Collaborative Annotation as a Teaching Tool

The guidelines presented here were developed in a seminar aimed at M.A. and advanced B.A. students. They are based on narratological theories by Marie-Laure Ryan, Gérard Genette, William Nelles, Ansgar Nünning, John Pier, and Viveca Füredy. Our contribution focuses on how collaborative annotation tasks can be used in university seminars, especially in the context of teaching students how to critically assess and compare theoretical frameworks and definitions. We also highlight the students’ impression that developing and using annotating guidelines improved their close-reading skills and that the task sensitised them to some of the core challenges of distant reading (e.g. questions of ambiguity and interpretation).

The in-class discussions soon drew our attention to fundamental problems that arise when trying to transform vague or even contradictory narratological theories into unambiguous, widely applicable annotation categories. The first issue was the definition of narrative itself. In particular, when does a dialogue, which is part of a narrative, become a narrative of its own? For example, is the statement "I went to the supermarket and bought some fruit" already a narrative? As a simple working definition we decided to choose "a report of connected events" 1 . This is important because, for example, Marie-Laure Ryan has an even wider definition, 2 which leads, as we think, to obscuring matters by a proliferation of narratives. The example, however, indicates a wider problem: there needs to be a clearly defined research question before starting to define and annotate narrative levels. For example, when one wants to find out whether novels from the eighteenth century tend to have more embedded narratives than twentieth-century novels, using annotation guidelines that are primarily based on Ryan's theory (see Change of narrative levels (Genette) and Change of narrated worlds below) might distort one's results because the crossing of an illocutionary or an ontological boundary does not necessarily establish an embedded narrative. Hence, even within the field of embedded narrative, there is no such thing as a 'universally marked-up text' that has to be annotated once and then can be re-used for many different research purposes.
The discussion of Frankenstein alerted us to another problem, namely the question of who, actually, is the narrator in a given passage: in the novel, Walton does not hear the Creature relate its own story; instead, it is filtered through Frankenstein. Who, then, is the narrator of the passages concerning the early life of the Creature? The Creature who related them to Victor, Victor who tells them to Walton (and maybe slightly manipulates them), or Walton who writes them down (and maybe does not transcribe Victor's tale verbatim)? For the sake of simplicity, we decided to go for the original source and assumed that the Creature is the most relevant narrator of its own tale. The aspect of time (e.g. whether a certain part of the narrative occurs in a prolepsis) also had to be discarded since otherwise our guidelines would have become too complex. Furthermore, when contemplating how to annotate two narrative levels that describe different worlds, we decided not to use separate tags for dreams, beliefs, delusions, and the like. This would have led to a proliferation of tags and would have made annotation too dependent on the interpretation of the text (e.g. we sometimes cannot be sure whether a character is hallucinating/dreaming or not). Instead, according to our guidelines, annotators need only indicate whether the world depicted in the narration of a lower level is factually dependent on the world depicted in the higher level or not (see Change of narrated worlds below). We also agreed that it would be helpful to annotate whether a narrative on a lower level is embedded in, or framed by, the narrative of the higher level. (For the theoretical background see Embedding vs framing narrative below.) The problem was, again, one of drawing a clear line between framing and embedding. For example, when the narrated passage on the lower level is just as long as the narrated passage on the higher level, is the former embedded in, or framed by, the latter?
Hence, in our systematization of narrative levels we focused on the features that define narratives within narratives: the narrator (position) (see Narrator's position and part in the narrative (Genette)), the narratee (see Narratee (Nelles)) and the (in)dependence of the narrated world (see Change of narrated worlds). We furthermore determined whether or not the narrative within a narrative is (quantitatively) the main narrative of the whole text and if it is fully enclosed (see Embedding vs framing narrative and Opened vs closed narratives). Last but not least, we took into account if the boundary between narrative levels is strictly observed or if there are cases in which, although we may notice a separate level of narration in some respects, the boundary is transcended in others (see The nature of the boundary between the levels (Füredy)).
A question that came up time and again during our discussions was which aspects our guidelines should cover in the first place. We might try to only annotate features that can be identified without much prior interpretation, but this would mean to exclude exactly those issues that make literary analysis so intriguing. The students also wondered whether it is possible to develop guidelines that can be used for all literary texts. When we annotated the short texts provided by the organisers of SANTA, we soon realised that some of the phenomena that we included in our guidelines were not to be found in these texts, whereas some features that we identified in the texts were not covered by our guidelines. Hence, developing guidelines that are too specifically tailored to one text or genre will make the guidelines useless for analysing other texts, but when the guidelines are too general, they tend not to yield interesting results.
During our in-class discussions, it became clear to what extent annotation depends on definitions and interpretations. Students pointed out that, in the future, they would never rely on studies based on corpora without first considering the guide-lines that were used to annotate them. Even though many of them were critical as to the applicability of annotation for their purposes as literary scholars, they appreciated the development and use of annotation guidelines as a tool for close reading: rather than let an ambiguous text stay ambiguous, they simply had to decide for one option in order to be able to annotate a passage and had to justify their choice with reference to the whole text or to adapt the guidelines in order to address and document the ambiguity. Likewise, they had to precisely identify the location of changes (e.g. of level or narratee) in the text. Students also liked the idea of creating guidelines that were to be used by others as it provided a welcome contrast to writing term papers that no one but their lecturer would read. However, they would have appreciated to get the guidelines and annotated texts of all other participants and to receive feedback on their own guidelines (either by the organisers or by the participants who used them to annotate).
The biggest problem was that it was not really clear which research question the guidelines were designed to tackle. Depending on this, we could have shifted the focus of our guidelines by adding or omitting certain categories. Overall, our students enjoyed the SANTA competition because it enabled them to practice their close reading skills as well as to learn and critically evaluate a new method of conducting literary studies.

