1. The narrator as a central parameter of narrative perspectivization
Linguistic representations are perspectival in (at least) two aspects. On the one hand, every representation is grounded on segmentation and selection of alternative viewpoints from which something is “seen” and, as such, reflects a “viewing arrangement”, i.e. a “relationship between the conceptualizers and the situation being viewed” (Langacker, 2001: 16; see also Verhagen, 2007, and Sweetser, 2012). On the other hand, language is perspectival since it offers linguistic elements like deictics and evaluative expressions whose interpretation is dependent on an observer’s viewpoint (Mitchell, 1968). Narratives are perspectival in this general sense, but also in a more specific way. They are multiperspectival since they prototypically integrate viewpoints of different characters and narrators and thus offer a set of possible alternate perspectives that allow for viewpoint switches. Furthermore, narratives have traditionally been seen as linguistic artefacts that are mediated by someone telling the story, i.e. as representations of events that are perspectivized as ‘seen’ from a narrator’s point of view. Within the traditional model, the narrator’s perspective thus constitutes a (or even the) central component of narrative perspectivization (see also Szabó, 2015: 124).
As such, the narrator is discussed extensively in literary stylistics and narratology, where the close link between narration and perspectivization has always been a central interest. In linguistics, the study of viewpoint in narratives is also not a new topic (for example, see in particular the grounding work by Kuroda,  2014, and Kuno and Kaburaki, 1977), but has recently become a vital new area of research (see, among others, van Krieken, Sanders and Sweetser (eds.), 2019; Głaz and Trofymczuk, 2020; Lee, 2020; and Altshuler and Maier, in press). Despite this increasing interest, the concept of the narrator itself has not been the main focus of the linguistic discussion (also see Eckardt, 2015), and in studies that investigate how narrator’s and characters’ viewpoints can be shifted and disambiguated (for example, see Kuroda,  2014; Hinterwimmer, 2017; and van Krieken, 2017), the narrator’s perspective is often seen as the default case.
How to define the narrator’s perspective in linguistic terms is, however, not a trivial question. First, it is obvious that the narrator is qualitatively different both from the actual “speaker” of a story as well as from characters within the story (for example, see Kuroda,  2014; and Cohn, 1978). The asymmetry between narrator and character lets us assume that not every viewpoint indicator is valid for both character and (every kind of) narrator. Second, it is also obvious that there is not an overt narrator in every narrative. This has led to the still unsolved question of whether every narrative has a narrator by definition, as prominently discussed in Banfield ( 2015) (and, more recently, in Lee, 2020). In this respect, the narrator’s perspective deviates from an intuitive notion of perspective that is commonly linked to a perceiving subject.
The narrator’s perspective is thus a crucial, but problematic, parameter from both a theoretical and empirical point of view. If we want to explore the possibilities of a unified approach to narrative perspective, as it is the aim of the present special collection, a closer look on basic concepts like the narrator constitutes a possible starting point for such an endeavor. In order to take a (small) step in this direction, this paper addresses the narrator by (1) a review of the linguistic discussion and (2) an empirical analysis on a specific narrative construction in German. After a short overview on the linguistic problems linked to the narrator (Section 2), Section 3 offers an overview on the current state of the art with respect to the theoretical status of the narrator’s perspective against the background of general principles of perspectivization. The discussion leads to the conclusion that the narrator is best conceptualized as a covert viewpoint potential provided by narrative context that can be actualized to different degrees. As such, the narrator’s perspective should not be identified with “plain narration”, but as a perspective that has to be established by means of linguistic clues. In a second step (Section 4), this is demonstrated by an empirical analysis of a specific narrative constellation in German that will show that the narrator’s perspective does not necessarily arise per default. In consequence, it is examined which linguistic cues can disambiguate the narrator’s perspective. The results with respect to the general principles of narrative perspectivization are summarized in Section 5.
2. The narrator as a linguistic problem
In narratology, the “narrator” has often been seen as a (or even the) crucial element of narration. If a narration is defined as the result of “someone telling someone else that something happened” (Herrnstein Smith, 1980: 232), it requires a teller who is committed to the assertions in the text, a “someone” who constitutes the central element within the act of mediacy as well as the deictic center of the story’s representation. This has led to extensive discussions about the theoretical status of the narrator in narratives and the possible actualization within the story as unreliable, homodiegetic vs. heterodiegetic, covert narrators, and so on (for an overview of the controversial discussion of the narrator in narratology, see for example Patron, 2009; Margolin, 2012; and Birke and Köppe (eds.), 2015). In linguistic studies, by contrast, “the narrator” has not been a central theoretical issue. This does not mean that “the narrator” is not part of the descriptions and analyses of narratives, but linguistics has not paid the same attention to the theoretical discussion (see also Eckardt, 2015). In discourse analytical investigations that focus on the sequence of events, coherence phenomena and anaphora in narrative texts, the narrator is rarely mentioned. But also in investigations on narrative perspectivization in a narrow sense, such as studies on Free Indirect Discourse, the narrator is seldom defined. Terms like “narauthor” (Fabricius-Hansen, 2002: 7) seem to indicate that a precise definition is not seen as necessary, since both author and narrator can be seen as forms of “speakers” in terms of speech act theory. The narrator would thus be the agent who tells the story: “the reporter, i.e. the utterer, writer, or otherwise producer of a reporting sentence” (Maier, 2014: 144).
