Narrative levels (also referred to as diegetic levels) are an analytic notion whose purpose is to describe the relations between an act of narration and the diegesis, or spatiotemporal universe within which a story takes place. At the outermost level, external to the intradiegetic (or diegetic, i.e. first-level) narrative, the extradiegetic narrator recounts what occurred at that first level; a character in that story can, in turn, become an intradiegetic narrator whose narrative, at the second level, will then be a metadiegetic narrative. This process can extend to further meta-levels, forming a series of narratives patterned recursively in the fashion of Chinese boxes or Russian dolls. Characterized by a relation of inclusion, narrative levels are distributed vertically when a change of both (diegetic) level and speaker and/or addressee occurs, and horizontally when no change of speaker takes place (as in a digression) or when several parallel stories are recounted by different speakers but at the same narrative level (as in Boccaccio’s Decameron). Narrative levels are most accurately thought of as diegetic levels, the levels at which the narrating act and the narratee are situated in relation to the narrated story.
According to Genette, who first proposed the term, narrative levels are one of the three categories forming the narrating situation, the other two being the time of the narrating (subsequent, prior, simultaneous or interpolated) and person (heterodiegetic or homodiegetic) ( 1980: chap. 5). Introduced for the purpose of systematizing the traditional notion of embedding, narrative levels mark “the threshold between one diegesis and another,” and more particularly “the fact that the second diegesis is taken charge of by a narrative fashioned within the first diegesis” (Genette  1988: 84, original emphasis). This threshold results from the fact that as every narrative, beginning with the first-level narrative, is produced by an act of narration which is of necessity external to that level: “any event a narrative recounts is at a diegetic level immediately higher than the level at which the narrating act producing this narrative is placed […]. The narrating instance of a first [i.e. first-level] narrative is therefore extradiegetic by definition, as the narrating instance of a second (metadiegetic) [i.e. second-level] narrative is [intra]diegetic by definition, etc.” (Genette ( 1980: 228–29, original emphasis). It is important to bear in mind that first-level narrative (récit premier or récit primaire) is intradiegetic and not, as stated by some commentators, extradiegetic.
Narrative levels are frequently understood to correspond to narrative framing or embedding. The two notions coincide to some extent, but it is essential to remember that narrative levels extend into areas not generally taken into account in non-narratological discussions of framing and embedding. From the perspective of narrative levels, framing or embedding occurs between the intradiegetic and the metadiegetic levels—not between the extradiegetic and the intradiegetic levels: narrative levels come into play at all three levels, even in the absence of any frame story (or metadiegetic narrative), it being important to remember that extradiegesis, where the narrative act occurs, lies “outside” the intradiegetic level. Lanser (1981: 134) puts it quite simply in postulating levels A, B and C, where a tale within a tale corresponds to level C (cf. Fludernik 1996: 342; Wolf 2006a: 181). In an attempt to resolve certain difficulties found in accounts of narrative framing or embedding, Genette (not without analogy to the differentiation between “who speaks?” and “who sees?” in his analysis of point of view and focalization) distinguishes level from voice. A second aspect of narrative levels is that they operate in close conjunction with voice, constituting a four-part typology of narrator status.
Another point is that much discussion about narrative levels has resulted from two apparently incompatible ways of organizing them: by definition, levels are distributed vertically whereas framing and embedding are operations that involve inclusion. Genette embraces both, stating that a narrative event “is at a diegetic level immediately higher than” the narrative act ( 1980: 228) but ultimately pleading in favor of an inclusionary relation, as illustrated with a series of stick figures and balloons ( 1988: 85–6) where the second-level narrative is “inside” the first-level narrative. The contradiction between these two configurations has led some critics to revise the original concept and others to reformulate the concept in different terms.
Finally, a third set of issues, closely intertwined with the previous two, concerns the types of relations between intradiegetic narrative and metadiegetic narrative, extending from the explanatory to the thematic to the narratorial. Succeeding discussions have sought to specify the functional nature of these relations.
