and communication: the former is defined from the viewpoint of the receiver and the latter “cannot be defined without reference to the viewpoint of the transmitter” (122). As for the literary work, Yacobi asserts that usually there is no doubt “about the very existence of communicative intent on the author’s part” and that the relations “between implied author and reader are by definition functional and hence located within the framework of an act of communication” (122–23).
Here Yacobi also considers the variability of context. However, the context is only textual, for it concerns “the modalities of the unreliable source(s) of narration vis-à-vis authorial communication” (123). Yacobi distinguishes between two kinds of unreliable narrators: the unself-conscious versus the self-conscious, the latter’s unreliability being “harder to detect than the unsuspecting monologist’s” (124). This position is unequivocally rhetorical: the implied reader “detects” unreliability through the textual features encoded by the implied author prior to interpretation. In such contexts, Yacobi is not placing unreliability “within the reader’s organizing activity” but in the narrator and the author, and consequently the yardstick of unreliability is the implied author’s norms or “overall design” (125).
Yacobi’s more recent essay (2005) is entitled “Authorial Rhetoric, Narratorial (Un)Reliability, Divergent Readings.” As shown by her own analysis (e.g. 1981: 124–25), in order to grasp the “authorial rhetoric,” a critic must try to enter the implied reader’s position so as to arrive at the authorial reading. By contrast, in interpretive practice we find “divergent readings” attributable to the differences among actual readers and various contexts. It is very important to investigate divergent actual readings—unreliability in different actual readers’ eyes—either synchronically or diachronically (see Zerweck 2001; V. Nünning 2004; Yacobi 2005). But if we acknowledge, in Yacobi’s own words, that a literary narrative is a “communicative act” that “cannot be defined without reference to the viewpoint of the transmitter,” we must avoid taking actual readers rather than the implied author as the basis for narratorial unreliability.
In the work of Ansgar Nünning, another representative of the constructivist/cognitivist approach, we also see shifts to the rhetorical position. In Nünning (1997b), a constructivist stance is adopted: “a structure is not by its nature inherent in a literary text; rather the structurality is construed by the perceiving human consciousness” (115), but it stands out particularly in the following assertion: “The information on which the projection of an unreliable narrator is based derives at least as much from within the mind of the beholder as from textual data. To put it quite bluntly: A pederast would not find anything wrong with Nabokov’s Lolita; a male chauvinist fetishist who gets his kicks out of making love to dummies is unlikely to detect any distance between his norms and those of the mad monologist in Ian McEwan’s ‘Dead As They Come’." (Nünning 1999: 61) Here the measure of unreliability rests with the ethically problematic reader’s “norms” in conflict with the implied author’s norms, a matter of the former subverting the latter. By contrast, Nünning (1997a) focuses on “the textual and contextual signals that suggest to the reader that a narrator’s reliability may be suspect” (83). In such places, Nünning’s reader is in accord with “the value and norm system of the whole text” (87) and is therefore identical with the implied reader that the rhetorical approach focuses on.
Later, Nünning (2005) attempts to synthesize the constructivist/cognitivist and rhetorical approaches. He explicitly criticizes the former approach for neglecting authorial or textual function (105), but the rhetorical approach is also criticized for failing to give sufficient attention to readers’ interpretive strategies or conceptual frameworks (91–9). Nünning’s synthetic “cognitive-rhetorical” approach asks questions such as: “What textual and contextual signals suggest to the reader that the narrator’s reliability may be suspect? How does an implied author (as redefined by Phelan) manage to furnish the narrator’s discourse and the text with clues that allow the critic to recognize an unreliable narrator when he or she sees one?” (101, emphasis added). These questions, however, come only from the rhetorical side of Nünning’s “synthesis.” The constructivist/cognitivist approach will ask very different questions such as: When faced with the same textual features, what different interpretations might readers come up with? What different conceptual frameworks or cultural contexts underlie the divergent readings?
Significantly, one can take a cognitive approach to unreliability without dropping the rhetorical yardstick. A good case in point is Vera Nünning (2004), who draws attention to different readers’ changing interpretive frames across historical contexts. The essay begins with a quote from Booth ( 1983: 239): “The history of unreliable narrators from Gargantua to Lolita is in fact full of traps for the unsuspecting reader.” Adopting Booth’s rhetorical standard, Vera Nünning tries to reveal various traps of interpretation—how different historical contexts affect readers’ conceptual schema and distort the original meaning, resulting in “misreadings” (A. Nünning 2005: 99).