As such, we would expect that the narrator behaves just like any other “speaker” of a linguistic utterance. However, the narrator has been described as behaving in a special way as s/he is the only one that can represent the inner world from a third person perspective (for example, see Hamburger,  1968: 73, and Cohn, 1978) and thus possesses knowledge that is not available for an ordinary speaker in the everyday world. This has grammatical consequences in languages such as Japanese, where there is a grammatical opposition between information that belongs to the territory of the speaker and information that does not. Kuroda (, ,  2014), for example, has shown that some grammatical forms are used within narration only, so that “one would be led to conclude that the omniscient narrator uses a special grammar of his own” (Kuroda,  2014: 81).
Furthermore, the narrator can be actualized in different forms (for example, see Kuroda,  2014, on the “omniscient narrator”; and Bimpikou, 2020, on the different prominence status of first vs. third person narrator). Also, not every story has a narrator as an “embodied self” (Stanzel, 2008: 127ff.) who is visible within a story, which has led to the question of whether every narrative has a narrator by definition, as discussed in Banfield,  2015. The narrator thus also behaves differently to the protagonists within the story.
As the narrator is both different from the “speaker” as the agent who tells the story and the protagonists within the story, the status of the narrator is not a trivial question and calls for a specification in linguistic terms. Such a specification is closely linked to perspectivization in general, as the narrator – as will be seen in the following – is shaped by an intuitive notion of perspective but also deviates in central aspects. It is thus the relationship between the narrator and the conception of perspective in general which will constitute the starting point for the discussion.
3. The narrator’s perspective in linguistic terms
In order to specify the narrator’s perspective, we have to specify what a narrative perspective is. At first sight, perspective can be seen as a directed relation between a perceiving subject and the perceived aspects of an object in focus, the latter being dependent on the observer’s viewpoint, and hence a binary relation between “subject” and “object” (see, among others, Linell, 2002: 43). Such an intuitive definition derives from the concept of perceptual perspective which is bound to (at least) three premises. First, although perspective always presupposes the existence of possible alternative perspectives, an observer in real life is, due to his local-temporal position, commonly restricted to one viewpoint only. If one is standing before a statue, one is able to see only its front, and not its back. Changing the perspective is possible but this requires a change of the localization between the observer and the observed object. As such, perspectivization has often been described in terms of “viewpoint-shift” and “perspective-taking”, relying on the premise that a choice of perspective is linked to the selection of one perspective out of several others, and, as such, is a choice of either – or. Second, this selected perspective is commonly seen as person-bound since it presupposes an animate subject that is able to perceive and observe. Linked to this, it is also a common premise that perspectivization always requires an origin of perspective:
|(1)||Premises of perceptual perspective|
|P-1:||Perspectivization is a form of perspective-taking.|
|P-2:||Perspectivization presupposes an animate subject.|
|P-3:||Perspectivization requires an origin of perspective.|
In symbolic systems such as language, perspectivization deviates from the premises of perceptual perspective as we are mentally capable of processing and evaluating more than one perspective at a time (see Zeman, 2017, for more details on the relationship between linguistic and cognitive perspectivization). Narratives, in particular, have been described as multiperspectival since they rely on multiple viewpoints which can be situated on the level of the narrated world or the level of the (real or fictive) act of narration (see recent work by Lee, 2020: 47). Furthermore, multiperspectivity is also an inherent feature of the linguistic structure of narratives (Zeman, 2017; see also Lee, 2020: 47). While in everyday discourse the context of utterance, thought and perception are usually the same, the structure of narrative relies on a potential difference between the author’s and the narrator’s viewpoint. As such, narration not only offers a bigger potential for perspectivization than everyday discourse but also allows for the simultaneous activation of more than one perspective (for example, see Bakhtin’s 1981 studies on polyphony; Dancygier, 2012, on “viewpoint blending”; and Zeman, 2017, on viewpoint embedding).
Nevertheless, the conceptualization of the narrator’s perspective has also been shaped by the premises in (1). Both narratological and linguistic accounts have been looking for linguistic markers that indicate whether the viewpoint in a story can be attributed to the “speaking” narrator or the “perceiving” characters. As such, narrative perspectivization has often been seen as a form of perspective-taking, as a choice between either the narrator’s or protagonist’s viewpoints. Terms like “narrator” and “protagonist” furthermore suggest that narrative perspectives presuppose a “Subject of Consciousness” (i.e. a referent “to whom all expressive elements are attributed”; see Banfield, 1973: 29, and Banfield, 1982: 93),1 since only animate beings are capable of speaking and perceiving. It would thus not be reasonable to refer to the perspective of an object like a stone, unless it is anthropomorphized as a conscious observer within a story (see Richardson, 2006, for “unnatural” narratives, and Trompenaars et al., 2018, on the linguistic animacy features of “inanimate narrators”). Furthermore, many accounts rely on the assumption that every story presupposes an act of narration and thus a “speaker” or “narrator”. The premises of perspectivization thus correspond to the theoretical questions about the narrator’s perspective, as outlined in Section 2 above:
|(2)||Q-1:||Is the narrator’s perspective a form of perspective-taking?|
|Q-2:||Does the narrator’s perspective presuppose a “Subject of Consciousness”?|
|Q-3:||Does every narration have a “narrator”?|
As an intuitive notion of perceptual perspective is not capable of grasping the characteristics of narrative perspectivization; the following section aims to discuss these questions with respect to narrative perspectivization in general and the narrator’s perspective in particular. If the narrator is “a grammatical problem” (Kuroda,  2014: 62), it seems promising to have a closer look at grammatical constructions that are characteristic for narrative perspectivization in order to examine the premises of narrative perspectivization and the specific characteristics of the narrator’s perspective, in particular. As such, as a reference point for the discussion, I take a grammatical construction which leads to the disappearance of the narrator, i.e. Free Indirect Discourse.