One of the principal interests of narrative levels, a concept formulated by structuralist narratology, is that it has been effective in setting out terms for re-examining the traditional approaches to framing and embedding and in opening up new lines of debate and inquiry. Accordingly, this section comments on embedding and framing from the perspective of narrative levels (3.1), outlines a number of the responses to the various configurations of these levels (3.2) and comments on the functional relations between intradiegetic and metadiegetic narrative (3.3).
The terms embedding and framing themselves merit some clarification. Often used synonymously, the domains covered by the two notions are somewhat different. Embedding, along with linking and alternation, represents one of the ways that narrative sequences can be combined within a narrative instance or in different ones, and in this sense it is a device that pertains to story (histoire), independently of any change of level (Prince  2003: 5, 25, 48–9). The corresponding terms employed by Bremond (1973: 132) are, respectively, enclave (one sequence developing within another), bout-à-bout (the end of one sequence succeeded by the beginning of another) and accolement (bracketing simultaneous sequences together). Similarly, Todorov (1972: 379) proposes enchâssement (order: 1-2-1), enchaînement (order: 1-2), and entrelacement or alternance (order: 1-2-1-2). Earlier, however, Todorov ( 1973: 83–5), employing the same terms, had linked embedding to narrative levels, thus associating it with discours. In “Narrative-Men” (originally published in 1967), it is stated that embedding coincides, not gratuitously, with the syntactic category of subordination in modern linguistics (e.g. “Scheherazade tells that Jaafar tells that the tailor tells that…”), and it is concluded that “embedding narrative is the narrative of a narrative” (Todorov  1977: 71). Strictly speaking, however, likening narrative embedding to the concept of embedding in transformational grammar, a concept developed in place of subordination in traditional grammar, is not defensible: a sentence such as “Hamlet knew that his father had been murdered” cannot be described as an example of narrative embedding (cf. Pier 2011: 120–21). In order to avoid any misleading superimposition of linguistic categories on narrative categories, Greimas and Courtés (1979: 123) prefer to speak of “intercalation” in narrative texts rather than embedding. It is thus useful to bear in mind that even though embedding is the consecrated term in narrative theory, the process concerned is actually one of intercalation, the insertion of one story in another, i.e. metadiegetic narration, a relation which is not, in all cases, one of subordination.
Emphasizing intercalation as the specific narratological sense of embedding serves both to stake out the parameters of the concept and to avoid the risk of assimilating it into phenomena that are actually of another nature. With reference to the criteria of punctuation and continuum, boundary and logical levels that characterize embedding in fields as diverse as linguistics, logic, psychology, communication, computer science, Füredy (1989) identified the more extreme forms of embedding found in artistic representation: (a) intact and multiplying boundary (e.g. mise en abyme, which in principle is open to infinite recursion); (b) intact but reified boundary (escape from the undecidable and oscillating boundary built into Escher’s Drawing Hands is possible only through access to an otherwise inviolate metalevel); (c) transgressed boundary (metalepsis; Pier → Metalepsis). Ryan (1986, 1991: 156–74) employs the term “embedded narrative” in a way she characterizes as “idiosyncratic” (1991: 274, n. 2) but which, requiring no speech act or verbalization, is nonetheless a logical extension of the principle as outlined above. For her, embedded narratives are not only narratives that “reflect the events of the factual domain” but also those that “delineate unactualized possibilities” such as “dreams, fictions, and fantasies” as well as “plans, passive projections, desires, beliefs concerning the history of TAW [textual actual world], and beliefs concerning the private representations of other characters.” (156)
In the field of conversation analysis, by contrast, embedding, referred to as “embeddedness,” concerns not a change of level but the context of surrounding discourse and social activity. Thus a narrative of personal experience may be embedded in an explanation or a prayer and will be more or less embedded into the surrounding social activity according to the frequency and length of turn-taking and the degree of thematic and rhetorical integration into the general conversation (Ochs & Capps 2001: 36–40; on the related notion of “situated communication,” see Young 1987: chap. 4). Indeed, if the story within the story is more characteristic of written narrative than it is of oral storytelling (Fludernik → Conversational Narration – Oral Narration), this is largely due to the attempt to restore a sense of orality to the written text and to simulate oral storytelling. Yet another angle on embedding is taken by deconstructive approaches to narrative. Considering that story is embedded in the objects, subjects, and bodies of the world, that they are in effect “texts,” these theories tend to assimilate text into its contexts (cf. Punday 2003), thus stepping beyond the question of how one story is embedded into another.
