What narrative is

1 Descriptive narratology, the diversity constraint and the centrality of categorization disputes

The subtitle of this text is inspired by P. F. Strawson’s 1959 book Individuals. An Essay in Descriptive Metaphysics in which Strawson (1996: 9) argues that traditional metaphysics is often revisionary rather than descriptive, inadvertently prescribing how to think about the objects in the world rather than describing how we actually think about them. I believe the same criticism applies to narratology. Many narratologists who seemed to believe that they describe what narrative is or how narrative works, have inadvertently offered revisionary accounts. They did so, among other things, because they failed to sufficiently assess the extension of expressions such as “narrative” or “telling a story” in ordinary language. Descriptive narratology can be defined as an approach that seeks to describe rather than prescribe how people think about (aspects of) narrative(s). For instance, the paradigm of narrative most narratologists operate with is still word-based, while most people would agree that there are many narratives that are not. Even though Roland Barthes (2005: 109), Claude Bremond (1964), Tzvetan Todorov (1969: 10) and many others have long argued that not all narratives are verbal and despite the “transmedia turn” in narratology, presentations on films and pictures were rare at the ENN 2017 in Prague (less than 10 % each) and less than 5 % treated wordless music. This might be linked to the fact that the existence of mimetic narrative^[1] and narrative in wordless music is still under dispute. However, here like elsewhere good reasons might not be the most influential factor: As most narratologists first learn about narratives through texts (Hausken 2004: 394), they might easily fall prey to a cognitive bias which Liv Hausken calls “media blindness”. As Hausken (2004: 392) explains, “theories that are seemingly independent of the medium are usually implicitly tied to a particular medium”. While narratology has discovered video games, newspaper articles, oral utterances and abstract comics as objects of study, its paradigm cases are still complex, verbal and often long literary narratives. This may lead narratologists to consciously or unconsciously hold that the specificities of such verbal texts – for instance double timeline – characterize narrative tout court and cause problems in categorizing less paradigmatic productions.

Even if most theorists have an opinion about what a narrative is, explicit categorization disputes only engage a handful of them. Most scholars rather focus on developing general theories or specific concepts and on analyzing paradigmatic creations. Let me start by explaining why this disinterest in categorization can be problematic. General theories need to respect the “diversity constraint” (Lopes 1996: 32): they must be general enough to account for all entities of the kind under consideration. A theory of pictures needs to account for a child’s drawing, a cubist painting as well as a trompe-l’oeil or a photographic snapshot. Likewise, if we want to be sure that our theories of non-mimetic narratives, unnatural narratives, unreliable narration, point of view, suspense and curiosity, etc. do not only apply to a subset of narratives, for instance verbal ones, we need to know where narratives and narrativity occur. An analogue example may further clarify why correct categorization matters for interpretation.

When we define a game as “A system in which players engage in an artificial conflict, defined by rules, that results in a quantifiable outcome” (Salen and Zimmerman 2003: 80), cooperative activities are excluded from the realm of games. But are badminton or volleyball not games when participants try to make it easy for their partners to hit the shuttle or ball? Our intuition seems to tell us that they are. If this intuition turned out to be stable across subjects, we could want to replace (a) “artificial conflict” by something else to account for what we can call “folk categorization dispositions”, perhaps (a') “artificial conflict or cooperation” or (a'') “artificial interaction”. Such a change will not only lead to a definition that adequately categorizes badminton and volleyball in the park and other cooperative games as games. It will have much wider implications. Making (a') or (a'') instead of (a) a central element in a definition of games implies a different understanding of what a game is and why people play games. Therefore it modifies what is considered salient in gameplay, even when we analyze games dominated by competition. Based on an understanding of games that integrates (b'), scholars of game studies will naturally start to account for cooperative elements, even in mainly competitive games. With a definition integrating (b''), which remains unspecific as to the nature of the interaction, all kinds of exchange between players can be seen as important to the game as game. Perhaps competition and collaboration are only some of the interactions that are essential. Other forms of exchange, such as hindrance or encouragement might suddenly appear to be elements of gameplay, perhaps even when they do not appear in the rules^[2]. Ultimately, theorists could then relativize the importance of rules for games and develop a new theory of gaming. If discrepancies between folk categorizations and expert theories are taken seriously – and they certainly should be – they can thus have a strong impact on a discipline. This shows not only how important definitions are, but, more importantly, how definitions are important: When definitions are modified to include non-paradigmatic entities, the ways we analyze paradigmatic ones changes too. Influencing which aspects of the entities at stake appear as salient, definitions guide interpretations and can ultimately affect general theories. Therefore categorization problems are to be taken seriously, even by scholars who see their task as analyzing prototypical creations, such as verbal narratives in narratology (or competitive games in game theory).