3.1 Is the narrator’s perspective a form of perspective-taking?
If perspectivization is seen as the result of a selection based on several choices with respect to the representation of the story, narrative perspectivization can be defined as the whole complex of factors that determine the representation of the events within a story (Schmid, 2014: 121). In a narrower sense, perspectivization becomes particularly relevant when the represented consciousness does not coincide with the speaker of the utterance and there is more than one conceptualizer to whom judgements and responsibility can be ascribed. In this respect, Sanders and Spooren (1997) have drawn a distinction between “subjectivity” and “perspective”. While subjectification is defined as “the explicit binding to the subject (i.e., the current speaker)”, perspectivization is “the binding to a subject other than the speaker” (Sanders and Spooren, 1997: 93). Under this definition, narratives are particularly interesting objects of investigation as they integrate viewpoints of different characters and narrators as alternative viewpoints besides “the speaker” and offer a potential for “viewpoint shifts”. In consequence, one central aspect of linguistic studies on narrative perspectivization has been the question of which factors indicate whether a proposition can be ascribed to the narrator’s or to a character’s viewpoint (for examples, see Hinterwimmer, 2017, and van Krieken et al., 2017).
What at first sight looks like perspective-taking (“Is it the perspective of the narrator or the protagonist?”) is, however, more than just choosing one perspective out of many others. On the one hand, several investigations (for example, see Sanders et al., 2012; Sweetser, 2012; Dancygier, 2012; van Duijn and Verhagen, 2019: 220f.; Dancygier and Vandelanotte, 2016; Evans, 2005; Spronck, 2012; van Krieken, 2017; Zeman, 2017) have shown that narratives are commonly characterized by multiple viewpoints which can, but must not be, equivalent alternatives, since the viewpoints of characters and narrators can be embedded in each other. In this respect, Sanders et al. (2012) assume a “recursive patterning which can represent not only the author’s subjectivity, but also the subjectivity of embedded subjects, such as speakers and actors in a narration” (Sanders et al., 2012: 209). Furthermore, these different viewpoints require a global perspective from which the different perspectives are organized. Based on these considerations, it has been proposed to investigate the relations between viewpoints in terms of mental space theory as “network constellations” (Dancygier and Vandelanotte, 2016) or “thoughtscapes” (van Duijn and Verhagen, 2019). In these accounts, viewpoints are seen as hierarchically ordered mental spaces and can be linked to embedded, parallel, alternative or independent spaces. Narrative perspectivization can thus not be adequately described as “perspective-taking”. Rather, the description of such phenomena requires that the whole perspectival constellation is taken into account, and, as such, the relationships between the different viewpoints.
What does this mean for the narrator’s perspective? In order to approach this question, it is instructive to take a closer look at the multiperspectival constellation of Free Indirect Discourse (FID). FID is a phenomenon that is restricted to narrative contexts only and has been the main focus in many studies due to its striking combination of past tense and third person pronouns and deictic adverbials linked to the “here-and-now” of a protagonist. For example, see (3) below:
|(3)||Tomorrow was Monday, Monday, the beginning of another school week!|
|[Lawrence, Women in Love; cited in Banfield  2015: 98.]|
Sentences like (3) have been characterized both as “univocal” and “bivocal” (see Vandelanotte, 2009: 246–251, for an overview). “Univocal” accounts foreground the criterion that FID offers access to mental contents of the protagonists which are normally only accessed by introspection and linked to one (i.e. the character’s) perspective only (e.g. Banfield, 1982). “Dual-voice” accounts in the tradition of Pascal, 1977, on the other hand, foreground the criterion that FID integrates two perspectives, namely the character’s and the narrator’s perspective at the same time (for example, see Fludernik, 1993). Eckardt (2014), for example, describes FID as a “double context” (C,c) in the sense that FID has to be evaluated relative to two utterance contexts <C,c>, i.e. the “external” context C of the narrator vs. the “internal” context c of the character. For Schlenker (2004), the difference is correlated with the distinction between the Context of Thought (CoT, θ; i.e. “the point at which a thought originates”) vs. the Context of Utterance (CoU, υ; i.e. “the point at which the thought is expressed”) (Schlenker, 2004: 279).