If embedding can be thought of as inserting or placing something within a larger unit, framing is normally understood in the sense of enclosing. This nuance stands out in the earliest definition of the frame tale: “The concept is taken from framed pictures and this means that one tale encloses [umschliesst] another like a frame,” its two forms being “cyclical frame tales” and “framed individual novellas” (Merker 1928–29: 1; for a historical account, see Jäggi 1994 and Kanzog  1977; for framing in different genres, see Duyfhuizen 1992; for frame tales in Indian, Arabic and other cultures, see Picard 1987; Williams 1998: 104 comments on the frame metaphor from the visual arts). In addition to Merker’s two forms, involving “one-story framing” and “plural-stories framing” (or “interpolated framing”), note that a narrative frame can be “complete” or “closed,” but also that an introductory frame may be paired with a “missing terminal frame” just as a terminal frame may be paired with a “missing initial frame” (Wolf 2006a: 185–88; cf. Fludernik  2009: 28–9).
Framing is generally regarded as a presentational technique: the frame tale is of limited length and varying significance, serving to render the more ample inset or inner tale (Binnenerzählung) accessible and/or to authenticate it, imbuing it with a “narratorial illusionism” (Nünning 2004: 17), particularly in simulations of oral storytelling. Many but not all authors rely on quantitative criteria to characterize frames. Thus Williams (1998) argues that “a frame articulates a discernable narrative scenario, focusing on the rhetorical dynamic of narrative exchange” (107–08), and he goes on to outline a typology of preliminary, introductory and prologue frames (120–25). However, Williams does not appear to distinguish consistently between the frame and the framed and at one point even inverts them: “Framed narratives specify place and time—a setting—for the act of narrative” (110). Embedding, which, technically, also occurs in frame tales, is not concerned with the presentational relations between the two levels, but rather with identifying the “threshold” or differential relation between the narrating act and diegetic level. The principle of narrative levels does not seek to sort out these distinctions, and indeed Genette’s discussion wavers between narrative subordination and thematic precedence precisely when he takes up the question of framing ( 1988: 86–90).
Fludernik, speaking from the holistic perspective of natural narratology, seems to be one of the few commentators to have distinguished between framing and embedding. She observes that “[w]ith regard to length, frame and inset are […] in inverse proportion to the relation obtaining between a story and the embedded story within it,” and she concludes: “If the tale is conceptualized as subsidiary to the primary story frame, a relationship of embedding obtains; if the primary story level serves as a mere introduction to the narrative proper, it will be perceived as a framing device.” (Fludernik 1996: 343) This distinction is vital since, notably, it is only in the first case, where the embedding discourse is dominant, that a mise en abyme can occur, as in portions of the romance The Mad Trist that parallel certain incidents in Poe’s “The Fall of the House of Usher” (cf. Dällenbach  1989). A sort of “reverse” mise en abyme is the mise en cadre, where an element of the subsidiary frame may proleptically illustrate some feature of the dominant embedded story, as in Conrad’s Heart of Darkness (cf. Wolf 2010).
Narrative levels differ from traditional concepts of embedding and framing because they articulate these questions in significantly new terms. Indeed, the concept cannot be isolated from other aspects of Genettian narratology, notably the time of the narrating and person which, together with narrative levels, constitute the narrating situation. A full treatment of narrative levels would thus bear not only on embedding and framing but would also take into account the so-called narrator’s status. The narrator’s status, broken down into the well-known four-part typology of narrators combining level (extradiegetic/intradiegetic) and the relationship of presence or absence of the narrator in the diegesis (heterodiegetic/homodiegetic) (see Genette ( 1980: 248), is a topic that has attracted a considerable amount of commentary and revision. Here, mention can only be made of some of the issues raised by the two components of the narrator’s status. (For an explanatory note on Genette’s use of the term diegesis, see Pier  2010; for a discussion of person, level and voice, see Walsh 2010.)
