Annotation Guideline No. 7 (revised): Guidelines for annotation of narrative structure

Analysis of narrative structure can be said to answer the question “Who tells what, and how?”. The key part of our annotation scheme is related to the “who?”, and to this end we distinguish between narration and fictional dialogue. Furthermore, with respect to the latter we keep track of turns, lines, identities of speakers and addressees, and speech-framing constructions, which provide the narrator’s cues about the circumstances of the speech. We also annotate voice, that is, whether the narrator is ever present in the story or not. Our annotation of the “what?” includes embeddings of narrative transmission levels to capture stories in stories, and embeddings of fictional dialogue to capture characters quoting other characters. Our annotation of the “how?” includes focalization, that is, the perspective from which the narrative is seen and how much information the narrator has access to.

Our annotation of the "what?" is more rudimentary. In addition to keeping track of the characters taking part in fictional dialogue, we annotate embeddings of narrative transmission levels to capture stories in stories, and embeddings of fictional dialogue to capture characters quoting other characters. On the other hand, we do not distinguish events, let alone how timewise and causally related events form a plot.
Our annotation of the "how?" is also rudimentary. We annotate focalization, that is, the perspective from which the narrative is seen and specifically how much information the narrator has access to.
In developing our annotation scheme, we have adopted a hierarchical tagset with mutually exclusive tag categories within the same layers to increase the transparency of the scheme, and thereby hopefully inter-annotator agreement. For simplicity, we use embedded (in-line) annotation in the original text. A summary of the annotation is given in Summary of annotation layers.

Revisions
This section summarizes the changes relative to the first version of the guidelines, published in Cultural Analytics on January 15, 2020.
1. We have removed the category <SCENE>.
2. We have removed the category <AUTHOR>. We thus only annotate two levels of narrative transmission: <NARRATOR> and <CHARACTERS>. In addition, <CHARACTERS> is embedded into <NARRATOR> instead of iterating the two at the same level.
3. We have added the category <NC> (for narrative construction) to tag speechframing constructions in lines in characters' discourse, for example: (1) "He looks like a thief", <NC> said Sellén, watching him from the window with a sly smile </NC>.

Narrative transmission
Our starting-point is the following levels of narrative transmission, from higher to lower; 4 compare Figure 1: 1. Transmission from the real (historical) author to any real reader of the text.
2. Transmission from an implied author to an implied reader. The implied author is the "hypothetical individual who seems to be implied by the content of the text". 5 The characteristics and intentions of this are sometimes taken to differ from those of the real author as well as from the narrator. 3. Transmission from a narrator to a narratee. We refer to this as narrator's discourse.
4. Transmission between characters in the story by means of direct speech. We refer to this as characters' discourse, or sometimes (fictional) dialogue.
As pointed out by Jahn, 7 it is only in narrator's and characters' discourse that transmission occurs in the text (that is, at the fictional or intratextual level), whereas the transmission between author and reader occurs outside of the text (at the nonfictional or extratextual level). Since we are only concerned with annotating the intratextual level, we only include narrator's and characters' discourses in our annotation scheme. As stated in Dolezel, 8  The reason for distinguishing narrator's and characters' discourses in our annotation is that they constitute different modes of story-telling, and that we assume that to some extent they require different forms of narrative analysis (whether automatic or manual). For example, while fictional dialogue is an independent narrative mode, its mimetic quality means that it is to some extent amenable to linguistic analysis using models aimed at human interaction, such as speech-act theory and Conversation Analysis. 10 To annotate characters' and narrator's discourses, we use opening and closing tags indicating the type of addresser and addressee at the respective level, namely, <NARRATOR--NARRATEE> and <CHARACTERS--CHARACTERS>. 11 For convenience, we may abbreviate <NARRATOR--NARRATEE> as <NARRATOR> and <CHARACTERS--CHARACTERS> as <CHARACTERS>. Furthermore, since each of these levels can be embedded into each other (for example, to capture stories in stories; see Embeddings of narrative transmission levels), we use indexing, with 0 as the highest level, 1 for the next embedded level, and so on; for example, <NARRATOR_0>, <NARRATOR_1>, <NARRATOR_2>.
Below is an example text which contains passages of what we annotate as narrator's and characters' discourses. 12 Melissa had her camera on the table and occasionally lifted it to take a photograph, laughing self-deprecatingly about being a 'work addict'. She lit a cigarette and tipped the ash into a kitschy-looking glass ashtray. The house didn't smell of smoke at all and I wondered if she usually smoked in there or not. I made some new friends, she said. Her husband was in the kitchen doorway. He held up his hand to ac-knowledge us and the dog started yelping and whining and running around in circles. This is Frances, said Melissa. And this is Bobbi. They're poets. He took a bottle of beer out of the fridge and opened it on the countertop. Come and sit with us, Melissa said. Yeah, I'd love to, he said, but I should try and get some sleep before this flight.
Here is how the text would be annotated for narrator's and characters' discourses: Melissa had her camera on the table and occasionally lifted it to take a photograph, laughing self-deprecatingly about being a 'work addict'. She lit a cigarette and tipped the ash into a kitschy-looking glass ashtray. The house didn't smell of smoke at all and I wondered if she usually smoked in there or not. <CHARACTERS_0> I made some new friends, she said. </CHARACTERS_0> Her husband was in the kitchen doorway. He held up his hand to acknowledge us and the dog started yelping and whining and running around in circles. <CHARACTERS_0> This is Frances, said Melissa. And this is Bobbi. They're poets. </CHARACTERS_0> He took a bottle of beer out of the fridge and opened it on the countertop.