One could argue that FID is an instance of perspective-taking in the sense that the perspective either of the narrator or the character is “taken”. On the other hand, FID is more than just perspective-taking, since the viewpoints involved are not equivalent viewpoints but are situated on different levels of discourse. This is indicated by the terms by Eckardt (2014), who distinguishes between an “external” and an “internal” context. “Internal”, however, can mean two different things: it can refer to the “inner” thoughts of the protagonist that are being reported (“they are inside his mind”), and it can mean that these inner thoughts are “internal” since they are situated inside the diegetic world whereas the heterodiegetic narrator is situated “outside”. In the latter sense, “internal” vs. “external” refers to the hierarchical viewpoint constellation between the narrator’s and the characters’ discourse which constitutes the structure of narratives (for a narratological perspective see Margolin 1991: 519, for example). This hierarchical structure is crucial for narratives, since it implies different access hierarchies: due to its position outside the described situation, the perspective of the narrator allows for simultaneous knowledge of the course of the story which can include the inner thoughts of the protagonists and its representation in discourse. The narrator “knows” what the protagonist knows, while the reverse would lead to a metaleptic rupture. This allows for the embedding of “internal” in “external” perspectives; see example (4), discussed in Phelan (2005):
|(4)||And I catch myself thinking today that our long journey had only defiled with sinuous trail of slime the lovely, trustful, dreamy, enormous country that by then, in retrospect, was no more to us than a collection of dog-eared maps, ruined tour books, old tires, and her sobs in the night – every night, every night – the moment I feigned sleep.|
|[Vladimir Nabokov, 1955, Lolita, Penguin II/3, p. 199.]|
In (4), I catch myself thinking today refers to the narrator-I who offers an evaluative view on the past events (i.e. an “external” perspective from outside the diegetic world). With by then, in retrospect a temporal shift is introduced. While the list of items describes the perception of the character on the story level as a “psychonarration” presented from “outside”, the parenthesis (– every night, every night”) gives insight within the protagonist’s earlier thoughts (i.e. an “internal” view that represents the mental content, but embedded in the report of a narrator). In Phelan’s words: the “narrator’s focalization contains the character’s” (Phelan, 2005: 118; see also van Krieken (2017: 14) for the observation “that the telling (by the narrator) and the seeing (by the character) can fuse to a greater or lesser extent, such that an observation can be made by both the narrator and the character at the same time”). The double-layered structure thus provides not only a viewpoint potential as the basis for a choice of perspectives but also allows for representing more perspectives at one time, and as such, more than “perspective-taking”.
3.2 Does narrative perspectivization presuppose a “Subject of Consciousness”?
A perspective establishes a relation that is linked to a certain vantage point. Within an intuitive conception of perspective, this vantage point is constituted by a conscious subject: the act of “perceiving” and “perspective-taking” requires that the subject is both animate and aware of his perceiving acts. In this sense, perspectivity is closely connected to the concept of “subjectivity”, i.e. “the presence of this character’s active consciousness to which the mental state is attributed (see Sanders & Redeker 1996)” (Sanders and Spooren, 1997: 85). According to Verhagen (2007), the common denominator of the notions of perspective is that “they capture aspects of conceptualization that cannot be sufficiently analyzed in terms of properties of the object of conceptualization, but, in one way or another, necessarily involve a subject of conceptualization” (Verhagen, 2007: 48; see also Smith, 2002). Perspectivization is thus more than a link between some kind of vantage point and an object but is seen as a linguistic process that involves “an implicit locus of consciousness” (Langacker, 2008: 77). In a narrow sense of the term, the description of intrinsic localizations, such as in The cat is sitting in front of the house for example, would thus not be perspectival, as it does not necessarily presuppose an observer. (In a wide sense, it can be seen as perspectival since it represents a conceptualization of the speaker.)
This premise is reflected in definitions of narrative perspective as “the embedding of a subject’s point of view in the narrator’s discourse reality” (Sanders and Redeker, 1996: 291; emphasis in original). Similarly, Dancygier and Vandelanotte (2016: 14) see perspectivization as “a discourse participant’s alignment with an aspect of a frame or situation”. “Perspective” and “viewpoint” are thus conceptualized as person-bound categories, relying on the assumption that every act of perspectivization presupposes a perspectivizing subject.
On the other hand, viewpoints are not linked to homogenous subjects as terms like “viewpoint shift” and “perspective-taking” could suggest. In tradition of Benveniste (1974), Ducrot (1984), and Jakobson (1957), it has been frequently stated that viewpoints do not refer to “real” physical persons in the extralinguistic world but to functional roles within discourse. Virtually, we are, for example, capable of splitting ourselves into an observing subject and an observed object if talking about ourselves (see Dancygier, 2004; Zeman, 2018). This principle is most obvious in first person narratives and soliloquies where the narrator and the protagonist are the same “person” but refer to two different instances within the text, as seen in the fact that narrator-I and character-I usually hold different states of knowledge (see also Virdee, 2019: 423). The “subject” is thus not a unified person, as also stated by Banfield:
“in first-person narration the I is divided by time into a SELF caught always in the NOW of consciousness and a SPEAKER narrating in a moment for which the NOW of consciousness is always past.” (Banfield,  2015: 195)
From a linguistic perspective, the narrator can thus not be seen as a homogenous narrating persona (although that might be the case from the reader’s perception).