In his discussion of narrators and levels, Schmid ( 2010: 67–70) proposes to make Genette’s system more user-friendly by modifying the terminology. In place of extra-, intra- and metadiegetic, the terms primary, secondary and tertiary are employed to designate narrators and levels of embedding while diegetic vs. non-diegetic is adopted instead of homo- vs. heterodiegetic to replace the traditional first-person/third-person dichotomy. The translation of one system into the other is thus as follows: extra- heterodiegetic narrator → primary non-diegetic narrator; extra- homodiegetic narrator → primary diegetic narrator; intra- heterodiegetic narrator → secondary non-diegetic narrator; intra- homodiegetic narrator → secondary diegetic narrator; meta- heterodiegetic narrator → tertiary non-diegetic narrator; meta- homodiegetic narrator → tertiary diegetic narrator. These emendations do clarify some of the terminological issues, but it should be pointed out that Schmid’s schema, extending from the primary to the tertiary level, presents itself as a system of “levels of embedding, the degree of framing” (67), not as a typology of narrators (for which a separate set of criteria are enumerated; cf. 66–7). In Genette’s system, as already pointed out, embedding occurs between the intradiegetic and the metadiegetic levels (Schmid’s secondary and tertiary levels), not between the extradiegetic and intradiegetic levels (primary and secondary levels). For Schmid, by contrast, the secondary level of narration functions as a “quoted world” of the primary level and thus already as framed or embedded narrative. Genette works out a typology of narrators that includes no metadiegetic (or tertiary) level; the question of embedding and framing is an extension of that typology that raises a specific set of issues.
As can be seen from the above example, the commentaries that the notion of narrative levels has given rise to and the revisions put forth by various authors center in large part around the prefixes added to the word diegesis and the vertical and horizontal dimensions of embedding. The prefix “meta-” in particular has drawn considerable attention, for Genette does not employ the term metadiegetic in the sense of metalanguage, i.e. a language used to speak about an object language. Rather, “the metanarrative [metadiegetic narrative] is a narrative within the narrative, as the diegesis […] designates the universe of the first narrative.” (228 n.1, original emphasis) (Note that metanarrative [métarécit] must not be confused with Lyotard’s grand récit, sometimes translated as “metanarrative,” or with “metanarrative comments”; cf. Nünning 2004: 15; Neumann & Nünning → Metanarration and Metafiction.)
In one of the best-known critiques, Bal, adhering to a metalinguistic perspective, redefines embedding in terms of subordination, dominance and hierarchy: “[a]n embedded unit is by definition subordinate to the unit which embeds it” (1981b: 48). In place of “meta-” she thus adopts “hypo-” (meaning “under”) and proposes to replace “metadiegetic”/ “metadiegesis” with “hypodiegetic”/“hypodiegesis,” reserving “meta-” to the “superior level” (45; cf. Bal 1977: 35; Rimmon-Kenan  2002: 92–6 adopts this system of “subordination relations” between levels). As a result, Genette’s system of narrative levels is inverted so that it is the narrative act that is above or higher than the narrative event.
In his reply Genette reaffirms the “inclusionary” approach as opposed to the metalinguistic conception, maintaining that a metanarrative occurs “within” the narrative, that it is not a narrative “on” narrative ( 1988: 91–2). Actually, Bal partly rallies to this position when she states that quoted discourse is metalinguistic in relation to embedding discourse but that metadiscourse, when it is without quotation marks, must be qualified as “hypo-discourse” (1981b: 54–5). Genette further stresses the essential connection between metadiegetic and metalepsis, a connection which is blurred when hypo- replaces meta-.
Despite this realignment, the term hypodiegetic continues to be widely used in place of metadiegetic. Adopting metalanguage as a model for narrative embedding, however, is not unproblematic: an embedding discourse (by which Bal means quotation) is not a metalanguage, nor is an embedded discourse an object language; and while an embedded discourse might be said to “depend” on the embedding discourse by which it is taken in charge, this is hardly the case of an object language examined with the use of a metalanguage (indeed, quite the opposite is true). Another point is that diegesis, in the sense of the spatiotemporal universe in which the story takes place, is largely abandoned by Bal, for in her system hypodiegetic narrative results from the embedding of the subject and object of narration, focalization and acting—a synthesis of Greimas’s actantial roles and of Genette’s focalization (1981b: 45; on the embedding of focalization, see Bal 1981a). She distinguishes between embedding and framing on the basis that in the former there occurs subordination of both actor and action whereas in the latter only one or the other is subordinated. Interestingly, narrative embedding is taken up again later under the heading “Levels of Narration,” but with no reference to hypodiegetic/hypodiegesis, concentrating instead on speech representation and the relations of embedding between fabula and text (Bal  1997: 43–66).