2 Definitions and intuitions

A major reason for theorists to want to define important concepts of their domain narrowly has been that they wanted to avoid that “anything goes” (cf. Revaz 1997: 69). As Shlomith Rimmon-Kenan (2006: 17) has put it: “if everything is narrative, nothing is”. But this underestimates people’s capacity to intuitively categorize ordinary things, even when they cannot provide criteria. As Wittgenstein (cf. 2009: § 68–71, § 76–77) argued, the fact that we are not able to explicitly state the criteria for applying a concept doesn’t fundamentally impair our use of it: “For how is the concept of a game bounded? What still counts as a game, and what no longer does? Can you say where the boundaries are? No. You can draw some, for there aren’t any drawn yet. (But this never bothered you before when you used the word ‘game’ [emphasis mine].)” (Wittgenstein 2009: § 68). In many cases, we are unable to clearly define a concept, yet we still know when it applies and when it does not. Arguably, the purpose of defining an ordinary concept such as game, narrative or story, is to provide explicit criteria for judgments we are already able to make (Schöttler 2016: 164)^[3]: “definitions need to stand the test of intuitions, not the opposite.” (Speidel 2013 a: 179). If definitions oppose intuitions, they lead to revision, not description. The definition of narrative as “the representation of actions or events” is a textbook case of revisionary narratology because it counter-intuitively turns recipes and weather forecasts into narratives. When Tzvetan Todorov (1973, 44) declared that recipes are “imperative narratives” to save his definition, this was certainly witty, but also quite ad hoc and unconvincing. The fact is that it proved notoriously difficult to devise a definition of narrative that does not misclassify recipes and weather forecasts or exclude real narratives. At the same time, it was always easy for anyone to intuitively understand that Cinderella tells a story and that the weather forecast does not (cf. Prince 1999: 43). Intuitive categorizations by ordinary speakers should therefore guide us when we design and assess definitions. How do we learn about intuitions? By checking if definition-based categorizations match intuition-based ones. As Stephen Laurence and Eric Margolis explain:

Forming an intuition is a matter of determining how we would categorize things in a given situation. Thus intuitions correlate with categorization dispositions and categorization dispositions correlate with content. The result is that intuitions are broadly correlated with content on virtually any theory of content. For this reason, they can be used as evidence for content, that is, evidence that a concept truly applies when intuition says it does. (Laurence and Margolis 2003: 279)

But given the reliability of intuitions, why do we need definitions at all? To put a first answer briefly: Intuitions inform us how to categorize this or that, but not why; they tell us which artifacts are narratives, but they cannot tell us what a narrative is. Intuitions can hardly be discussed, definitions can. Therefore we would not be doing science any more – at least not in the sense where most of us understand it today – if we only relied on intuitions. To understand what something is (and how it works), we need more than the ability to reliably categorize different productions (cf. Chatman 1983: 55). This is where definitions of ordinary concepts have a role to play. They provide reasons for categorizations we already know how to make. To conclude this section, we can say that intuitions not definitions are essential when we want to settle categorization disputes. Definitions certainly play important roles in research, but – as opposed to a common misconception and the ensuing misuse – being gatekeepers for a specific subject domain should not be one of them.

