VPIP: A Lexical Identification Procedure for Perceptual, Cognitive, and Emotional Viewpoint in Narrative Discourse

Although previous work on viewpoint techniques has shown that viewpoint is ubiquitous in narrative discourse, approaches to identify and analyze the linguistic manifestations of viewpoint are currently scattered over different disciplines and dominated by qualitative methods. This article presents the ViewPoint Identification Procedure (VPIP), the first systematic method for the lexical identification of markers of perceptual, cognitive and emotional viewpoint in narrative discourse. Use of this step-wise procedure is facilitated by a large appendix of Dutch viewpoint markers. After the introduction of the procedure and discussion of some special cases, we demonstrate its application by discussing three types of narrative excerpts: a literary narrative, a news narrative, and an oral narrative. Applying the identification procedure to the full news narrative, we show that the VPIP can be reliably used to detect viewpoint markers in long stretches of narrative discourse. As such, the systematic identification of viewpoint has the potential to benefit both established viewpoint scholars and researchers from other fields interested in the analytical and experimental study of narrative and viewpoint. Such experimental studies could complement qualitative studies, ultimately advancing our theoretical understanding of the relation between the linguistic presentation and cognitive processing of viewpoint. Suggestions for elaboration of the VPIP, particularly in the realm of pragmatic viewpoint marking, are formulated in the final part of the paper.

Although previous work on viewpoint techniques has shown that viewpoint is ubiquitous in narrative discourse, approaches to identify and analyze the linguistic manifestations of viewpoint are currently scattered over different disciplines and dominated by qualitative methods. This article presents the ViewPoint Identification Procedure (VPIP), the first systematic method for the lexical identification of markers of perceptual, cognitive and emotional viewpoint in narrative discourse. Use of this step-wise procedure is facilitated by a large appendix of Dutch viewpoint markers. After the introduction of the procedure and discussion of some special cases, we demonstrate its application by discussing three types of narrative excerpts: a literary narrative, a news narrative, and an oral narrative. Applying the identification procedure to the full news narrative, we show that the VPIP can be reliably used to detect viewpoint markers in long stretches of narrative discourse. As such, the systematic identification of viewpoint has the potential to benefit both established viewpoint scholars and researchers from other fields interested in the analytical and experimental study of narrative and viewpoint. Such experimental studies could complement qualitative studies, ultimately advancing our theoretical understanding of the relation between the linguistic presentation and cognitive processing of viewpoint. Suggestions for elaboration of the VPIP, particularly in the realm of pragmatic viewpoint marking, are formulated in the final part of the paper.