We can extend the point that only the rhetorical yardstick is valid by considering conceptual frames. In investigating Nabokov’s Lolita, Zerweck (2001: 165) points out that, “depending on whether real-world frames or literary frames are applied by the individual reader,” the novel can be read in two opposing ways: either as “a highly unreliable narrative” or “as a subtle metafictional game” being played with the literary convention of unreliability. In this situation, the rhetorical critic will choose the more powerful interpretive hypothesis as the one intended by the implied author. By contrast, the cognitivist critic can merely describe opposing readings. But it is the interpretive frames that the implied Nabokov had in mind—frames that he expected the implied reader to recognize and share with him—that really count in terms of the intended meaning of the novel.
Unreliable narration “can be found in a wide range of narratives across the genres, the media, and different disciplines” (A. Nünning 2005: 90). Although both the rhetorical and the cognitivist/constructivist approaches to unreliability have focused on prose fiction, some narratologists have turned their attention to unreliable narration in film and autobiography, among other media or other genres. Chatman (1978: 235–37, 1990: 124–38) extends the discussion of unreliability to film, where more dramatic effect may emerge, since a voice-over depicting story events may be belied by what the audience sees on the screen. Interestingly, the cinematic camera can also be used to mislead the audience temporally for certain effects (Chatman 1978: 236–37, 1990: 131–32; see also Currie 1995; Bordwell 1985; Kozloff 1988).
As regards the non-fictional genre of autobiography in the verbal medium, there are, on the one hand, the same manifestations of unreliable narration as in fiction: misreporting, -interpreting, -evaluating or underreporting, -interpreting, -evaluating. On the other hand, misreporting and underreporting figure much more prominently here, since in this “non-fictional” genre, whether the report is accurate or adequate often forms the focus of attention. In terms of this “factual” kind of unreliability, while in fiction—whether verbal or visual—the indicators are usually intratextual problems (textual inconsistencies or incongruities), in autobiography, the case is more complicated, since unreliability can occur not only at the intratextual level but also at the extratextual and intertextual levels. If the events depicted in an autobiography, however consistent the text itself is, do not tally with extratextual reality, we will be faced with “extratextual unreliability”; and if two or more autobiographies representing the same life experiences do not accord with each other, this will result in “intertextual unreliability” (see Shen & Xu 2007 for a detailed discussion).
In terms of the relation between author and narrator, there exists an essential difference between autobiography and fiction. In autobiography, the (implied) author and narrator often collapse into one, since it is usually “an art of direct telling from author to audience” (Phelan 2005: 67) where the author is the narrator. As distinct from fiction, unreliability, in the autobiographical norm of “direct telling,” is usually a matter of the “cognizant” reader’s judgment at the expense of the “I” as the second self of the narrator-author (Shen & Xu 2007: 47–9). Moreover, in autobiography, markers of “factual” unreliability exist that are not found in fiction, e.g. features indicating that the autobiographer (author-narrator) is fictionalizing her/his experiences (see Cohn 1999).
As a non-fictional genre, autobiography shares essential characteristics with other non-fictional narratives, such as those in news reporting or daily conversation. What has been said about autobiographical unreliability therefore applies in varying degrees to narratorial unreliability in other types of non-fictional narratives as well (see Currie 1995: 19; cf. Fludernik 2001: 97–8; Bamberg → Identity and Narration).
(a) Unreliable narration in non-verbal media and in verbal genres other than prose fiction. (b) In prose fiction, unreliability in postmodern fiction, second-person narration, simultaneous narration, etc. (c) Unreliability in poetry, e.g. in the “dramatic monologue.” (d) The relation between unreliable narration and gender, class or racial issues. (e) In dealing with textual incongruities, whether there are other integration mechanisms or conceptual frames apart from those already identified. (f) How different critical theories lead to different conceptions of the same textual incongruities. (g) When a text is translated into another language, how the different cultural context with different social norms bear on the interpretation of unreliability. (h) Whether there are other causes underlying unreliable narration. (i) Whether there are other indicators of unreliable narration. (j) How to carry out a rhetorical investigation of unreliability more effectively, especially in terms of a text produced in a different historical or cultural context.