Bal’s approach to narrative levels, carried out within the framework of structural linguistics, was succeeded by models that postulate two types of embedding: vertical, occurring in shifts between levels; and horizontal, without change of level but narrated by different narrators. A good example of this tendency is Nelles (1997), who, critical of Bal’s account of voice, subscribes to this distinction. He goes on to present a number of cases in which a vertical change of levels takes place accompanied not with a change of narrator but a change of narratee (e.g. the general narrator of The Canterbury Tales, who relates a second-level tale). He also notes examples of horizontal embedding (such as dreams) where no change of narrator takes place and where, rather than a change of level, there is a change in the nature of the diegesis or universe within which the story takes place (132–33). The events of the dream, he explains, “take place in an alternate universe created by a character’s mind rather than being physically carried out in the spatial-temporal universe of the rest of the narrative” (134). This adds a significantly new dimension to the question of narrative levels, leading Nelles to distinguish, with reference to McHale’s (1987) characterization of the “epistemological” dominant of modernist fiction as opposed to the “ontological” dominant of postmodernist fiction, and also analogously to Ryan (cf. § 3.1 above), between “epistemic” or “verbal” embedding (communicating knowledge) and “ontological” or “modal” framing (modes of being). García Landa (1998: 303–4) points out that the latter form of framing, which he calls “semiotic insertion,” is associated with Bal’s embedded focalization.
With regard to verbal embedding, one possibility is to consider this from an enunciative perspective. Coste (1989: 165–74), for instance, gives precedence to the notion of “overall narrator” over the distinction homodiegetic vs. heterodiegetic narrator. He then sharply separates the subject of the enunciation from the subject of the enunciated, breaking down the subject as narrating instance into present storyteller and past (or future) character. Based on these and other premises, Coste sets forth two types of embedding: hypotactic, resulting from grammatical subordination and materialized in the form of delegated narration; paratactic (juxtaposition, coordination), forming a system of “parallel” narrators at the same diegetic level. In its complex forms, hypotaxis, a grammatical form that also applies to framing, tends to blur the origin of enunciation, resulting in effect in pseudo-diegetic narration: a narrative second in origin but which, lacking a diegetic relay, is narrated as though it were diegetic (Genette  1980: 237–43). As for narrational parataxis, this occurs when, without change of level, narratives are combined in one of three ways: by sequential relay (several narrators tell the same story chronologically); through concurrent/conflictive versions of the same story; in narrational crossfire (as in Faulkner’s As I Lay Dying, where the central character, Addie Bundren, is absent). By highlighting discrete narrational acts rather than the threshold between levels, parataxis is conducive to multiple points of view and polyphonic narration, and is thus related to dialogism (Coste 1989: 188–205; Shepherd → Dialogism), tending to call into question the very notion of narrative levels. Coste’s model departs from the traditional focus of embedding and framing on the story within the story; however, as the distinction hypotactic vs. paratactic narration is configured vertically and horizontally, the model does nevertheless echo multi-level embedding and plural stories framing. (On the convergence of narrational parataxis and the horizontal plane, see García Landa (1998: 302; for a discussion of multiperspectivity, frame tales and paratextual framing, see Wolf 2000).