Introduction
During the long ride to the subway station, she and her husband did not exchange a word, and every time she glanced at his old hands, clasped and twitching upon the handle of his umbrella, and saw their swollen veins and brown-spotted skin, she felt the mounting pressure of tears. (Nabokov, 1948) Stories allow us to do what seems to be impossible in real life: to get a glimpse inside someone else's mind (Cohn, 1978;Palmer, 2004). As we read the short excerpt of Nabokov's Signs and Symbols (1948) above, we gain access to the perceptions, thoughts and feelings of one of the story's characters, an older woman whose son has just tried to take his own life. Through the use of perceptual verbs like glanced and saw we come to see what she sees: the hands of her seemingly agitated husband. And in the last clause, we get an impression of her sadness through the use of the verb felt. These types of verbs are characteristic of narrative discourse and are considered viewpoint (or perspective) techniques, i.e., linguistic elements that grant us access to the internal and subjective viewpoints of characters within a narrative.
Viewpoint refers to the expression of a subject's position in relation to objects and scenes. In this conceptualization, viewpoint has two aspects: a vantage point from which an object or scene is presented or viewed, and an orientation, i.e., the resulting depiction of the object or scene from that specific point (Langacker, 1987).
If the vantage point changes, so does the orientation. This implies that viewpoint is by its very nature subjective, and that its manifestation in language creates a personalized -and therefore restricted -account of an object or scene.
In discourse studies, viewpoint is typically conceptualized as a multidimensional concept. Vandelanotte (2017), for instance, distinguishes deictic viewpoint from cognitive viewpoint: whereas deictic viewpoint refers to the spatiotemporal position from which a subject views an object or situation (that is, the vantage point), cognitive viewpoint captures all of the subject's mental states and activities such as thoughts and attitudes (see also Farner, 2014;and Uspensky, 1973, for similar views). Additional dimensions that have been put forward include perceptual, emotional, and moral viewpoint (Sanford and Emmott, 2012;Van Krieken, Hoeken, and Sanders, 2017). The emotional and moral dimensions overlap with the concept of stance, which refers to 'the linguistic means by which speakers and writers convey their personal attitudes and emotions, their evaluations and assessments, and their level of commitment towards propositions' (Gray and Biber, 2014: 219). Lexical stance markers are expressions of personal feelings, attitudes, value judgments, or assessments, such as evaluative adjectives (beautiful, angry) and adverbials (surprisingly, unfortunately; see for example Pearce, 2005). Such markers give expression to a subject's emotional or moral experience of something and, therefore, to the subject's viewpoint. The difference between stance and viewpoint is that the latter concept includes more categories than stance alone, such as the sensory perceptions of a subject which do not necessarily mark the subject's stance towards an object (e.g., He saw an orange tree). Thus, while stance markers are typically also (moral or emotional) viewpoint markers, not all viewpoint markers are stance markers.
The study of viewpoint in narrative discourse has a long tradition in both literary studies and linguistics. A central aim of these studies is to elucidate how language is exploited in narrative discourse to describe events and situations from the subjective viewpoints of characters, and how this language use contributes to the aesthetic, rhetorical, functional, affective, and cognitive effects of narratives. The linguistic manifestation of viewpoint is studied both at a text-wide level, with a focus on the use of grammatical person (first, second, or third), and at lower levels of the discourse, with a focus on linguistic phenomena at lexical and sentence levels. Studies of this latter category have mainly adopted qualitative methods to analyze viewpoint in stretches of fictional as well as nonfictional narrative discourse, including the use of verb tense and free indirect discourse (e.g., Dancygier and Vandelanotte, 2009;Dancygier, 2017;Nikiforidou, 2010;Van Duijn, Sluiter, and Verhagen, 2015). The present article aims to foster quantitative research on local-level viewpoint phenomena in narrative discourse, by developing an identification procedure for lexical viewpoint markers.
Thus far, relatively few studies have employed quantitative methods to study narrative viewpoint. These studies have identified a range of linguistic manifestations of viewpoint, at multiple levels of analysis and in different types of narratives. For example, Habermas (2006) and Habermas and Diel (2010) examined non-fictional Eekhof et al: VPIP 4 oral narratives on the propositional level for the presence of viewpoint by analyzing the use of mental verbs, direct and indirect speech, and the historic present. Other studies have analyzed viewpoint techniques in journalistic narratives by examining speech and thought representations (e.g., Sanders, 2010;Van Krieken and Sanders, 2016a), referential expressions and grammatical roles Van Krieken and Sanders, 2016b), and tense and temporal adverbs (Van Krieken and Sanders, 2019). Similar analyses have also been applied to literary fiction (e.g., Ikeo, 2014).
General conclusions to be drawn from these quantitative studies are, first, that the expression of viewpoint is constitutive in narrative discourse and that linguistic markers indicating such viewpoints are ubiquitous in narrative discourse. Second, the linguistic manifestation of viewpoint in narrative discourse is highly diverse, which can be explained by the multidimensional nature of viewpoint (e.g., Farner, 2014;Uspensky, 1973). A Linguistic Cues Framework was recently presented that establishes connections between specific linguistic viewpoint markers on the one hand and the viewpoint dimension they give expression to on the other, distinguishing between spatiotemporal viewpoint, perceptual viewpoint, cognitive viewpoint, emotional viewpoint, moral viewpoint, and embodied viewpoint . A central premise of the framework is that these six dimensions are independently regulated by the use of particular linguistic cues. For example, verbs of perception (e.g., see, hear) are argued to indicate that a character's perceptual viewpoint is represented, while verbs of cognition (e.g., think, want) are argued to indicate that a character's cognitive viewpoint is represented. Each of these markers is thus, in its own way, an instruction to interpret a particular part of the discourse from a subjective viewpoint.
Taking the Linguistic Cues Framework  as an anchor, a lexical identification procedure can be developed for the identification of these viewpoint markers in narratives. Such a procedure could help establish a unified approach to the study of viewpoint in narrative, which currently appears to be scattered across disciplines and methods, and advance quantitative analyses of viewpoint markers. This may in turn benefit experimental research on the effects and processing of viewpoint by providing a ground for the identification and manipulation of viewpoint markers in narrative stimuli. This is important, because previous empirical research on viewpoint has often exclusively focused on text-wide viewpoint manipulations, comparing first-, second-, and third-person narration (i.e., grammatical viewpoint, e.g., Brunyé et al., 2016;Brunyé et al., 2011;Brunyé et al., 2009;Child, Oakhill, and Garnham, 2018;Ditman et al., 2010;Mulcahy and Gouldthorp, 2018). Studies on intra-textual viewpoint markers remain scarce, although there is evidence that these markers play a pivotal role in readers' experience and interpretation of narrative discourse Sanders and Redeker, 1993). For example, Van Krieken (2018) has shown that the presence of viewpoint markers, such as perception verbs like look, guides readers' interpretations of ambiguous perceptions such that these perceptions are represented as coming from the story character rather than the narrator. Similar effects are to be expected for narrative processes such as narrative engagement, transportation and persuasion (see Van Krieken et al., 2017, for example).
In this article, we therefore introduce the ViewPoint Identification Procedure, 'VPIP', a method for identifying the presence of perceptual, cognitive, and emotional viewpoint in narrative discourse. The VPIP was developed with three important goals in mind: 1) user-friendliness: the VPIP should be easy to use for both established viewpoint researchers as well as researchers from other fields (e.g., psychologists, cognitive neuroscientists, and psycholinguists) who wish to study viewpoint in an experimental context; 2) replicability: in order to be able to replicate experiments and analyses studying viewpoint across researchers and texts, the VPIP should be as straightforward and consistent as possible; and 3) implementability: the output of the VPIP should ideally align with the most detailed measures of linguistic processing (e.g., online measures like eye-tracking) that can be used in viewpoint experiments. This means that our procedure aims at analyzing narrative discourse on the lexical level, which is also the smallest level on which viewpoint information can be conveyed (see Krippendorff, 2018). For now, the VPIP Eekhof et al: VPIP 6 focuses on perceptual, cognitive, and emotional viewpoint. The reason for this is threefold: first of all, we believe these levels can be unequivocally and meaningfully identified on the lexical level, contrary to, for example, spatiotemporal viewpoint, which is also expressed through grammatical relations and choices that transcend the word level. Secondly, we believe these three dimensions of viewpoint are of relevance to a broad range of researchers who wish to study the processing and effects of viewpoint, as they have clear correlates in cognition and behavior (e.g., mental imagery, mindreading, and empathy). Finally, at least for moral viewpoint, analytical approaches are already available (that is, in terms of evaluation : Hunston and Thompson, 2000;appraisal: Martin and White, 2007;and stance: Biber et al., 1999).
We hope that the resulting procedure presented here will prove to be as useful and In what follows, we will first introduce the procedure. We will then discuss some special cases before illustrating the use of the procedure by applying it to three short Dutch examples (a literary narrative, a news narrative, and an oral narrative). The reliability of the procedure will be tested on a full-length Dutch news narrative. Finally, we will describe the possible applications, optional extensions, and limitations of our procedure in the discussion.