If we look again at FID, it becomes apparent that its inconsistent description is due to the fact that different dimensions of the narrator have been seen as the defining features for “the narrator”. In “univocal” accounts, it has been stated that FID leads to the impression that the viewpoint of the narrator is effaced and represents only the perspective of the protagonist on the story level, without any intrusion by the narrator. The latter position has famously been defended by Banfield ( 2015). One of her main arguments is that FID cannot integrate statements from the narrator, as seen in the fact that represented thoughts cannot be pragmatically cancelled by contrasting perspectives. For example, see (5):
|(5)||#No, she would never lie, although she often did.|
According to Banfield ( 2015: 215), the incompatibility of the contrasting statements in (5) shows that FID represents only the mental contents of the character. This makes it clear that Banfield ( 2015) is talking about a viewpoint bound to a consciousness of a (real or fictive) personal subject. In approaches that describe FID as “bivocal”, in contrast, the “narrator” is not necessarily a conscious subject but rather conceptualized as an abstract deictic vantage point outside the narrated world that is indicated by third person pronouns and the past tense. In this sense, FID integrates two perspectives, but not necessarily two Subjects of Consciousness. Only the character’s viewpoint is linked to a Subject of Consciousness, whereas the narrator’s mental state is not at issue. It is thus necessary to distinguish between two different parameters of perspectivization, i.e. the mental content of a protagonist’s or narrator’s “inner world” and their structural position within the architecture of the narrative, or, in Sanders and Spooren’s (1997) terms, between a “referential center”, i.e. a vantage point that can be qualified spatially and temporally, and the “subject of consciousness” (see also Vandelanotte, 2019: 172 who distinguishes between a “deictic” perspective, i.e. the relation to a deictic center, leading on from Bühler (1934), from which the situation is construed, and a “cognitive” perspective which pertains to experiences like thoughts, beliefs, feelings, attitudes and perceptions).
With respect to the narrator’s perspective, FID shows that a cognitive dimension is not necessarily present. This has important implications for the question about the theoretical status of the narrator, since it has led to more abstract concepts of the narrator. For Dancygier (2012: 61), for example, arguing within the Mental Space Framework, it is not the narrator per se, but a viewpoint space which constitutes the primary reference point for the overall structure of the narrative. This basic deictic ground is the location where the narrator can be profiled or not. The narrator is thus a “‘disembodied’ construct” (Dancygier, 2012: 136) which “does not necessarily rely on a viewing subject, so that it is not necessarily “someone’s” point of view, but it relies crucially on a selection of an aspect of space topology which provides a filter through which the events are narrated” (Dancygier, 2012: 61). Van Krieken et al. (2016) introduce a “virtual observer” as a “derivative of the narrator projected into the narrative” (van Krieken et al., 2016: 149). The narrator’s “personhood” can thus be reduced to zero (Lee, 2020: 47). Such an abstract conception of “narratorship” presupposes neither a speaking person nor a Subject of Consciousness (see also Gƚaz and Trofymczuk, 2020).
3.3 Does every narration have a “narrator”?
As seen above, narrative perspectivization has traditionally been investigated as a question of either – or: as someone must be the origin of perspective, the narrative content can either be attributed to the viewpoint of a protagonist or the narrator, as also reflected in Genette’s ( 2007) distinction between “who speaks?” vs. “who perceives”. This dichotomic distinction implies the “default-hypothesis”: if there is no indicator that marks a viewpoint shift to the character, the narrator’s perspective is the default case. For example, see Banfield ( 2015):
“But the missing premise is none other than the conclusion: if it doesn’t represent the character’s, it must represent the narrator’s voice.” (Banfield,  2015: 189)
Banfield has famously argued that there is a third option: sentences of narration in a narrow sense such as Free Indirect Discourse, which (although they can contain a “speaker”) erase the communicative frame. As such, “pure narration” is “descriptive language which is in some sense disembodied”, “where no first person need intervene” and which does not have any addressee (Banfield,  2015: 178). In this sense, sentences of narration are “unspeakable” and “outside any framework structured by the communicative relation between I and you” (Banfield,  2015). Likewise, Hamburger ( 1968) has argued that (third person) narrative fiction is not a form of a “statement about reality” (“Wirklichkeitsaussage”) and therefore categorically different from discourse. A comparable position is also found in Kuroda ([1973, 1974, 1979] 2014) who has argued that sentences of narration do not serve a communicative, but rather an objective function of language that cannot be accounted for in any framework of speech act theory. Similarly, Benveniste (1974: 269) has defined narration as a sequence of events that speak for themselves without any traces of an act of narrative mediacy. As a result, a narrator no longer exists: nobody “speaks”.2 Such conceptions argue against “the almost universal assumption that a narrative simply has to have a narrator” (Spearing, 2005: 17), since every speech act presupposes a “speaker” (see Köppe and Stüring, 2015, for an overview on so-called “pan-narrator-approaches” and the on-going dispute in narratology as to whether every narration implies an (overt or covert) narrator).
The discussion shows that linguistic conceptions of the narrator are not consistent but are referring to different perspectival dimensions of the narrator. For Dancygier (2012), who defines the narrator in terms of an abstract “narratorship”, there is always a deictic ground which serves as a primary reference point for the story, “but the degree to which such a vantage point is associated with a narrating subject or not accounts for the various choices made” (Dancygier, 2012: 64). She thus rejects Banfield’s 1982 concept of an “empty” deictic center as she assumes that “every narrative act assumes a space wherein there is a source of viewpoint, which can then take the form of a narrator” (Dancygier, 2012: 207, Fn. 5). In a similar way but using a different framework, Eckardt (2015: 154) has argued that the narrator can be seen as a reader’s impression that is indicated by linguistic elements, but that in “the extreme case, the fictional narrator […] can be completely absent”. She assumes that a formal speaker parameter is in play in every narration, but that the narrative information is sometimes “insufficient to create the fiction of a narrator” (Eckardt, 2015: 183). Since this “effaced” narrator is semantically presupposed, it is, as shown by Altshuler and Maier (in press), always possible to accommodate a discourse referent “for the (already implicitly present) narrator”.