Theoretical Explanation
Change of narrative levels A threshold between one narrative level and another: according to Genette, strictly speaking, only a second narrative (metadiegetic level) within the first one (the intradiegetic one). 3 1. Level within the global text at which the telling of the narrator-characters' story occurs 2. Level at which the primary narrator's discourse occurs 3. Level outside of the narrative act situated outside the primary narrator's discourse

Borders
When annotating narrative_level, please also include the punctuation marks that appear immediately before and after the first and last words occurring in this level (e.g. quotation marks or full stops) in your annotation.
When annotating level_change and narrator_change, please locate the tag between the two passages that differ in level or narrator.

What Does Not Belong Here?
Distinction between homodiegetic, heterodiegetic, and autodiegetic narrator. (For this, see below.)

Frequent Markers & Test
level_change and narrative_level: It is the overall goal of our guidelines to identify changes of narrative levels. Hence, all criteria discussed here and below (change of narrator, change of narratee, or change of the narrated world) can be indicators of a level change. A change in time can also be an indicator.
narrator_change: Often, the change of narrator is signaled in the text itself. E.g. "And then he began to tell his story…" or "In her letter she wrote the following…" The change of the position of the narrator with respect to the story or the change of the participation of the narrator in the story can also be indicators (see below, Narrator's position and part in the narrative (Genette)).

Examples
(1) LEVEL CHANGE: </level_change value=no > So strange an accident has happened to us, that I cannot forbear recording it, although it is very probable that you will see me before these papers can come to your possession.
[. . . ] </level_change value = yes > It is with considerable difficulty that I remember the original aera of my being: all the events of that period appear confused and indistinct.
(2) NARRATIVE LEVELS: <narrative_level number =2 > This manuscript will doubtless afford you the greatest pleasure: but to me, who know him, and who hear it from his own lips, with what interest and sympathy shall I read it in some future day! </narrative level> <narrative_level number = 1a > I am by birth a Genevese; and my family is one of the most distinguished of that republic. My ancestors had been for many years' counsellors and syndics; and my father had filled several public situations with honour and reputation. </narrative_level> <narrative_level number = 1b > I lay on my straw, but I could not sleep. I thought of the occurrences of the day. What chiefly struck me was the gentle manners of these people; and I longed to join them, but dared not. </narrative_level> (3) NARRATOR CHANGE: </narrator_change value= yes > It is with considerable difficulty that I remember the original aera of my being: all the events of that period appear confused and indistinct. </narrator_change value="no"> Nothing is more painful to the human mind, than, after the feelings have been worked up by a quick succession of events, the dead calmness of inaction and certainty which follows, and deprives the soul both of hope and fear.