Procedure
The procedure for identifying perceptual, cognitive, and emotional viewpoint markers is graphically represented in Figure 1, below. In what follows, we will discuss the steps in more detail.
Step 1) Read the text Raters should first read the text thoroughly to get a global impression of its meaning and use of viewpoint techniques.  Step 2) Divide the text into lexical units For the purposes of the present identification procedure, all words can be considered lexical units. The only special cases that deviate from this rule are (complex) phrasal verbs (e.g., Dutch opmerken ('to notice'), as in hij merkte het lawaai op (lit. 'he noticed the noise up'); see also Steen et al., 2010), which should be considered single lexical units. We used the electronic version of the Van Dale Groot woordenboek der Nederlandse taal (Den Boon and Geeraerts, 2005) to identify phrasal verbs.
Repeat the following steps for every lexical unit: Step

3) Determine the word type
Although viewpoint can be expressed by both function words and content words, perceptual, cognitive, and emotional viewpoint are almost exclusively expressed through content words. 1 Function words usually give rise to other dimensions of viewpoint. For example, interjections like Gee! or Wow! express a moral evaluation or attitude (i.e., stance), while determiners play a role in spatiotemporal viewpoint . As the VPIP is concerned with perceptual, cognitive, and emotional viewpoint, the remainder of the procedure is solely applied to content words: a. If the lexical unit is a content word (nouns, verbs, adjectives, adverbs), continue to step 4.
b. If the lexical unit is a function word (interjections, determiners, prepositions, complementizers, pronouns), mark it as not a perceptual, cognitive, or emotional viewpoint marker and go back to step 3 for the next lexical unit.