On the background of the observations so far, the appearance of the narrator’s perspective can thus be described as the result of an activation of a covert viewpoint potential as provided by narrative context (Zeman, 2017): the perspectival structure allows for the gradual actualization of (different dimensions of) a narrating persona, but does not necessarily lead to the construction of a fictional narrator. Such a conceptualization of the narrator has important implications for the analysis of viewpoint attribution: In order to analyze the narrator’s perspective in a narrative text, we would have to ask which perspectival dimensions are actualized and which linguistic cues facilitate the actualization of the narrator’s perspectival dimensions. Both questions will be pursued in the following section, which will examine an empirical case of the narrator’s perspective.
4. The narrator’s perspective from an empirical point of view
As already said above, investigations on viewpoint shifts have generally focused on linguistic indicators that mark a switch from the narrator’s to a character’s perspective, whereas the narrator’s perspective has been seen as the default case. According to Sanders et al. (2012: 209), for example, “the accessibility of the speaker’s epistemic space is less complicated than that of a narrative character’s SoC. The epistemic space of a third person character does not come “for free”, as is the case with the speaker’s epistemic space, which is always there; it requires evocation from the context” (Sanders et al., 2012: 209; see also Kaiser, 2015). In contrast, a character’s perspective requires activation by viewpoint markers that indicate the viewpoint of an experiencer who is not the speaker of the story, but a locally prominent protagonist (Hinterwimmer, 2017; Bimpikou, 2020). If such a viewpoint marker is introduced, readers continue to attribute the narrative information to the protagonist’s viewpoint. If a viewpoint marker is missing, narrative information is attributed either to the narrator, or the participants “refrain from clearly attributing observations to either the narrator or the character” (van Krieken, 2017: 13).
In the following, I want to approach viewpoint attribution from an opposite angle and ask which linguistic cues facilitate the actualization of the narrator’s perspectival dimensions. In order to examine this question, I will have a look on a specific proleptic structure in German that gives an outlook on the events to come in the future of the storyline with a high degree of certainty in the sense of a “Future of Fate” (FoF) and thus presupposes the viewpoint of a narrator, i.e. the German modal verb construction sollte (“should”) + infinitive, see (6):3
|(6)||Wenige Stunden später sollte sie eine böse Überraschung erwarten.|
|“A couple of hours later, a nasty surprise was to (literally: should) await her.”|
|[Welt.de webpage, available at https://www.welt.de/vermischtes/article166697218/Dieb-klaut-Fahrrad-Besitzerin-stiehlt-es-sich-zurueck.html, last accessed 13 November 2020.]|
Similar to FID, the meaning of the construction in (6) affects both the deictic and cognitive dimension of narrative perspectivization. It inherently presupposes two perspectives, linked to two states of knowledge. The narrator “knows” what will happen in the future of the protagonist, while the protagonist herself is unaware. However, the beliefs and thoughts of the protagonist are not at issue. Instead – unlike in FID – the narrator’s viewpoint is foregrounded. With respect to their perspectival structure, FID and FoF can thus be seen as complementary constellations. This is also supported by the fact that FoF cannot occur in contexts of FID and vice versa (Zeman, 2019a; Zeman, 2019b). While investigations of FID have led us to the definition of linguistic viewpoint markers that indicate the perspective of the character, FoF should thus lead us to viewpoint markers that indicate the perspective of the narrator.
The following observations rely on a corpus study of 6,000 occurrences of sollte + inf., including 338 instances with FoF reading (Zeman, 2019a).4 The analysis revealed that many occurrences of sollte + inf. that would allow for a FoF reading are actually ambiguous when they are examined in isolation from their context, since the modal construction is polyfunctional and can display also deontic and epistemic modal meanings, and, as such, can be attributed to a character’s viewpoint. The instances could thus not be classified without a closer analysis of the context beyond the sentence; see (7):
|(7)||Frank sollte dieses Problem lösen.|
|Literal meaning: “Frank should solve this problem.”|
|a.||It was Frank’s task to solve this problem.|
|→ obligation/necessity; event realization = uncertain|
|b.||Frank was to solve this problem.|
|→ Future of Fate; event realization = certain|
As one would expect, FoF readings frequently combine with kataphoric adverbials such as later on. However, these temporal markers are not able to unambiguously indicate the FoF-interpretation. Also prototypical viewpoint indicators are not sufficient to elicit the narrator’s perspective, since they can be attributed to both the narrator and the protagonist. See (8–9) (discussed in Zeman, 2019b: 238):
|(8)||Ich wußte es besser – und klagte doch auch mit ihm.|
|Aber ich sollte sie wiedersehen!|
|Es war ein stiller Abend. […]|