3 Major positions concerning pictorial narrative and the role of examples

In order to evaluate different ways to tackle categorization disputes and explore a new way to deal with them in concreto, I here want to look more closely at the debate concerning pictorial narrative. But the approach I describe would also be relevant to assess the narrative potential of other marginal cases, such as, for instance, drama, wordless music or abstract comics. Since Lessing (1853) first distinguished the spatial and the temporal arts in 1766, suggesting that the former should only – but often misinterpreted as saying that they could only (Speidel 2013a : 181–184) – be used to depict states, the central question concerning pictorial narrative has been: Can some single pictures be said to autonomously convey stories – and thus be classified as narratives (Wolf 2003: 180) – and if yes, which kind of pictures? Wendy Steiner (1988, 2004), Aron Kibédi Varga (1989, 1990), Jean-Marie Schaeffer (2001), Werner Wolf (2002, 2003, 2011), Michael Ranta (2013), Klaus Speidel (2013a, 2013 b, 2018a), Peter Hühn (2015) and Tobias Schöttler (2016) have addressed this problem within a narratological framework. They all develop a compatibilist argument, arguing that there are single pictures that autonomously convey stories. Looking more closely, we can however distinguish different kinds of compatibilism and the accompanying incompatibilisms. Monochrony – or strong – compatibilists like Ranta (2013), Speidel (2013a, 2013 b, 2018a) and Hühn (2015) believe that single pictures only explicitly depicting one moment in time can autonomously convey stories. This is something monochrony incompatibilists like Steiner (2004: 172), Wolf (2002: 96), Kibédi Varga (1990: 97) and Schöttler (2016: 178) deny. These authors believe that only works that explicitly depict several moments in time, for instance by repeating figures or by juxtaposing several pictures, can communicate stories^[4]. Let us call them polychrony – or medium – compatibilists. As opposed to them, Seymour Chatman (1990), Françoise Revaz (1997), Shlomith Rimmon-Kenan (2006) and Jean-Marie Schaeffer (2001) are incompatibilists concerning narrative in the single picture. They embrace sequence – or weak – compatibilism, i. e. the belief that only picture series and films can convey stories. Sequence (or weak) compatibilism and polychrony compatibilism are probably the most common positions in narratology. Full incompatibilism concerning pictorial narrative, the position denying that there is any intersection between the set of pictures and the set of narratives is relatively rare, but was prominently held by Gérard Genette (1980, 1983). More recently, Paul Barolsky (2010: 112) has also attempted to show that “there is no such thing as narrative art”.

4 Arguments for and against pictorial narrative

Any categorization question can be represented in terms of set theory. To say that pictorial narrative in monochronic pictures exists is, for instance, to affirm that there is an intersection between the sets of P and N, where P is the set of monochronic pictures and N is the set of narratives. There are two recurrent strategies to find out if there is an intersection: assessing if there is any conflict between essential criteria of the definitions of Ns and Ps. If there are none, Ps can be Ns, otherwise Ps cannot be Ns. Alternatively, it is possible to try to find at least one specific example that is both P and N. If there is a P that is also an N, this implies that Ps can be Ns. Like in other categorization disputes, both the first approach, where the compatibility of criteria is assessed, and the approach where we look for examples have been frequently used in arguments about pictorial narrative. The role and importance of examples in compatibilist and incompatibilist theories however differs greatly. Incompatibilists usually suggest that there is some kind of incompatibility between the concept of narrative as properly defined and what can be communicated with pictures. Pictures are then used as illustrations without much argumentative weight^[5]. As opposed to this, compatibilists usually make extended use of specific examples in their arguments. Given that incompatibilists deny that pictorial narrative of a certain kind exists, a single undisputed case of narrative in a specific kind of picture would be enough to prove that incompatibilism concerning this kind of picture is wrong.^[6]

Over the years, I have identified three major strategies to establish if single pictures can tell stories.

The argument from definition has largely dominated in narratology. Theorists argue based on definition alone or use definitions and examples, trying to show that some specific single picture corresponds to the criteria of a plausible definition of narrative.

The argument from intuition suggests that we should trust our intuition that single pictures communicate stories or that they do not.

The argument from expertise urges us to rely on expert judgment concerning the existence of pictorial narrative.

These strategies are often intertwined. Thus we, as experts, can reject or embrace definitions based on our intuitions concerning specific artifacts, accept definitions because they have been developed by experts in our field or look for expert confirmation of our intuitive assessments. To get a clear sense of the specific limits of each strategy, it is however useful to discuss them separately. After doing so, I will introduce a fourth strategy: the argument from ordinary language. It claims that definitions of terms like “telling a story” or “narrative” should not be incompatible with the ways they are used in ordinary language as assessed in valid experiments. I will then expose an experiment to find out what kind of pictures are naturally said to tell a story by non-experts, with the aim of settling the categorization dispute concerning pictorial narrative.