<CHARACTERS_0>
Come and sit with us, Melissa said.
Yeah, I'd love to, he said, but I should try and get some sleep before this flight. </CHARACTERS_0> </NARRATOR_0> As illustrated above, characters' discourse represents direct speech, that is, utterances that are rendered as quotations by the narrator. Other terms that have been used for this are direct discourse and quoted speech. 13 The form of the speech can be dialogue, but other forms are spoken monologue (soliloquy) or interior monologue (thoughts). Direct speech is usually marked typographically with quotation marks or dashes and may involve paragraph breaks. In addition to quotations, passages of direct speech may include speech-framing constructions, which provide the narrator's cues about the circumstances of the speech. 14 Such a construction may include the identity of the speaker (and possibly the addressee), the narrator's assessment of the speech, and information about events associated with the speech or that happen simultaneously. Here is an example: 15 "He looks like a thief", said Sellén, watching him from the window with a sly smile.
In this case, "said Sellén" provides the identity of the speaker and "watching him from the window with a sly smile" reports what the speaker does (and how) while speaking. In our annotation scheme, we tag speech-framing constructions using the tag <NC> (narrative construction), as shown below: (3) "He looks like a thief", <NC> said Sellén, watching him from the window with a sly smile </NC>.
Fictional speech can be rendered in many ways, of which direct speech is only one. Rimmon-Kenan 16 distinguishes seven types of speech representation, ranging from the diegetic (where the speech of the characters is expressed through the narrator) to the mimetic (where the characters' speech is rendered in quoted form). Five of these belong to indirect speech. Basically, indirect speech means that the narrator describes a speech event as seen from his or her point of view. A possible rendering of the example above as indirect speech is as follows: Sellén, watching Olle from the window with a sly smile, said that he looked like a thief.
In the words of Genette, 17 "the narrator takes on the speech of the character, or, if one prefers, the character speaks through the voice of the narrator, and the two instances then are merged". In our annotation scheme, all kinds of indirect speech belong to narrator's discourse. 18

Narrator's discourse
As mentioned above, narrators and narratees may be more or less overt, but this distinction is not annotated per se according to our scheme; rather, passages of narration are annotated as narrator's discourse regardless of the degree of overtness.
Here is an example of an narratee who is overt in the sense of being explicitly addressed, though not identified as in an epistolary novel: Here is another example where the narratee is again overt in the sense of being explicitly addressed. In addition, the narrator is overt in the sense of appearing as an "experiencing I", which would be annotated as a homodiegetic voice (<VOICE_1>, see Voice).

