What is literature? You'd think this would be a central question for literary theory, but in fact it has not seemed to matter very much.
These are the opening words of the second chapter, "What is Literature and Does it Matter?", in Jonathan Culler's Literary Theory: A Very Short Introduction (Oxford University Press, 1997). I will skip Culler's discussion of why this has been so, even though it is worth reading, and skip to his discussion of the question whether there are "any essential, distinguishing features that literary works share".
The modern sense of literature as "imaginative writing" is roughly 200 years old, even though we use the term to denote writings that are much older. (The Gilgamesh questions on this site, for example, are about texts that are up to 4,000 years old.) As a consequence, the term refers to many types of texts from a great diversity of cutures and time periods. This makes the concept very slippery, so Culler suggests that rather than asking "what is literature?" we should ask "what makes us treat something as literature?
Culler goes through a number of thought experiments to find out what to find what under what conditions we are more likely to treat a text as literary and concludes (page 25),
They [the thought experiments] suggest, first of all, that when language is removed from other contexts, detached from other purposes, it can be interpreted as literature (though it must possess some qualities that make it responsive to such interpretation). If literature is language decontextualized, cut off from other functions and purpose, it is also itself a context, which promotes or elicits special kinds of attention.
One of the assumptions that readers bring to texts is the "hyper-protected cooperative principle". When two people communicate with each other, each assumes that the other person is cooperating and that what they say is relevant to the conversation. This is the "cooperative principle". We also make this assumption about texts. However, when texts have gone through a process of selection (typically reviewing and publishing), we assume that the text is worth reading, thus making the cooperative principle "hyper-protected". This leads us to the assumption that features in a literary text, such as rhymes, metaphors and even apparent nonsense, are relevant at some level.
Sometimes it is the literary context (e.g. a text being somehow presented as "literature") that makes us treat a specific text as literature, sometimes it is specific features ("special ways of organizing language", p. 27) that make us treat it as literary. Neither the context alone nor the language alone are sufficient to turn a text into literature. Culler concludes his chapter on the definition of literature by describing five perspectives on literature (rather than features of literature) that theorists have used to describe the nature of literature, none of which, however, fully encompasses the other ones.
- "Literature as the 'foregrounding' of language" This is most obvious in poetry, which may draw attention to the language through sound patterns such as end rhyme, alliteration and assonance. See, for example, the poems Inversnaid and The Windhover by Gerard Manley Hopkins. However, prose can also do this, even though we don't always pay attention to it. This foregrounding of language is not sufficient for a text to be treated as literary; tongue-twisters, for example, are usually not considered literature.
- "Literature as the integration of language" Literature brings various elements of a text into a complex relationship, e.g. sound and meaning (e.g. by making specific words rhyme) or grammatical organisation and themes. However, Dwight D. Eisenhower's presidential campaign slogan I like Ike also creates a relationship between sound and meaning without being literary.
- "Literature as fiction" Literary works project a fictional world including a narrator and fictional events. Deictic references in a literary work don't refer to the real world but have meaning only in this fictional world. (For example, "this morning" in Hopkins's "The Windhover" is unrelated to the day on which I'm writing this Stack Exchange contribution.) The relationship between what the narrator (or any character) says and what the author thinks is a matter on interpretation.
- "Literature as aesthetic object" Unlike, for example, recipes or advertisements, literary works are separated from a practical context, and without this practical communicative goal they draw the reader's attention to the relationship between form and content. This is what makes a literary work an aesthetic object. Culler also cites Kant's concept of "purposiveness without a purpose": works of art "are made so that their parts will work together to some end", but that end is the work of art itself. When we examine a text as literature, we look at how its parts contribute to the effect of the whole.
- "Literature as intertextual or self-reflexive construct" Literary works exist between or among other texts that they reference, repeat or transform. For example, Alexander Pope's The Rape of the Lock contains elements that allude to or parody earlier epic poems. This is known as "intertextuality". In addition, literary works can also be read as at some level about literature itself; this is referred to as "self-reflexivity". For example, Gustave Flaubert's Madame Bovary can be read as a novel about how romantic literature (e.g. of the kind that Emma Bovary reads) and Flaubert's own novel itself make sense of the kind of experiences that Emma Bovary has. Note that intertextuality is not unique to literature or the arts; it even exists in advertising.
The five aspects listed above can be read as properties of literary works, or they can be read as the results of a specific kind of attention that we accord literary works. Culler points out that "[t]he qualities of literature can't be reduced either to objective properties or to consequences of ways of framing language" (page 35). This is because language resists the frames we try to impose on it when we read literary texts; we have to "work on it" for the analysis to work.