Looking at narrative levels from the perspective of vertical and horizontal distribution also opens the way to examining the concept in its various historical contexts. Thus Tomassini (1990: 43–67) stresses the fact that narrative levels, which involve the delegation of a speaker by the narrator, are rooted in Plato’s two modes of narration, pure narration (haplē diēgēsis) and imitation of the heroes’ discourse (mimēsis), and their mixed forms (a connection implicit in Genette and Bal but explicit in e.g. García Landa 1998: 301, and Walsh 2010). Later, Renaissance Italian poetics identified four modes of narration—linear, quasi-linear, oblique, quasi-oblique—two of which correspond to internal narratives. Quasi-linear narrative, in which a character becomes a narrator, is equivalent to metadiegetic narrative, but it also suggests more clearly than the modern concept the effect of reticence on the part of the intradiegetic narrator and that of curiosity on the part of his/her narratee. Quasi-oblique narrative, occurring without change of level (thus horizontally), is close to the rhetorical figure of digressio which, in narrative contexts, qualifies not as an analepsis but is sufficiently autonomous to suspend development of the principal narrative. One form of digression is excursus, where the extradiegetic narrator directly addresses the reader (as in Tristram Shandy), the other being narrative ekphrasis, as in the description of Achilles’ shield in the Iliad, a description which, however, verges on narration. Conceived in this way, digression cuts across modern distinctions such as explanatory metanarrative analepsis (Genette  1988: 93), models that exclude digression from framing and embedding (Williams 1998: 107–08) and extradiegetic narrative comment (Nünning 2004). Clearly, further historical research on these matters is required.
Defining narrative levels in accordance with change of speaker and diegetic level proceeds from structuralist narratology. As the discussion above shows, however, a number of issues remain unresolved. Working from the perspective of artificial intelligence, Ryan (1986, 1991: chap. 9), proposes to rearticulate the question of narrative levels in terms of boundaries, frames and stacks.
To start with, story and discourse are defined not in the sense of the “what” and “how” of narrative but as ontological (semantic) boundaries and illocutionary (speech act) boundaries, respectively. In the simplest case, a single speaker and the utterances are situated at the same level of reality, and there is no crossing of boundaries (1). Boundaries can be crossed in one of three ways, either actually (a) or virtually (b): change of speaker where the speakers are in the same world (2a); speech act of a character presented through that of a narrator (2b); crossing of ontological but not of illocutionary boundaries (change in levels of reality in Alice in Wonderland reported by the primary narrator) (3a); virtual crossing of the ontological boundary but not of the illocutionary boundary (a dream described from an external perspective) (3b); crossing of both boundaries, a fiction within a fiction (4a); actual crossing of the illocutionary boundary but virtual crossing of the illocutionary boundary (primary narrator speaking as though he were a secondary narrator but never entering the world of the projected story) (4b).
Types (4a) and (4b), both of them forms of framing/embedding, can be represented visually as in a picture frame (cf. Ryan 1991: 178). However, modeling these processes in this way fails to distinguish illocutionary from ontological boundaries and also to take into account the chronological sequence of levels in a given narrative. To complement frame structures, Ryan thus introduces the notion of “stacks,” a metaphor drawn by computer science from the “pushing” and “popping” of a stack of cafeteria trays: as trays (or embedded stories) are added or removed, the stack (series of embedding stories) is pushed down or it pops up so that the topmost level remains in view. Where frames provide a static model of the text’s semantic domain and a map of boundaries, stacks are dynamic, capturing moments of that domain and modeling the mechanisms of boundary crossings. Developed independently of Genette’s narrative levels, Ryan’s narrative frames and stacks nevertheless provide a basis for the analysis of metalepsis, as her comments on McHale’s (1987: chap. 8) discussion of strange loops, contamination of levels and other such cases shows (Ryan: 1991: 191–210; 2006).
The original account of narrative levels identified three types of relations between metadiegetic narrative and first-level narration, extending from explanatory (causal relation) through thematic (contrast, analogy) to narrational, independent of metadiegetic content (distraction, obstruction) (Genette  1980: 232–34). This list was later expanded so as to incorporate Barth’s (1981) typology of the frame tale into a six-part “functional” typology whose poles are diegetic content and the narrating act: (a) explanatory (by metadiegetic analepsis); (b) predictive (by metadiegetic prolepsis); (c) purely thematic; (d) persuasive; (e) distractive; (f) obstructive (Genette  1988: 92–4). For Genette, Barth’s interest is in the thematic relations between the two levels. However, it is also true that for Barth stories within stories are a form of digression or postponement of the main story (“Digression and return is a variation on the theme of theme and variation”; 1981: 62), a feature reserved by Genette to the distractive and obstructive functions (cf. Pier 2011: 122–24).