Characters' discourse
A characters' discourse consists of dialogue between characters, using one or more turns, each of which is associated with one speaker and one or more addressees. A characters' discourse is thus a sequence of turns uninterrupted by narrator's discourses. A turn is distinguished by having a single speaker (or possibly the same set of speakers in chorus). A turn consists of one or more lines, each of which consists of one or more utterances. In contrast to a turn, a line is distinguished by having a single addressee (or the same set of addressees, if the spoken message is intended for more than one person). In other words, different lines within a turn have the same speaker(s) but different addressees. 19 We annotate turns using the tag <TURN>, and lines with their speakers and addressees along with speech-framing constructions, as in the example below. 20 We're all on the same side here, <NC> Derek said </NC>.
<Derek--ALL> Nick, you're an oppressive white male, you back me up.
<Derek--Nick> </TURN> <TURN> I actually quite agree with Bobbi, <NC> said Nick </NC>. Oppressive though I certainly am. <Nick--Derek> </TURN> </CHARACTERS_0> Here, the first turn is divided into two lines since there is a change in addressees. The second turn corresponds to one line consisting of two utterances.

Embeddings of narrative transmission levels
This section discusses how the two types of discourse can be embedded.

Narrator's discourse embedded in narrator's discourse
A narrator's discourse embedded in a narrator's discourse corresponds to what has been called narrative level 24 , in other words, a story within a story. Our criterion for making this embedding is that there is a switch of narrator, as in the following example: The story that I have to tell came to my knowledge more than half a century ago in the house of my great-grandmother, the wife of Senator Feddersen, when, sitting close up to her armchair one day, I was busy reading a number of some magazine bound in blue cardboard, either the Leipziger or Pappes Hamburger Lesefruchte, I have forgotten which. I still recall with a tremor how the old lady of more than eighty years would now and then pass her soft hand caressingly over her great-grandchild's hair. She herself, and that day, have long been buried and I have sought in vain for those old pages, so I can just as little vouch for the truth of the facts as defend them if anyone should question them. Only one thing I can affirm, that although no outward circumstance has since revived them in my mind they have never vanished from my memory.

<NARRATOR_1>
On an October afternoon, in the third decade of our century </NARRATOR_1> -thus the narrator began his tale -<NARRATOR_1> I was riding in very bad weather along a dike in northern Friesland. For more than an hour I had been passing, on the left, a bleak marsh from which all the cattle had already gone, and, on the right, uncomfortably near, the marsh of the North Sea. [. . . ] </NARRATOR_1> <NARRATOR_0> Note that, by virtue of the embedding, "-thus the narrator began his tale -" belongs to the first level (<NARRATOR_0>).
Here is an example illustrating what the annotation looks like if the narrator who appears first in the story does not belong to the first level:

Characters' discourse embedded in characters' discourse
The event of a character who is quoting the speech of another character is annotated as an embedding of a characters' discourse representing the quotation in the present characters' discourse. Analogously with embedded narrations, it is annotated as an additional opening of <CHARACTERS> with an incremented index inside the present characters' discourse, as in the following example: 25

Narrative situation
This section describes two notions related to the position and perspective of the narrator. We refer collectively to these as narrative situation, inspired by Genette, 27 but we do not use the term in the full meaning developed there.

Voice
The notion of voice concerns the narrator's relationship to the story, and more specifically whether the narrator is ever present in the story or not. 28 If the story is told by a narrator who appears as a character in the story at some point, we say that we have a homodiegetic narrative. Such narrators usually refer to themselves in the first person, but there are exceptions to this, such as Caesar's De Bello Gallico in which the narrator refers to himself in the third person. In contrast, if the story is told by a narrator who is never present as a character in the story, we say that we have a heterodiegetic narrative. Such narratives thus do not have an "experiencing I", and actions in the story (whether events or speech) are expressed in the third person.