Narrator's Position and Part in the Narrative (Genette)
Theoretical Explanation

Borders
Please include the punctuation that encloses a narrator's story when annotating this narrator's story. Example: And then he began to tell his story. <narrator participation="autodiegetic"> <narrator position="intradiegetic"> "When I was a little boy…" </narrator>.

What Does Not Belong Here?
For the sake of simplicity, we do not annotate the narrator's focalization.
Pronouns are a good indicator for determining whether a narrator is heterodiegetic or homodiegetic. The presence of homodiegetic and autodiegetic narrators is often explicitly marked; they are usually overt narrators. Heterodiegetic narrators are often (but not always) covert narrators.

Example
(4) (Beginning of Chapter 7 of Frankenstein) <narrator participation "homodiegetic narrator"> On my return, I found the following letter from my father: </narrator> < narrator position "intradiegetic narrator"> "My dear Victor, "You have probably waited impatiently for a letter to fix the date of your return to us; and I was at first tempted to write only a few lines, merely mentioning the day on which I should expect you. But that would be a cruel kindness, and I dare not do it. What would be your surprise, my son, when you expected a happy and glad welcome, to behold, on the contrary, tears and wretchedness? </narrator>

Theoretical Explanation
We have included this category since sometimes narrative levels are only to be distinguished by a change of narratee. In other words, the narrator may remain the same, and the narrated world (see Change of narrated worlds below) may remain the same, but the person to whom the story is told may become a different one. (E.g. when the autodiegetic narrator of the first-level narrative tells a story to a specific person within that narrative.) 7 There are two possibilities:

Borders
The tag occurs at the point between two passages that are told to two different narratees, i.e. it annotates the blank between text rather than the text itself.

What Does Not Belong Here?
Different narratees that only differ in that they are different persons on the same narrative level, in the same time, in the same world, etc. are not annotated as "changed narratees". E.g. when an autodiegetic narrator directly addresses different 'readers' and says "You, Sir, will probably not believe my story, but you, Madam, will certainly trust me when I say [. . . ]", 'Sir' and 'Madam are not counted as different narratees.
A change in narratee is usually signaled by mentioning the name of the new narratee or by any other expression that makes clear who the embedded narrative is directed at.

Theoretical Explanation
We have included this category since it is a key to providing significant information about the relation of the different narratives to each other: do they depend on each other or are they fictions within fictions? Just as fictional texts are counterfactually independent of the actual world 8 , second-level narratives may be counterfactually independent of the world of the first-level narrative. Examples are inserted narratives (as in the Decamerone or the Canterbury Tales).
Ryan describes in her theory the crossing of boundaries, either illocutionary or ontological. An ontological crossing of boundaries refers to a change of reality. These kinds of reality shifts affect the narratological structure and are therefore important for our guidelines. A shift of reality occurs when narratives refer to two different worlds that are not dependent on each other.
Our category of narrated worlds is similar but not identical with Ryan's "ontological boundary", 9 which is, however, not strictly logical and therefore impracticable.
In the case of narrated dreams it may sometimes be difficult to decide if there is a change of worlds, but even though in dream worlds different physical laws might apply, the dream world is dependent on the narrative world, either due to influence of the experiences of the dreamer or due to their prophetic character. This is why we recommend tagging dreams, as a rule, as "same world".