Step 4) Determine the viewpoint dimension
Is the lexical unit related to: a. …the perceptions by one of the senses (visual, auditory, tactile, olfactory, taste) and/or bodily sensations of one of the characters or narrators of the story? If yes, continue to step 5a. If not, continue to step 4b.
b. …the thoughts, beliefs, intentions, and/or desires of one of the characters or narrators of the story? If yes, continue to step 5b. If not, continue to step 4c.
c. …the emotions of one of the characters or narrators of the story? If yes, continue to step 5c. If not, score the lexical unit as not a perceptual, cognitive, or emotional viewpoint marker and go back to step 3 for the next lexical unit.
Step 5) Determine whether the lexical unit is a viewpoint marker for that dimension a. Perceptual dimension I. Is the lexical unit a verb of perception or a content word morphologically related to such a verb? If yes, score the lexical unit as a perceptual viewpoint marker. If not, continue to step 5a.II.
II. Is the lexical unit a verb of bodily sensation or a content word morphologically related to such a verb? If yes, score the lexical unit as a perceptual viewpoint marker. If not, score the lexical unit as not a perceptual, cognitive, or emotional viewpoint marker and go back to step 3 for the next lexical unit.
b. Cognitive dimension I. Is the lexical unit a verb of cognition or a content word morphologically related to such a verb? If yes, score the lexical unit as a cognitive viewpoint marker. If not, continue to step 5b.II.
II. Is the lexical unit an epistemic modal adverb? If yes, score the lexical unit as a cognitive viewpoint marker. If not, score the lexical unit as not a perceptual, cognitive, or emotional viewpoint marker and go back to step 3 for the next lexical unit. c. Emotional dimension I. Is the lexical unit a verb of emotion or a content word morphologically related to such a verb? If yes, score the lexical unit as an emotional viewpoint marker. If not, continue to step 5c.II.
II. Is the lexical unit an adjective of emotion or a content word morphologically related to such an adjective? If yes, score the lexical unit as an emotional viewpoint marker. If not, score the lexical unit as not a perceptual, cognitive, or emotional viewpoint marker and go back to step 3 for the next lexical unit.
The different viewpoint markers in step 5 can be identified using the definitions and examples in Table 1. In case of uncertainty when determining whether a lexical unit meets these definitions, an additional paraphrase test can be done: if the lexical unit under investigation can be replaced by or paraphrased with the use of one of the basic forms of a particular dimension, it is a viewpoint marker for that dimension.
For example, in the sentence Hij snakte naar een avondje rust ('He craved a quiet evening'), the verb snakken ('to crave') can be replaced by the basic form willen ('to want'): Hij wilde een avondje rust ('He wanted a quiet evening'). Therefore snakte can be considered a cognitive viewpoint marker. Paraphrase tests have proven useful and reliable for the detection of similar linguistic phenomena. 2 Table 1 shows an overview of the markers and their definitions by viewpoint dimension and provides the basic forms and examples. Note that all content words morphologically related to (but not necessarily derived from) any of the categories of viewpoint markers are also considered to be viewpoint markers (see the column on the right of the table).
To further facilitate the viewpoint identification process, we have compiled a list of Dutch verbs of perception and bodily sensation (steps 5a.I and 5a.II), verbs of cognition and epistemic modal adverbs (steps 5b.I and 5b.II), and verbs and adjectives of emotion (step 5c.I and 5c.II) that can be found in the Appendix. The list was developed as follows. For the four types of verbs, verb classes from work by Levin (1993) were identified that satisfied the definitions from Table 1. These can be found in Table 2.
Verb of bodily sensation A verb that denotes an internal bodily state experienced by a body part.

Pijn doen ('to hurt')
Duizelen Verb of cognition A verb that represents thoughts, beliefs, intentions, and/or desires.

Morphologically related words
Epistemic modal adverb An adverb that indicates a subject's certainty about a claim or description of reality (see Klein, 1998;Salverda, 2003).

Not applicable
Emotional Verb of emotion A verb that denotes the experience of emotion.

Schamen
('to feel ashamed') Enthousiasmeren Adjective of emotion An adjective that denotes the experience of emotion.
The verbs of these verb classes were then translated into Dutch, taking only those English meanings and Dutch translations into consideration that were related to the perceptual, cognitive, and emotional viewpoint dimensions and satisfied the definitions given in Table 1. In addition, the closed class of epistemic modal verbs was added to the list of verbs of cognition. With these verbs, speakers indicate commitment to the validity of a proposition on the basis of their estimation of the probability that a particular state of affairs is the case (Sanders, 1994: 146). Such estimates are subjective by nature and thus express subjective viewpoints (see also Spooren, 1996, 1997).
The class of emotion adjectives was compiled based on work by Hevner (1936) on emotional adjectives used to describe music, a revised version of Hevner's adjective list (Schubert, 2003), and Dutch translations of the adjectives used in the Multifaceted Empathy Test (Dziobek et al., 2008;Eekhof, van Krieken, Sanders, and Willems, in preparation). All translations were made using the electronic version of The epistemic modal adverbs were taken from Salverda (2003) and the electronic version of Van Dale.