|[Ludwig Schefer, 1845. Der Kuss des Engels. Verlag von Veit & Comp.]|
|“I knew better – and still lamented with him. But I was to see her again!|
|It was a silent evening. […]”|
|(9)||Da plötzlich erschrak ich. Wahrhafftig – wie fiel mir das erst jetzt ein – ich sollte sie wiedersehen, aber wo und wann? Sie konnte unmöglich wissen, wo ich abgestiegen war – und ich hatte mich ja bereits vergeblich nach ihr erkundigt. Stumm und ungeschickt hatte ich dagestanden, und nun, – war sie mir nicht fast so gut wie abermals verloren?|
|[Ludwig Salomon, 1908. Die Blüchertrompete. B. Elischer.]|
|“Suddenly, I was frightened. Indeed – how did this cross my mind only now – I should see her again, but where and when? She could not possibly know where I was staying – and I had already asked after her in vain. Silently and clumsily I had stood there, and now, – wasn’t she practically lost to me again?”|
In (9), sollte + inf. is used in a context of FID, which is indicated by deliberative questions (aber wo und wann?), evaluative expressions (vergeblich), modal particles (ja) and pauses for reflection, indicated by the dashes in the text, and does not allow for a FoF reading. On the other hand, (8) displays a FoF reading, despite speaker-oriented features like first person pronouns and exclamative expressions. Formal indicators are thus not sufficient in order to distinguish between both readings. Furthermore, both (8) and (9) are attributed to first person narrators. The difference lies only within the viewpoint position: the viewpoint of the narrator-I in (8) is situated outside the story world which allows for a global view on the whole story world, while the narrator-as-character-I in (9) is restricted to the knowledge present at the “story now”. This “external” perspective can be elicited by contextual means which indicate that the narrator “knows” more than what can be known at the actual “story now” from the protagonist’s perspective. This is also the case in third-person-narratives; see (10):
|(10)||Da wusste Carola Wichmann noch nicht, dass alles bald viel schlimmer kommen sollte.|
|“At this time, little did Carola Wichmann know that soon everything was to become even worse.”|
|[DWDS-Korpus – Die ZEIT, 4 January 2018, Nr. 01]|
In (10), the (third person) narrator’s perspective is elicited by several features: an event is announced that is – when seen from the perspective of the character – situated in the future. Its later actualization is presented as certain, and, as such, presupposes the knowledge about the future sequence of events. Furthermore, an evaluation of the future situation as “even worse” presupposes an observer’s viewpoint that is situated outside the diegetic world, linked to an asymmetry with respect to the states of knowledge as made explicit in the first part of the sentence (“little did she know”).
Within the corpus, the following factors have been found that facilitate the FoF readings:
|(11)||Indicators that facilitate the FoF reading|
|(i)||Metaperspectival expressions such as evaluative content words (e.g. scandal, later generation), specific details about the future events (e.g. as the Italian newspaper ‘La Stampa’ was to celebrate), temporal adverbials that refer to a longer time span in the future and can include the discourse now (her whole life, until the present day).5|
|(ii)||An unexpected contrast between the future events and the story now (e.g. But this was to change very soon) which presupposes the narrator’s knowledge of the further events of the story.|
|(iii)||An asymmetry between the knowledge states of the narrator and the character, often leading to dramatic irony: the narrator knows what is going to happen whereas the protagonist is both unaware and unable to control the events. Within the corpus, this correlates with the amount of impersonal, non-agentive subjects (28% non-animate subjects; see Zeman, 2019a: 243).6|
The context factors in (11) facilitate a global perspective on the future events and thus elicit a narrator’s viewpoint outside the diegetic world. In addition, the FoF reading can be elicited by metanarrative comments which make clear that the character has no control over the events to come and rules out the protagonist’s viewpoint. The disambiguation is most successful if there are indicators of the narrator’s metaperspective and, at the same time, indicators of the suppression of the character’s viewpoint, which suggests that viewpoint attribution is based on the processing of both the narrator’s and characters’ viewpoints.
Furthermore, the analysis of the FoF supports the above assumption that character and narrator viewpoints are not equivalent alternatives but are situated on different levels of the hierarchically ordered discourse structure, as seen in the fact that the narrator’s knowledge includes both the knowledge state of the characters and the knowledge about the further line of discourse. This hierarchy allows for the combination of representing a future event with a high degree of certainty with respect to its actualization, which is uncommon in non-narrative discourse. The FoF can thus be seen as a grammatical structure that reflects the hierarchical constellation of narration on the micro-linguistic level (see also Currie, 2013: 175, who sees the future-of-the-preterite as a model of narration per se) and, as such, confirms the relevance of the hierarchically-ordered discourse structure for the interpretation of grammatical means.