5 The argument from definition

Definitions of narrative or the criteria of such definitions usually play such a prominent role when pictorial narrative in single pictures is discussed that we might come to believe that checking if single pictures can correspond to a sound definition of narrative is the only available strategy. Thus researchers like Gérard Genette (1980, 1983), Seymour Chatman (1990), Françoise Revaz (1997, 2009), Emma Kafalenos (1996), Jean-Marie Schaeffer (2001), Marie-Laure Ryan (2004) and Shlomith Rimmon-Kenan (2006) have argued that (autonomous) narrative in the single picture does not exist because according to a plausibly narrow definition such a picture cannot convey what is necessary for narrative. Marie-Laure Ryan (2004: 139) argues that “since pictures, left by themselves, lack the ability to articulate specific propositions and to explicate causal relations, their principal narrative option is [...] the illustrative mode”.^[7] She considers that pictures cannot represent causal links between story events and therefore cannot convey a story autonomously. Seymour Chatman (1990: 7) and Jean-Marie Schaeffer (2001: 15) believe like Lessing (1766) that pictures cannot convey stories because as opposed to texts, they do not have double temporality, they do not govern the order of access to their elements^[8]. Werner Wolf (2002, 2003) and Wendy Steiner (2004) have argued from definition for polychrony compatibilism. I myself have also argued from definition on several occasions (Speidel 2013 a, Speidel 2013 b, Speidel 2018a), defending monochrony compatibilism. For the reasons given in sections I and 2, I have however come to consider the strategy from definition to be problematic when it is used to settle categorization disputes. Why shouldn’t we then argue from our intuitions?

6 The argument from intuition and its problems

The argument from intuition is based on the conviction that intuitive categorizations can be trusted. In an article where the narrativity of pictures is at stake, the philosopher Nelson Goodman argues for monochrony compatibilism based on his intuitive assessment of the narrativity of different kinds of pictures. The crux of his argument is his description of The Conversion of Saint Paul (1567), a monochronic picture by Pieter Brueghel, of which Goodman (1981: 333) affirms that “[it] tells a story and tells it so compellingly that we tend to forget that nothing in the picture literally moves, that no part of the picture precedes any other in time, and that what is explicitly shown is not actions taking place but a momentary state [emphasis mine]”. It is on the basis of this experience that Goodman states that even pictures which only depict one moment can tell stories. But is the intuition of a single expert that one monochronic painting tells a story sufficient to ground monochrony compatibilism? Many narratologists are skeptical... .

Aron Kibédi Varga (1989: 108) believes like Marie-Laure Ryan (2004: 139) that the evocation of familiar stories is the only option available for single monochronic pictures. Like Jean-Marie Schaeffer (2001: 12), he believes that a painting like The Conversion of Saint Paul merely reminds viewers of a verbal narrative they already know. If the pre-text is culturally well entrenched, it may well happen that a picture which only reminds spectators of a familiar story is experienced as communicating it. Goodman, who certainly knew the story of the conversion of Saint Paul, may be a case in point. Given that his monochrony compatibilism is based on his intuition that Brueghel’s Conversion of Saint Paul tells a story and this intuition is problematic, his whole argument falls apart. This case shows that intuitions are prone to be influenced by previous knowledge, some of which can be quite specialized, and are not to be trusted under all circumstances. Therefore intuitions we have alone in an armchair are not a sound basis for our arguments. One way out of the predicament of a single-handed intuitive judgment of one expert is to look at the assessment of an expert community.

7 The argument from expertise and its problems

The argument from expertise is based on the conviction that categorizations by domain experts can be trusted. Bence Nanay (2009) has thus argued for monochrony compatibilism concerning pictorial narrative based on the fact that art historians often consider it unproblematic that certain monochronic pictures tell stories. Like Göran Sonesson (1997), Nanay holds that the representation of a single action is sufficient for pictorial narrative^[9]. More precisely, Nanay’s theory of what he calls “narrative pictures” is based on the identification of “goal-directed actions” in representations (Nanay 2009: 125). In Nanay’s account, the representation of such actions comes quite close to a necessary and sufficient condition for narrative. However, this criterion is essentially backed by his appeal to expertise, namely “the way the term ‘narrative picture’ is used by art historians” (Nanay 2009: 123)^[10]. Art historians’ categorization dispositions then become the tortoise on which the elephant of Nanay’s definition of narrative stands^[11]. There are both specific and general reasons that make it relatively shaky ground. First of all, many theorists have called art historical expertise concerning matters of narrative into question. Wendy Steiner (1988: 2) notes that “the general art historical use of the term ‘narrative’ seems incomprehensible to literary scholars.” The art-historian Lorenzo Pericolo writes: “Art history, which has mostly emerged from the early modern debates about the istoria, has stubbornly continued to appraise visual narratives with criteria that are frankly obsolete” (Pericolo 2011: 94) and according to Werner Wolf (2002: 24), “questionable narrativisations [fragwürdige Narrativierungen]” are frequent in art history, and the use of “narrative” in most texts about art is “purely intuitive and accordingly vague.” Again, I do not believe that the use of intuitions is a problem per se, but I do believe that just like the intuitions of narratologists, those of art historians may be defective because of their professional background. Art history students are taught to quickly grasp which narrative, if any, a picture refers to. A naked couple with a snake is likely to evoke the Fall of Man not only to Christian audiences but also to most art historians and by just looking at a figure’s pose a well-trained art-historian can often tell what character and story is referenced. However, depicting a pose or a few clues and attributes does certainly not amount to autonomously conveying a story in the ordinary sense of the expression. While narratologists are more theory-conscious, we have already seen in section 1 that they are prone to media-blindness. This is likely to influence intuitions as much as definitions. Further more, both the intuitions and the language of art-historians and narratologists are likely to be theory-loaded.