Literary theory in the 1980s and 1990s (Literary Theory: A Very Short Introduction was first published in 1997) did not focus much on the distinction between literary and non-literary texts. Instead, it focused on how it functioned (or was meant to function) in specific historical or ideological contexts. Literature has both been seen as a civilising institution (see e.g. Chris Baldick's The Social Mission of English Criticism 1848-1932) or as an instrument to increase acceptance of the current hegemony. Literature is also a paradoxical type of discourse because on the one hand, it relies on specific conventions while on the other hand constantly testing the limits of these conventions and going beyond them.
Culler finishes his chapter on the definition of literature by pointing out that the question "What is literature?" is not raised because people are concerned that they might mistake the message in a fortune cookie for a poem. The background to the question is that theory has highlighted the literariness of many types of texts that are usually not considered literary. The aim of discussing the definition of literature is "to promote what [critics and theorists] take to be the most pertinent critical methods and to dismiss methods that neglect the most basic and distinctive aspects of literature."
Many of the same properties mentioned by Jonathan Culler can also be found in *Terry Eagleton's Literary Theory: An Introduction (Blackwell, 1983). He cites the Russian formalists, who focused on "literary language as a deviation from the norm" (pages 2-4) but points out that literariness can also be found in non-literary texts (see Culler's first perspective). He points out we sometimes know something is literature because of its context (e.g. a work of fiction; page 5; see Culler's third perspective). Literature can also be seen as "non-pragmatic discourse" that is "taken to refer to a general states of affairs" (page 7). Literary works can "signal that what is at stake is a way of talking", i.e. it is self-referential language (page 8; see Culler's fifth perspective). He concludes that literature is not so much defined by an inherent set of properties but by "a number of ways in which people relate themselves to writing" (page 9).
Both Eagleton and Culler suggest that the concept of "literature" may be similar to "weed". "Weed" is not a type of plant that can be found in a biological taxonomy; it is a collective noun for plants that gardeners and farmers want to get rid of. In philosophical terms, both "literature" and "weed" are functional rather than ontological concepts: they are defined by what we do with them rather than by inherent properties. (This similarity to the concept of "weed" was suggested by John M. Ellis in The Theory of Literary Criticism: A Logical Analysis, 1974.) Which types of text count as "literary" has evolved over time: both Eagleton and Culler point out that our modern concept of literature is fairly recent and began to emerge during the Romantic period. What both theorists have done is teasing out some reasons why we, as descendants of the Romantics (at least in this regard), treat certain texts as literature and arguing that this concept is not set in stone.
Since Eagleton and Culler are both representatives of English-speaking academia, it seems worthwhile to compare their approach with a non-English one. The German textbook Französische Literaturwissenschaft. Eine Einführung (freely translated: "An Introduction to the Study of French Literature") by Maximilian Gröne and Frank Reiser (4th edition. Narr Francke Attempto, 2017) also discusses the definition of literature in the first chapter (pages 4-18). The authors discuss various characteristics that may make a text "literature":
- Poetic language deviates from everyday language; literary texts may deviate from ordinary language in order to foreground language. Gröne and Reiser point out that deciding whether a text's language is "poetic" or innovative becomes increasingly difficult as we read increasingly older texts. In addition, some literary texts don't deviate from ordinary language.
- Literary texts may be fictional, i.e. they present characters, events etcetera as not existing in the real world. (The authors draw the reader's attention to a distinction between "Fiktionalität" or fictionality on the one hand and "Fiktivität" or fictitiousness on the other. The characters in Voltaire's Candide, for example, are fictitious; the novella itself is fictional. The characters in a historical novel may be real but the novel presents actions, statements and thoughts that may never have occurred, so the novel is fictional. A study of Voltaire's Candide would contain many statements about the fictitious characters in the novalla, but that would not make this study fictional.)
Gröne and Reiser point out that it is not always easy to establish a text's fictionality; this requires knowledge from outside the text itself. The status of a text may even change over time. For example, the Old Testament's creation narrative was considered non-fiction for many centuries but is no longer generally considered as "true" today.
- However, even texts that use no poetic language and that don't appear as fictional may be read as literary texts. As an example, Gröne and Reiser cite the enumeration of tools and materials for decorating a room from George Perec's novel La Vie mode d'emploi / Life: A User's Manual. We read Perec's enumeration as a literary text because the text is clearly removed from pragmatic discourse ("entpragmatisiert", "défonctionnalisé").
The last point leads to a discussion of factors that make us treat a text as literary. One factor is that the context in which we encounter these texts presents them as literary, for example, by means of a bookcover that explicitly identifies a book as a novel (this is common in the German-speaking world, where the German world for novel—"Roman"—is usually printed on a novel's front cover) or by identifying the text as having been written by a specific author. Gröne and Reiser admit that "literature" is a category with vague boundaries and that the characteristics discussed above don't result in a definition that remains valid across time and across cultures.