Special Cases
In general, application of the identification procedure should be straightforward.
However, there are a few special cases that require extra attention. These are

Ambiguity
In some cases, words with multiple meanings can receive different viewpoint scores depending on the meaning that is intended given the context. For example, the Dutch word zullen ('shall/will') can either be used as a temporal auxiliary when forming the future tense (see Example 1), or to signal epistemic modality. Only when the verb is used in an epistemic modal sense, is it considered a cognitive viewpoint marker (see Example 2). Another example is the verb vinden ('to maintain' or 'to find'), which can either signal cognitive viewpoint (see Example 3), when it is used to express an opinion (see also Vis, Sanders, and Spooren, 2015), or perceptual viewpoint (see Example 4), when it is used in the sense of discovering something. Similarly, the verb moeten ('to need/to have to/should') can have multiple meanings including a deontic and epistemic interpretation. Only in those occurrences with an epistemic interpretation, related to the thoughts, beliefs, intentions, and/or desires of one of the characters of the story, should these be considered cognitive viewpoint markers (see Examples 3 and 5). Deontic interpretations, characterized by the presence of an external or objective force, as in Example 6, are not part of the cognitive viewpoint dimension (see Spooren, 1996, 1997

Inanimate Subjects
There are instances in which something inanimate, rather than one of the characters, is the subject in a sentence with a viewpoint marker, as in the examples below. If, in these cases, the viewpoint is nevertheless to be understood as coming from one of the characters or narrators, the lexical unit should still be considered a viewpoint marker. In the first example above, the intention that is expressed by the verb must is to be understood as coming from the he that is taking the pills. Hence, although the pills are the subject of must, the verb is still a cognitive viewpoint marker as it signals the intention of the character. Similarly, in the second example, the feeling that the time had been lonely and grim is experienced by the character. Therefore, these two adjectives should be considered emotional viewpoint markers.
Note that in more extreme cases, something inanimate might be the main character of a story, as in the Dutch novel Specht en Zoon ('Woodpecker and Son') by Willem Jan Otten, in which the main character and narrator is a painting canvas (see Trompenaars, 2018;Trompenaars et al., 2018

Three Sample Narratives
To illustrate the identification procedure, we will now discuss three sample narratives.
The first sample is an excerpt from a literary story by the Dutch literary author  Table 1.

Literary Fiction: The Byzantine Cross by A. F. Th. van der Heijden (2008)
The Byzantine Cross is a short story about a person with an obsession for scissors, which he uses to break into cars. The following passage describes a scene in the store where the man usually buys his scissors. He is afraid that the woman at the till will recognize him from his frequent visits to the store.

Probeerde (tried) = cognitive viewpoint marker
This lexical unit is related to the intentions of the main character, which falls under the cognitive dimension (step 4b). In step 5b.I, we see that the verb is not on the list of verbs of cognition in the appendix, but it does satisfy the definition of a verb of cognition given in Table 1: it is a verb that represents intention. As a result, this lexical unit is scored as a cognitive viewpoint marker.

Overwoog (considered) = cognitive viewpoint marker
This lexical unit is related to the thoughts of the main character about his course of action, which falls under the cognitive dimension (step 4b). In step 5b.I, we see that the verb overwegen ('to consider') is on the list of verbs of cognition in the appendix.
As a result, this lexical unit is scored as a cognitive viewpoint marker.

Leek (seemed) = cognitive viewpoint marker
This lexical unit is related to the beliefs of the main character about the suspiciousness of putting the scissors back. In step 5b.I, we see that the verb lijken ('to seem') is on 5 Literal translation: The / next / time / tried / I / at / an / other / till / particle / to / pay. / But / I / was / referred. / I / considered / still / the / thing / back / to / put, / but / that / seemed / me / completely / suspicious. / Everywhere / away / felt / I / cameras / on / me / pointed.

Eekhof et al: VPIP 20
the list of verbs of cognition in the appendix. As a result, this lexical unit is scored as a cognitive viewpoint marker.