As Thomas Kuhn (1996), Norwood Hanson (1958) and other theorists of science have shown, observation is influenced by the paradigms dominating a scientific community. As Hanson (1958: 19) puts it, “seeing is a ‘theory laden’ undertaking.” As expertise in most disciplines is linked to possessing theoretical knowledge and even holding theoretical beliefs, we may be particularly ill-advised to rely on expert intuitions concerning ordinary objects of inquiry. The way art historians or narratologists use phrases like “this painting tells a story” could be problematic precisely because they are experts. As Eugen Fischer recently explained, expert use of ordinary language may be particularly problematic:

Psycholinguistic findings reveal that some ordinary language is privileged – not normatively but psychologically: The uses of words a subject employs and encounters most frequently shape associative memory processes that duplicate semantic and pragmatic inferences [...]. In many cases, philosophers will use and encounter well established terms most frequently in certain familiar senses. When they give such terms a new use without realizing its novelty, they will hence unwittingly continue to make leaps of thought which duplicate semantic and pragmatic inferences licensed by the terms’ familiar use. (Fischer 2014: 132–133)

In other words: as philosophers, narratologists and art historians who specialize in narrative, we are particularly prone to using expressions like “telling a story” inconsistently because we cannot entirely forget the meaning of such terms in ordinary language while we theorize.

Given that experts can be affected by media-blindness, theory-leadenness and inconsistency, the approach which seems the most promising is working with non-experts, whose use of ordinary language is likely to be less biased. Working with a sufficiently high number of people should help compensate possible idiosyncrasies and the influence of theoretical knowledge some of them might have acquired.

8 Ordinary language as a basis for descriptive narratology

The idea that ordinary language use is a valuable indicator of the extension of concepts is quite old. P. F. Strawson’s (1996: 9) descriptive metaphysics, which we evoked above is based “upon a close examination of the actual use of words”, which he says to be the “the best, and indeed the only sure, way in philosophy”. Ludwig Wittgenstein (2009: § 43, § 92) and John Austin have also suggested that analyzing the ways ordinary people use words may help us to either solve or dissolve philosophical problems. Wittgenstein (2009: § 43) famously held that rather than abstract criteria, the “use in the language” is essential for the meaning of a word. The most articulate defense of this approach has been provided by Austin:

Our common stock of words embodies all the distinctions men have found worth drawing and the connexions they have found worth marking, in the lifetime of many generations: these surely are likely to be more numerous, more sound, since they have stood up to the long test of the survival of the fittest, and more subtle, at least in all ordinary and reasonably practical matters, than any that you or I are likely to think up in our arm-chairs of an afternoon – the most favoured alternative method. (Austin 1961: 130)

According to this, we should rely on the meaning terms have in ordinary language rather than the special meaning they sometimes receive in theoretical contexts. What is often overlooked by both critics and advocates of Austin’s approach is the fact that he speaks of ordinary and reasonably practical matters only. This seems to be a reasonable restriction as non-expert categorizations are likely to be useless for subjects like molecular biology. Even the nature of being or the freedom of will might not be sufficiently ordinary for non-experts to have much to say about them. Narrative, however, is quite different in this respect. After all, “all classes, all human groups, have their narratives” (Barthes 2005: 109) and storytelling performs many functions in society. It is thus both ordinary and practical. Looking at when ordinary people use the expression “to tell a story” should then be a relatively trustworthy indicator of what kind of representations really tell (or communicate) stories. Given that representations that tell stories are usually considered to be narratives (Gaudreault 1988: 84; Wolf 2003: 180; Currie 2010: 6), such investigations also teach us about what narratives are^[12]. The information obtained can then be used to evaluate our definitions and the ways they are usually interpreted.

