Interpretation is widely regarded as the core activity of literary studies. Still, the appropriate balance between the plurality and the limitation of possible interpretations is a non-trivial issue. Whereas it is sensible to accept that literary texts can generally have various meanings, it should not be possible to attribute any kind of meaning to a text. Therefore, while interpreters must be allowed to disagree in their analyses, it must at the same time be possible to review whether a disagreement is actually based on adequate reasons like, for example, textual ambiguity or polyvalence. In this paper, we propose a best practice model as one effective means to review disagreement in accordance with literary studies principles. The model has been developed during the collaborative, computer-assisted annotation of literary texts in a project in which short stories have been analyzed narratologically. The examination of inconsistently annotated text passages revealed four types of reasons for disagreement: misinterpretations, deficient definitions of the categories of analysis, dependencies of the relevant categories on preliminary analyses, and textual ambiguity/polyvalence. We argue that only disagreements based on textual ambiguity are considered legitimate or valuable cases of disagreement, whereas the other three types of disagreement can be resolved in a systematic way.

This paper investigates the question of whether it is possible to talk about aspects of the meaning of literary texts in a context-free manner. Its starting point is a detected discrepancy between the assumption that some (not purely formal or quantitative) approaches to literature operate in a context-independent manner, and the thought that processes of understanding are necessarily interpretive and/or context-dependent. The exemplary field of investigation is (structuralist) narratology, which is often said to be a »context-free« approach to literature.To determine whether narratology actually is context-independent, I first offer an explicative definition of ›context‹ applicable to the field of literary studies, based on aspects of the meaning of ›context‹ in everyday use. According to this definition, ›context‹ in literary studies is to be understood as a sum of additional extra-textual facts that may be consulted in order to foster the understanding of a text. This definition implies that neither the text itself nor any of its parts may be properly regarded as possible contexts for a given text; consequently, neither the sentences which a text consists of nor the propositions these sentences express are possible contexts of that text. A more general upshot is that any reference to the propositions expressed by the sentences of a text qualifies as a context-free approach if said propositions can be accessed without drawing upon contexts. The question of whether this is possible is subsequently investigated with the help of a three-stage model of grasping linguistic meaning, fit to analyze the processes which are involved when understanding linguistic utterances. As it turns out, linguistic meaning is indeed accessible without drawing upon contexts in many, but not all cases.The next step of the investigation involves a close examination of the application conditions of two types of narratological categories: categories for the analysis of the discours of a narrative, i. e. the presentation of a story, and categories for the analysis of the histoire, i. e. the elements of the story itself.Two major results emerge from this investigation. First, as it turns out, the question of whether an approach to literature requires reference to contexts should be distinguished from, on the one hand, the question of whether an approach is interpretive, and, on the other, the question of whether an approach puts forward a theory of »work meaning«. For while questions about whether an approach is contextual are determined by whether additional »input material« is used in order to foster the understanding of a text, whether the approach is interpretive con-

No abstract

The narratological concept of unreliable narration is subject to constant debate. While this debate affects different kinds of problems associated with unreliability, one of the central issues concerns the application area of ›unreliable narration‹. Here, theorists discuss, for example, whether there are certain types of narrators that cannot be unreliable, whether some kinds of narrators are necessarily unreliable, or in which way other characters apart from narrators can also be unreliable. It is the first one of these questions that I am addressing in this paper: Are there types of narrators that cannot be unreliable? As I lay out in the first section of my paper, my argumentative starting point is the observation that previous contributions to the application area discussion neglect two basic theoretical distinctions that are necessary to find robust and detailed answers to the relevant questions. The first of these theoretical distinctions will be addressed in the second section of the paper. It concerns the narrative phenomena that are usually referred to as »unreliable narration«. As I will argue, these phenomena are very heterogeneous, and we must distinguish at least five basic types of unreliability whose application areas partially differ:(1) fact-related utterance unreliability: the narrator’s claims about story world facts are false or in a relevant sense incomplete,(2) fact-related cognitive unreliability: the narrator’s beliefs about story world facts are false or in a relevant sense incomplete,(3) value-related utterance unreliability: the narrator’s evaluative utterances are in conflict with a relevant value system,(4) value-related cognitive unreliability: the narrator’s evaluative opinions are in conflict with a relevant value system, and(5) value-related actional unreliability: the narrator’s actions are in conflict with a relevant value system. In the third section of the paper, I will then proceed to show that four kinds of narrator types have been conflated or confused in the application area debate:(a) heterodiegetic narrators: narrators who are not part of the narrated story world,(b) non-personal narrators: narrators of whom we know no features apart from them telling a story, or narrators whom we are not invited to picture,(c) all-knowing narrators: narrators who have complete knowledge of the story world facts, and(d) stipulating narrators: narrators who generate the story world facts by narrating them. In discussions concerning the question of whether one or more of these narrator types cannot be unreliable, some theorists seem to assume that some or all of these types are necessarily connected. I will show, however, that there are hardly any necessary connections between them. After this preparatory work, I am showing in a step-by-step analysis in section four which of these narrators types can or cannot be unreliable in which way – and why. The results are as follows: Both heterodiegetic and stipulating narrators can be unreliable in all of the five ways outlined in section two. This outcome may seem surprising for the case of stipulating narrators. It becomes more comprehensible, however, if we bear in mind that only fact-related utterance unreliability is really impugned by a narrator’s ability to create facts by narrating them – and even here we can find a case where unreliability is very likely possible: the case of narratorial self-correction. All-knowing narrators, however, can only be unreliable in four of the five ways: It is, for conceptual reasons, impossible that all-knowing narrators are unreliable on the cognitive level with regards to the story world facts. Since they have complete knowledge of the story world facts, they cannot be wrong or ignorant about them. The case of non-personal narrators, finally, is the most complex. Here, it may first seem that non-personal narrators can never be unreliable – because as soon as a narrator is unreliable, we would know one significant feature of theirs, namely their being unreliable, which makes them personal. However, I will argue that, according to one reading of the non-personality concept, this type of narrator can in fact be unreliable on the utterance level both with regards to facts and values. This is because neither two conflicting reports by the same narrator nor the occurrence of problematic evaluative utterances in a narration – while often being sufficient for fact-related or value-related utterance unreliability respectively – necessarily invite us to picture a narrator. I am closing my paper in section five by summarizing the results and pointing to some possibly debatable theoretical assumptions on which my analyses are based.