A survey of the literature shows that a broad variety of other functions have been attributed to the two narrative levels as well, and also that a number of revisions of the system have been proposed. Numerous authors writing about frame tales have stressed the authenticating role (a “falsification” function has also been advocated; cf. Tomassini 1990: 175–82), and attention is frequently drawn to the fact that frame tales serve to restore an aura of oral storytelling to written narratives, some non-narratological discussions even restricting the frame tale to the reproduction of oral stories (Jäggi 1994: 62). The expository function of framing has been noted (e.g. Kanzog  1977: 322), and narrative framing as constitutive of “narrative circumstance” has also been defended (Williams 1998: 110ff.).
Closer to narrative levels proper is Shryock’s contention that Genette’s second typology, by adopting a functional perspective, implicitly shifts to a speech act approach. Shryock (1993: 6–8) points out that the explanatory and the predictive functions operate by virtue of their illocutionary force while the persuasive, distractive and obstructive functions can be qualified as such only by their perlocutionary effects (cf. Williams 1998: 101, 106). This suggests that intradiegetic narration, by serving as a “strategy of presentation” of metadiegetic narration, is not, as Genette would have it, “insignificant” ( 1988: 95); at issue, it seems, is the degree of saliency of narrative levels that prevails in specific narratives.
Nelles (1997: 138–49), referring to Genette’s and Barth’s typologies, maintains that embedded narrative is characterized by a dual function: dramatic, as it defers or interrupts the embedding narrative; thematic, by highlighting contrast or analogy. In light of these two functions he outlines an “interpretive strategy” for the study of narrative levels which incorporates Barthes’ hermeneutic, proairetic and formal codes.
The disparities between these and other accounts of the relations between the two levels, intradiegetic and metadiegetic, are due at least in part to the lack of a shared conception of function. Further progress in this area will thus require theoretical reflection on the notion of function in order to clarify its applicability to narrative levels.
One already existent line of inquiry into the functional nature of narrative levels can be traced back to Šklovskij. In an essay devoted to sjužet and the devices of repetition, postponement and digression, Šklovskij ( 1990: chap. 2) shows how these and other techniques, by retarding the development of the principal story, contribute to the deautomatization of perception, or defamiliarization, one of the principal aims of art. It has been noted by Seager (1991: chap. 1) how Todorov ( 1977), who stresses the importance of narrating within narration and thus the fundamental role of metadiegesis (a term not employed by Todorov) in narrative generally, effectively draws attention to the connection between the three functions of metadiegetic narration (explanatory, thematic, narrational) and the perception of retardation, an effect which, according to Šklovskij, is produced particularly by framing devices. Seager also credits Todorov (1979) with pointing out that Genette’s system of extra-, intra- and metadiegetic levels represents a rationalization of Šklovskij’s devices of sjužet composition. But at the same time he comments on the fact that neither of these narratologists follows up on the deautomatization of perception which, in Šklovskij’s poetics, is functionally indissociable from the devices of retardation. This is a point on which Šklovskij and structuralist narratology seem to diverge, but it is also a point that opens up perspectives for present-day narratology to develop a more functional approach to narrative levels.
By formulating embedding and framing in terms of change of speaker and diegetic level, narrative levels have proved fruitful not only in clarifying questions that have remained problematic in the traditional notions, but also for determining premises and raising questions for continued debate. As already mentioned, further consideration is required into narrative levels from a historical perspective, both in practice and in theory (§ 3.2.2), as well as into a functional approach to the concept (§ 3.3).
Because of the connection of narrative levels with framing in narrative, it is a natural step to inquire into the relevance of frames in other disciplines to this narratological category (e.g. Wolf 2006b). This begins with frame analysis (Goffman 1974) and extends, inter alia, to cognitive frames and scripts. To date, cognitive narratology has devoted little research to narrative levels (cf. Herman 2013: chap. 7; Herman → Cognitive Narratology); deictic shift theory (Duchan et al. eds. 1995) and contextual frame theory (Emmott  1999) are topics that open up prospects for further research in this direction.