9 Narratives in single pictures: Designing an experiment

Based on what has been said so far, an interesting approach to pictorial narrative emerges: taking pictures of a certain kind, for instance monochronic pictures that depict goal-directed actions, and asking people with no particular scholarly expertise on either narrative or pictorial art if they think that they tell stories. If categorizations obtained with such experiments oppose those based on theory, the burden of proof then lies with the theorist.

A well-designed and well-conducted experiment on pictorial narrative will, first of all, check whether participants agree that certain kinds of pictures tell stories. As storytelling is likely to be a question of degrees rather than categories alone (Currie 2010: 34; Prince 2012: 26), Likert scales seem more adapted than simple yes/no answers, where certain thresholds, above which an answer counts as a “yes” can be defined.

We should also make sure that participants are not misled by previous knowledge of the story communicated by a picture, which could systematically falsify their intuitions. A simple way to avoid this pitfall is to choose pictures that communicate stories that the participants are not acquainted with. Pictures that convey new stories and do not have a verbal pre-text or pictures that communicate relatively unknown stories can be chosen^[13], where works that are unknown to participants are preferable. As the latter cannot be determined beforehand, the effect of familiarity on other judgments will have to be assessed in the experiment.^[14] By choosing relatively unknown pictures while also showing a sufficient number of pictures to a sufficient number of participants, we can hope to find at least a few cases where the picture and story is unfamiliar to most viewers.

If previous knowledge of the story that is evoked was a condition for pictorial narrative, as different theorists (e. g. Kibedi-Varga 1989; Schaeffer 2001) have argued, there should be a strong correlation between knowing a picture beforehand and considering that it communicates a story.

Given that general reflections have to be applicable to diverse sets (Lopes 1996: 32), pictures should be chosen to reflect the diversity of picture-making and narrativity. Viewers should see pictures from different times and contexts. However, of course, one example of a kind of picture, where viewers agree that it does convey a story, would be enough to prove wrong the corresponding incompatibilism. To verify if viewers apply an “every-picture-tells-a-story” rule, which might arguably call into question the reliability of their intuitions, at least one picture that belongs to a genre traditionally considered to be low in narrativity by art historians (landscape, portraiture, allegory) and with a level of narrativity that is expected to be lower according to narrative theory should be included for each viewer.

A study also ought to include polychronic pictures which explicitly depict several moments, for instance by showing the same actors more than once, because, as mentioned above, theorists like Aron Kibedi-Varga (1989, 1990), Wendy Steiner (2004) and Werner Wolf (2002, 2003, 2011) consider polychrony to be a necessary criterion for pictorial narrative. If participants implicitly embraced this concept of narrative, the polychronic pictures should turn out to be the only ones with high storytelling ratings.

10 Experimental evidence

10.2 Stimuli

Fifteen paintings served as the stimuli. Three contemporary street art paintings (i. e. Banksy, 2006, 2011; Dran, n.d.), three paintings from the XVII century (i. e. Greuze 1777; Hogarth 1742; c.1743), three Baroque art paintings (i. e. Domenichino c.1615, c.1617, c.1635), three history paintings from the XIV century (i. e. Lorenzetti 1341; Lorenzetti 1332, 1339), and three pages without words from picture books for children (i. e. Nordqvist 1991, 1997, 2008) were used. All materials used were in a digital form and the entire experiment was set up on Social Science Survey (Leiner 2014).