Verdacht (suspicious) = cognitive viewpoint marker
This lexical unit is related to the (hypothetical) thoughts of one of the characters of the story, namely the cashier referred to as haar ('her'). In step 5b.I, we see that this adjective is morphologically related to the verb verdenken, which is on the list of verbs of cognition in the appendix. As a result, this lexical unit is scored as a cognitive viewpoint marker.

Voelde (felt) = perceptual viewpoint marker
The verb voelen ('to feel') can either refer to a physical sensation (e.g., I feel the sun on my skin), which falls under the perceptual dimension, or the experience of emotion (e.g., I feel bad), which falls under the emotional dimension. In this case, the perceptual dimension is evoked (step 4a), but with a hyperbolic interpretation: the man's perception of the cameras is probably affected by his anxious state, as it is unlikely that there are cameras pointed at him from every direction. In sum, the perceptual meaning of voelen ('to feel') is metaphorically projected on the emotional domain, providing an instance of subjectification (Kissine, 2010;Traugott, 1989).
However, because the VPIP is concerned with viewpoint rather than metaphor, we decided to code the semantically primary meaning. Based on these considerations, this lexical unit is scored as a perceptual viewpoint marker.

News Narrative: Crime Report (NRC Handelsblad, 2011)
This news narrative describes a shooting that took place in April 2011 in a shopping mall in the Dutch city of Alphen aan den Rijn. In this passage, a shop owner is interviewed about the aftermath of the incident.

Shock (shock) = emotional viewpoint marker
This lexical unit is related to the emotional dimension (step 4c) as it expresses the emotion that main character Charradi and others felt when they went back to the crime scene. In step 5c.I, we see that although the verb (to) shock, which is the English loan verb from which this noun is derived, is not on the list of verbs of emotion in the appendix, the Dutch counterpart schokken ('to shock') is. As a result, this lexical unit is scored as an emotional viewpoint marker.

Verschrikkelijk (horrible) = emotional viewpoint marker
This lexical unit refers to the emotions that main character Charradi experiences when she learns one of her colleagues has died. In step 5c.II, we see that the lexical unit matches the definition of an adjective of emotion given in Table 1: it is an adjective that denotes the emotion of horror. As a result, this lexical unit is scored as an emotional viewpoint marker.

Ziet (see) = perceptual viewpoint marker
This lexical unit is related to the visual perception of the main character, which falls under the perceptual dimension (step 4a). In step 5a.I we see that the verb zien ('to 6 Literal translation: Later / go / they / back. / 'Everyone / was / in / shock', / says / Charradi. / 'Diagonally / opposite to / us / is / also / a / fashion store. / The / (female) owner / there of / has / died. / Horrible. / A / colleague / that / you / every / day / see. / You / believe / it / not. / You / have / the / feeling / that / you / in / a / movie / are / in which / you / not / want / be'. see') is on the list of verbs of perception in the appendix. As a result, this lexical unit is scored as a perceptual viewpoint marker.

Gelooft (believe) = cognitive viewpoint marker
This lexical unit is related to the (dis)belief of the main character about the tragic situation, which falls under the cognitive dimension (step 4b). In step 5b.I we see that the verb geloven ('to believe') is on the list of verbs of cognition in the appendix.
As a result, this lexical unit is scored as a cognitive viewpoint marker.

Gevoel (feeling) = emotional viewpoint marker
This lexical unit refers to the emotions of the main character (step 4c). In step 5c.I we see that this noun is morphologically related to the verb voelen ('to feel'), which is on the list of verbs of emotion in the appendix. As a result, this lexical unit is scored as an emotional viewpoint marker.

Wil (want) = cognitive viewpoint marker
This lexical unit is related to the desire of the main character to not be in the situation she found herself in, which falls under the cognitive dimension (step 4b). In step 5b.I, we see that the verb willen ('to want') is on the list of verbs of cognition in the appendix. As a result, this lexical unit is scored as a cognitive viewpoint marker.