Focalization
We take focalization to correspond to the perspective from which the narrative is seen, and specifically how much information the narrator has access to; alternatively, in what ways this information is restricted. 29 We distinguish the following types: 1. Zero or unrestricted focalization. The story is narrated from a fully unrestricted or omniscient perspective. This often involves helicopter views of the story that no single character would be capable of, but it might also involve taking the perspectives or looking into the souls of the individual characters. As put by Todorov, 30 cited in Genette, 31 the narrator knows more than any of the characters, symbolized by Narrator > Character. We annotate this as <FOC_UNR> with a corresponding closing tag </FOC_UNR>.

2.
Internal focalization. The story is narrated from the inside perspective of a character in the story, limited by the perception and feelings of that character. As put by Todorov, 32 the narrator knows only as much as this character, symbolized by Narrator = Character. We annotate this as <FOC_INT> with a corresponding closing tag </FOC_INT>.
3. External focalization. The story is narrated from a perspective outside of the characters in the story, like using a camera, but without an omniscient perspective, as in <FOC_UNR>. Typically, the main components of such narratives are dialogues and neutral descriptions of events. As put by Todorov, 33 the narrator knows less than any of the characters, symbolized by Narrator < Character. We annotate this as <FOC_EXT> with a corresponding closing tag </FOC_EXT>.

Summary of annotation layers
Our tagset is hierarchically structured in four layers, ordered by an inclusion relation. From the top, these are voice, focalization, <NARRATOR> (narrator's discourse) and <CHARACTERS> (characters' discourse). The hierarchical inclusion relation means that a change of a higher-level tag affects all lower-level tags. For example, changing voice means that focalization and the other tags have to be reset even if their values do not change (see example below). (This is a price we pay for the hierarchical tagset.) A characters' discourse, <CHARACTERS>, consists of one or more turns, each of which is associated with one speaker (or a chorus of several speakers) and one or more addressees (who may vary). Thus, whereas the speaker (or set of speakers) is immutable throughout a turn, this does not necessarily hold for the addressee(s). Each turn consists of one or more lines, each of which is associated with one speaker (or a chorus of several speakers) and one addressee (or a set of addressees). Lines may contain speech-framing constructions, which provide the narrator's cues about the circumstances of the speech. In addition, a line consists of one or more utterances, that is, sentences or fragments typically distinguished by full stops in the text. Utterances are not part of the annotation, however.
Note that only lines are tagged with speaker(s) and addressee(s). Note also that we make a differerence between addressees and listeners: addressees are the recipients of lines, whereas listeners overhear lines without being recipients. We only annotate addressees. In sum, a characters' discourse is a sequence of turns uninterrupted by narrator's discourses. A description of the layers and discourse levels is shown in Table 1. more than an hour I had been passing, on the left, a bleak marsh from which all the cattle had already gone, and, on the right, uncomfortably near, the marsh of the North Sea. A traveler along the dike was supposed to be able to see islets and islands; I saw nothing however but the yellow-gray waves that dashed unceasingly against the dike with what seemed like roars of fury, sometimes splashing me and the horse with dirty foam ; in the background eerie twilight in which earth could not be distinguished from sky, for the moon, which had risen and was now in its second quarter, was covered most of the time by driving clouds. It was icy cold. My benumbed hands could scarcely hold the reins and I did not blame the crows and gulls that, cawing and shrieking, allowed themselves to be borne inland by the storm. Night had begun to fall and I could no longer distinguish my horse's hoofs with certainty; not a soul had met me ; I heard nothing but the screaming of the birds, as their long wings almost brushed against me or my faithful mare, and the raging of wind and water. I do not deny that at times I wished myself in some secure shelter.
It was the third day of the storm and I had allowed myself to be detained longer than I should have by a par-ticularly dear relative at his farm in one of the northern parishes. But at last I had to leave. Business was calling me in the town which probably still lay a few hours' ride ahead of me, to the south, and in the afternoon I had ridden away in spite of all my cousin and his kind wife could do to persuade me, and in spite of the splendid home-grown Perinette and Grand Richard apples which were yet to be tried.

<CHARACTERS_0>
<TURN> "Just wait till you get out by the sea," <NC> he had called after me from the door </NC> ,"you will turn back then; we will keep your room ready for you!" <Cousin--Narrator>