So far, we have thus seen that the FoF construction implies a narrator’s perspective. But what kind of narrator is elicited? In order to address the question which perspectival dimensions of the narrator are activated and whether the narrator’s perspective induced by FoF readings presupposes a Subject of Consciousness, it is necessary to have a closer look at the “narrator’s knowledge”. At first sight, FoF readings seem to require the perspective of an “omniscient” narrator, for whom the characters are “transparent minds” (Cohn, 1978): since the narrator is “outside” the diegetic world, he “knows” the beliefs of the characters in the story. In this respect, narratological classifications of perspectivization have traditionally distinguished between “zero focalization” (“The narrator knows more than the character knows”), “internal focalization” (“The narrator tells what the character knows”) and “external focalization” (“The narrator tells less than the character knows”) (Genette,  2007).7 At a first glance, FoF seems to be a prototypical example of zero focalization. However, what the narrator knows about the character’s knowledge is not the important aspect for the FoF readings. As the occurrences in the corpus show, the knowledge of the narrator primarily concerns events which can involve future thoughts and perceptions of the protagonist, but not necessarily do so. In most cases, the proleptic reading is independent of the cognitive states of the protagonist, as also seen in the fact that an external focalization is possible as well. For example, see (12), which is taken from a historical chronicle where the inner thoughts and beliefs of the protagonists are unknown to the narrator:
|(12)||In der ersten Zeit nach der Einnahme von Nan-kiṅ scheint der Tien-waṅ noch für Ausbreitung seiner Lehre gewirkt zu haben; 1853 wurde dort an der Bibelübersetzung gedruckt, die Gemeingut werden sollte Bald darauf muss sein Geist sich verwirrt haben;|
|“In the first time after the capture of Nan-kiṅ, Tien-waṅ seems to have still worked for the propagation of his doctrine; in 1853, the translation of the bible was being in press there, which was to become common property. Soon after that, his mind must have been confused;”|
|[Albert Berg, DTA 227, Die preussische Expedition nach Ost-Asien, 1873. Vol. 3. Berlin.]|
Thus FoF readings do not presuppose the narrator’s “omniscience”, as his knowledge can also be fragmentary. Furthermore, the FoF can also occur in contexts of internal focalization. In autobiographies, where narrator and character are referentially the same, the narrator knows his own inner thoughts within the context of the story world and the future events to come, but he does not know about the inner world of other people (see also the examples of first person narratives in (8–9)). The knowledge about the mental world of the protagonist is thus not the crucial parameter at stake. Rather, the certainty about the events to come is the relevant factor for the FoF readings. This knowledge does not necessarily come from an “inner” mental state linked to a (real or fictive) person but rather follows from the narrator’s structural position within the architecture of the narrative that allows an external view on the discourse. As such, this knowledge is not the result of perception or linked to an evaluative or epistemic stance, and is thus not necessarily linked to the Subject of Consciousness of a “narrating persona”.
This supports the assumption that the narrator’s perspective does not come for free but that the narrator’s perspectival dimensions can be actualized to different degrees. With respect to viewpoint attribution in narrative discourse, it is thus crucial not only to examine the indicators of the character, but also the contextual factors that can facilitate the different dimensions of the narrator’s perspective (see also Bimpikou, 2020).
In sum, the theoretical discussion and the empirical analysis have led to the following conclusions with respect to the narrator’s perspective:
- The narrator and characters do not represent equivalent viewpoints but viewpoints with different qualities. Due to the hierarchical structure of narrations, the narrator has access to different “internal” contents, i.e. the knowledge on the diegetic world, including the inner thoughts of the protagonist and a metanarrative knowledge about the story representation. This “knowledge” is not necessarily linked to a Subject of Consciousness. The “narrator” is thus not a “homogenous person” but a textual instance that can be described according to different perspectival dimensions.
- The narrator’s perspectival dimensions do not necessarily arise by default, but require contextual indicators in order to be distinguished from “plain narration”. While markers of subjectivity can indicate both the narrator’s and characters’ viewpoints, the narrator is unambiguously indicated by contextual means that presuppose a global perspective on the narration as a whole. The narrator can thus be seen as a covert viewpoint potential provided by narrative context that can be activated to different degrees.
If we compare these results with the intuitive notion of perspective (see Section 3 above), we can conclude that the narrator’s perspective deviates from the premises of perceptual perspective in relevant aspects. Since narrative viewpoints are linguistic constructions, the “narrator” does not necessarily rely on a narrating persona. Rather, there are several perspectival dimensions of the narrator which can be activated to different degrees. As such, the “narrator” can be reduced to an abstract deictic vantage point and loose its “communicative voice”, so that the perspectival relation between subject and object can be effaced (see also Gƚaz and Trofymczuk, 2020, on the conceptual difference between “voice” and “viewpoint”). It is important to note that this abstraction is not restricted to literary fiction, as seen in the fact that the “abstract” narrator elicited by the FoF readings is frequently attested in narrative passages of journalistic reports, sports coverage and popular science.
On the other hand, narrative fiction also allows that inanimate narrators become animate by language (see Richardson, 2006, on “unnatural” narratives). See example (13) below where the narrator is an ancient Sumerian bowl, which is linguistically represented as a self-reflective experiencer with a cognitive perspective:
|(13)||She puts me down on a low table, folds her arms, and looks down at me sternly.|
|‘Talk,’ she commands.|
|This is an idiotic, if not deranged thing to say to a bowl, even to a bowl like me, thin-walled, sporting the scorpion look of Samarra ware that was the rage of Mesopotamia six and a half thousand years before Rosa was born. Pottery, after all, isn’t renowned for its chatty nature, so why futilely address a vessel thus – even me, the bowl with soul? But Rosa is far from being unhinged.|
|Inevitably, I’ve been talked to, more than anyone would credit. Being inanimate doesn’t earn you any dispensation form being buttonholed.|
|[Tibor Fischer, 1997. The collector collector. New York: Metropolitan books, p. 5f.]|
Trompenaars et al. (2018) have shown that if (homodiegetic) narrators are “not a prototypical animate entity, lacking agentive properties”, their animacy is nevertheless presupposed by the use of pronouns, grammatical functions and voice (Trompenaars et al., 2018: 719). Conceptual inanimate narrators are thus treated grammatically as animate “tellers”. The “fictionalization” of the narrator can thus go in both ways: the “teller” of a story can be effaced and “lose” its consciousness – or an inanimate narrator can gain consciousness through language. In both cases, it is vital to gain more insights within the linguistic aspects of the narrator’s perspective.