Oral Narrative: The Road Trip (CGN, 2000)
In this excerpt, taken from the Spoken Dutch Corpus (CGN, 2000), three friends reminisce about their youth. One of the friends then proceeds to tell a short story about one of their road trips. Please note that names and place names have been removed for the sake of anonimity.
-God / wat / nog / een / keer / eens / gelachen EVP / in / in / zo'n / bussie -toen / moest / ik / rijden waren / we / naar / PLACE NAME / of / PLACE NAME / weet CVP / ik / veel / waar / we / waren hadden / we / eerst / NAME / enorm / z'n / bek / gehouden -die / wou CVP / niks / meer / zeggen -zaten / we / op / 't / terras / en / NAME / raakte / met / een / meisje / aan / de / praat / en / z'n / eerste / vraag / was / wat / studeer / jij -nou / dat / heb / ie / de / hele / hele / dag / moeten / horen PVP -hij / zegt / van / als / 't / zo / moet CVP / dan / gaan / dan / ga / ik / naar / huis 7 God on another occasion we laughed so much in such a van. That time I was the driver. We went to PLACE NAME or PLACE NAME, well I don't know where we were. We had first shut NAME's mouth so much. He didn't want to say anything anymore. We were sitting at an outdoor café and NAME started talking to this girl and his first question was 'what do you study?' Well, he has had to hear it from us all day long. He says like: if it has to go like this, then I'll go home.

Gelachen (laughed) = emotional viewpoint marker
This lexical unit is related to the emotion of the main characters, including the narrator (step 4c). In step 5c.I we see that lachen ('to laugh') is on the list of verbs of emotion in the appendix. As a result, this lexical unit is scored as an emotional viewpoint marker.

Moest (had)
It is clear that this lexical unit is not related to the perceptions or bodily sensations (step 4a) of one of the characters. When judging whether it is related to the cognitive dimension (step 4b), we should keep in mind that only epistemic interpretations of 7 Literal translation: -God / what / still / one / time / once / laughed / in / in / such a / van -then / needed / I / drive -were / we / to / PLACE NAME / or / PLACE NAME / know / I / much / where / we / were -had / we / first / NAME / tremendously / his / mouth / kept -he / wanted / nothing / anymore / say -sat / we / on / the / terrace / and / NAME / got / with / a / girl / on / the / chat / and / his / first / question / was / what / study / you -well / that / has / he / the / whole / whole / day / must / hear -he / says / like / if / it / so / must / then / go / then / go / I / to / home the verb moeten ('to have to') are considered to be related to this dimension. In this case, the interpretation is deontic: there is an (undisclosed) external force such that the narrator was the one who 'had to' drive. So, the lexical unit is not related to the cognitive dimension, nor to the emotional dimension, and is discarded after step 4c. This lexical unit is not a viewpoint marker of the perceptual, cognitive, or emotional dimension.

Weet (know) = cognitive viewpoint marker
This lexical unit refers to the beliefs of the narrator about the destination of the road trip, which falls under the cognitive dimension (step 4b). In step 5b.I, we see that the verb weten ('to know') is on the list of verbs of cognition in the appendix. As a result, this lexical unit is scored as a cognitive viewpoint marker.

Wou (wanted) = cognitive viewpoint marker
This lexical unit refers to the desire of one of the characters to not speak anymore, which falls under the cognitive dimension (step 4b). In step 5b.I, we see that the verb willen ('to want') is on the list of verbs of cognition in the appendix. As a result, this lexical unit is scored as a cognitive viewpoint marker.

Moeten (had)
Again, we have to analyze the interpretation of the verb moeten ('to have to') before we can rate it. Although not as straightforward as the previous occurrence of moest, this case can also be seen as having a deontic meaning by virtue of the presence of an external force: his friends' constant talking about the incident leads to the fact that Bert-Jan 'had to' listen to them all day. So, the lexical unit is not related to the cognitive dimension, nor to the emotional dimension, and is discarded after step 4c. This lexical unit is not a viewpoint marker of the perceptual, cognitive, or emotional dimension.

Horen (hear) = perceptual viewpoint marker
This lexical unit is related to the auditory perception of the narrator, which falls under the perceptual dimension (step 4a). In step 5a.I we see that the verb zien ('to see') is on the list of verbs of perception in the appendix. As a result, this lexical unit is scored as a perceptual viewpoint marker.

Moet (must) = cognitive viewpoint marker
For this final occurrence of moeten ('to have to') the context does not present any external forces that would justify a deontic interpretation. Rather, it is the character who internally reaches the conclusion that things will probably go a certain way.
So the lexical unit, in its epistemic interpretation, is related to the thoughts and beliefs of a character, which falls under the cognitive dimension (step 4b). In step 5b.I, we see that the verb moeten ('to have to') is on the list of verbs of cognition in the appendix. As a result, this lexical unit is scored as a cognitive viewpoint marker.