Span
The tag occurs at the point between two passages that are told about two different realities, i.e. it annotates the blank between text rather than the text itself.

Borders
The tag occurs at the point between two passages that are told about two different realities, i.e. it annotates the blank between text rather than the text itself.

Frequent Markers Test
To test whether an embedded narrative is counterfactually dependent or independent of its frame narrative, one can simply ask: "If aspect X changed in the reality of the frame narrative, would aspect X also have to change in the embedded narrative?" For example, a strange fairy tale in an embedded narrative is a change in reality (i.e. counterfactually independent of the 'real world' depicted in the frame narrative) because it does not have to adhere to the rules of the reality of the frame narrative. (E.g. in the frame narrative gravity exists and elephants cannot talk but in the embedded narrative gravity does not exist and elephants can talk.) (5) <narrative_level number= "1"> This manuscript will doubtless afford you the greatest pleasure: but to me, who know him, and who hear it from his own lips, with what interest and sympathy shall I read it in some future day! </narrative level> </change_narrator value="yes"> </change_narratee value="yes"> </change_reality value="no"> <narrative_level number "2"> I AM by birth a Genevese; and my family is one of the most distinguished of that republic. My ancestors had been for many years counsellors and syndics; and my father had filled several public situations with honour and reputation.

The nature of the level-change structure: Embedding vs framing narrative
Theoretical Explanation which can be used to compare the length of levels without using the ambiguous term "framing" and "embedding".
2. whenever there is a level change in the text, which should be tagged following the instruction in "Change of narrative levels (Genette)", read the following guidelines to add the information of level length.
When counting the words of "level n", first count the number of words "Ln" between the tag <narrative_level number="n"> and the first end of tag </narrative_level number="n"> after it (so that you do not count any other parallel level n that does not belong to the same narrative).
If there is no "level n+1" within "level n", L=Ln.

Span
The annotation can span a word, sentence, passage, chapter, or a whole text.

Borders
Punctuation marks are here not counted as words, but the punctuation immediately before and after the first and last words of the annotated passage is also included in the annotation. In order to make the automatized counting of words easier, contractions like "isn't" or "they're" are counted as one word.

What Does Not Belong Here?
Titles, chapter headings, or announcements like "The End" are not counted.

Theoretical Explanation
Both framing and embedding mentioned in Embedding vs framing narrative can have three kinds of structures concerning if they are complete: opened and closed, opened but never closed, and closed but never opened.

Category Attributes Possible Values narrative_levelchange completion Complete
Never closed Never opened

Span
The annotation can span a word, sentence, passage, chapter, or a whole text.

Borders
The punctuation immediately before and after the first and last words of the annotated passage is also included in the annotation.

Theoretical Explanation
This category is optional and should only be applied if there is at least one metalepsis that can be clearly identified in a text. 12

Strictly observed
Strict boundary between narrative levels. (<boundary transgression="no">) EXPLANATION: This category is applied when the boundary between narrative levels is respected. Strictly put: it is applied when a metalepsis does not occur and therefore cannot be applied.
EXPLANATION: a metalepsis is identified according to Genette's terminology. 13 Therefore, this category is only applied in instances were a transition between narrative levels can be identified and only if the following condition is fulfilled: any intrusion by a narrator or narratee from outside of the particular narrative level that transgresses its internal logic. This can occur when authors (or their readers) introduce themselves into the fictive action of the narrative, or when a character in a narrative intrudes into the narrative level of the author (or reader). Such intrusions disturb the distinction between levels.

Pseudo-diegetic narration
cf. Genette: second-level narrative told as first-level narrative 14 "a narrative second in origin but which, lacking a diegetic relay, is narrated as though it were diegetic". 15 This often means that we do not know who exactly is narrating in a passage.

Category Attributes Possible Values Boundary Transgression No
Metalepsis Pseudo

Span
The annotation can span a word, sentence, passage, chapter, or a whole text.