Reliability Analysis
The full news story from section 4.2 was used to test the reliability of our procedure.
First, the text was divided into 1145 lexical units by the first author. Then, the first and second author independently applied the VPIP to the story (see Table 3 for the results). Raters agreed on scores for 1128 of the 1145 lexical units, making interrater agreement almost perfect (98.52%, κ = 0.87; Landis and Koch, 1977). There were 17 cases of disagreement. In most of these cases (n = 14), one of the raters scored the lexical units as a viewpoint marker whereas the other did not. In the three remaining cases, raters agreed on the viewpoint marking nature of the lexical unit, but disagreed on the dimension. In total, 63 lexical units (5.50% of the text) were 6. Discussion

Contributions and Applications
In this article, we presented the VPIP, a lexical identification procedure for perceptual, cognitive, and emotional viewpoint in narrative discourse. The VPIP uses a detailed step-wise procedure to identify the lexical elements that signal the perceptual, cognitive, and emotional viewpoints of story characters and narrators. Application of the procedure is further facilitated by a large appendix of Dutch viewpoint markers.
We have demonstrated that the VPIP can be applied to a wide variety of narratives, ranging from spontaneous, oral narratives to stylized, literary narratives. In addition, the good results of the reliability analysis indicate that the procedure can be used by multiple raters with substantial agreement.
We believe the VPIP can be used by a broad and diverse group of scholars ranging from literary scholars to cognitive neuroscientists.  Sentence (b) represents an utterance spoken by Charradi, which is indicated by the explicit embedding through the reporting verb says, as well as by quotation marks.

Limitations and Optional Extensions
This entails that the validity of these particular words, both in content and wording, is limited to this subjective viewpoint. In addition, sentences (f-h) are pragmatically embedded as impressions of what Charradi and her colleagues exchanged when they went back (a) and found the horrible (e) news that the owner of the shop across the street had died (c-d). In (b), Charradi explicitly describes in a past tense sentence how they all felt: Everybody was in shock. In sentences (f-h), by contrast, she shows how they all felt by representing their impressions in present tense with a generic you, blending her own voice with the voices of the others. The present tense with you demarcates a shift to a free indirect speech representation mode (Sanders, 2010), which is embedded within the direct quote. A pragmatic analysis of viewpoint would allow for an analysis of such embedded viewpoints, and the different internal and external voices and viewpoints involved and intertwined in the narration, which could elucidate how linguistic perspective manifests itself and functions at different layers of the narrative. Hence, for researchers interested in these phenomena, the VPIP could be extended with a qualitative, more pragmatic analysis.
Researchers working with literary or otherwise stylized narratives in which multi-word units like idioms play an important role may wish to extend the VPIP to also include collocations, fixed expressions, and idioms (e.g., I am at my wits' end). One way this could be achieved is by adding another paraphrase test at the end of the identification procedure. After applying the VPIP to the full narrative, raters could look for multi-word viewpoint markers by checking whether any multi-world units present in the text can be replaced by a single-word viewpoint marker of one of the three relevant categories (e.g., I am at my wits' end can be rephrased as I am worried, justifying an emotional viewpoint marker rating; she feasted her eyes on the beautifully decorated cupcakes can be rephrased as she looked at the beautifully decorated cupcakes, justifying a perceptual viewpoint marker rating).
Finally, at present our procedure does not take into account the spatiotemporal, moral, and embodied dimensions of viewpoint as described in .
Spatiotemporal viewpoint is expressed by syntax and anaphora (Kuno, 1987; see also . Moral viewpoint is interpreted on the basis of evaluations in the narrative that underpin the rhetorical intentions of telling the story. Research in this tradition can be traced back to studies of oral storytelling by Labov and Waletzky (1967) and Tannen (1982), and to studies of story plots and narrative archetypes, rooted in Propp (1928) and Campbell (1949; see also Sanders & Van Krieken, 2018).
Among others, Martin and White (2007) have described how evaluations, attitudes or stance are expressed linguistically and how they can be analyzed in narrative discourse. Embodied viewpoint is an aspect of mental simulation of narrative as evoked by expressions of shape, orientation, and movement, and is studied in neuroimaging studies of sensory and motor simulation such as Zwaan (1999Zwaan ( , 2009).
In the future, the VPIP could be extended to also include these dimensions.
To conclude, we believe that in its current form the VPIP can be a helpful tool to systematically identify perceptual, cognitive, and emotional viewpoint markers. For researchers interested in the experimental and analytic study of these manifestations of viewpoint, the VPIP opens many horizons for the study of viewpoint.

Additional File
The additional file for this article can be found as follows: • Appendix. List of Dutch markers of perceptual, cognitive, and emotional viewpoint. DOI: https://doi.org/10.16995/olh.483.s1