The twelve art paintings and the three pages from the children’s books were chosen for diversity, because they are commonly unknown to non-experts and according to major criteria commonly used to define narrative in narratology, interpreted in such a way that they are compatible with monochronic single pictures. They are i) representation of humanlike agents, ii) event representation, iii) conveying the timeline of story events, iv) extraordinariness/eventfulness of these events, v) conveying causal connections between the events. I expected all of these to contribute to narrativity. Based on Speidel (2013a), the representational timeline criterion that is regularly used by incompatibilists, e. g. Christian Metz, Shlomith Rimmon-Kenan or Jean-Marie Schaeffer, was excluded. If this was problematic, none of the 10 pictures in the medium and high narrativity groups should appear as telling a story to the participants. I should here mention a last point concerning definitions: they usually leave room for interpretation. There are thus two ways different narratologists might fail less paradigmatic narratives: their definitions might be too narrow or the way they interpret and apply the defining criteria might be too restrictive given the ordinary concept of narrative. As mentioned in section 5 above, criteria iii) and v) have previously been used to justify monochrony or full incompatibilism. I believe that the underlying interpretation is problematic (Speidel 2013 a, 2013 b, 2018a). I consider that monochronic pictures can convey causality and several events of a story by showing different states and events that form part of a causal continuum, where viewers spontaneously use story scripts and world knowledge to reconstruct the rest. According to my interpretation, Hogarth’s Bagnio Scene corresponds to criterion v) because it conveys a causal relationship between a man who is bleeding, a bloodstained sword laying on the ground, a second man fleeing through the window (and many other picture elements). This picture is not at odds with iii) because it conveys previous events through traces or by showing states that are so unstable that viewers will most certainly think of past and, possibly, future events (cf. Speidel 2018 c: 441–443).

Figure 1

Criteria influencing narrativity as used in experiment

Based on these criteria according to my liberal interpretation, degrees of narrativity were defined, namely higher, medium and lower (Figure 1). While all pictures showed human-like agents, only the higher and medium pictures made it possible to systematically reconstruct temporal and causal relationships and showed extraordinary events. The events in higher pictures were considered to be more extraordinary than those in medium pictures. Extraordinariness was used to separate medium from higher narrativity pictures because since Artistotle’s peripeteia, conceived as an unexpected turn (para doxan), so many definitions have insisted that for ideal narratives, there needs to be some kind of disruption in the chain of story events (Herman 2009: xvi; Todorov 1992: 121) or that at least one of the represented events has to appear as “significant” (Schmid 2010: 8) or “extraordinary” (Hühn 2009) in the storyworld.

Pictures of lower narrativity either showed states rather than events or multiple disconnected events. They also tended to show more ordinary situations, such as a market scene, people resting by a river, children posing for a portrait, or an old man contemplating objects.

To clarify how our interpretation of the criteria motivates our classifications, let us look at the XVIII century painting style (Figure 2) with paintings by William Hogarth and Jean-Baptiste Greuze. William Hogarth’s “Marriage à la Mode: The murder of the count” (c. 1743) was chosen as a picture higher in narrativity because it represents human agents and several events (such as a woman interacting with a man who has been hurt, armed men coming through the door, while another person is escaping through the window). In accordance with Speidel (2013a: 181, 189–190, 2018a), it was also predicted to cue interpreters to draw inferences about the temporal and the causal relationships of the depicted events. The depicted scene is obviously high in extraordinariness – in fact it is a crime scene. This was contrasted to “The ungrateful son” (Greuze 1777), which also represents agents and events, namely a dispute. However, its timeline seemed less obvious and the causal relations more difficult to perceive (Did the father first get angry and the son decide to leave because of this or the opposite? What is the role of the man by the door?). Given this and the fact that a dispute is less extraordinary than a crime, this painting was assumed to be medium in narrativity. The third picture in the group was The Graham Children (Hogarth 1742). Being a group portrait, it contains human agents, who interact somewhat, but has no clear timeline, shows few events, few strong causal relations and nothing that had to be considered as extraordinary by a non-expert public given the context. It was therefore predicted to be lower in narrativity, according to theory.

Figure 2

Pictures of style 2 (XVIII century) classified in order of narrativity from higher to lower

For each style of pictures, similar possibilities were evoked, and choices and assignments were made and discussed in groups of experts in art history and psychology at the CReA and EVA labs of the university of Vienna respectively, thus combining art historical and psychological expertise.

10.6 Results

“Predicted degree of narrativity” that reflects higher, medium and lower in narrativity pictures will be referred to as “narrativity” from hereon, the ratings by participants as “storytelling”.

In general, pictures that were higher and medium in narrativity received higher storytelling scores than pictures lower in narrativity across all styles of paintings (Table 1).

Table 1

Means (μ) by Narrativity

Degree	Higher	Medium	Lower
Mean μ (Total)	5.79	5.86	3.84
Standard Error	.096	.102	.129

The mean scores of higher, medium and lower in narrativity pictures were μ=5.79 (SE=0.96), μ=5.86 (SE=0.102) and μ=3.84 (SE=0.129), respectively.

Frequencies were calculated for “I can’t tell” responses of the storytelling question. Participants could not tell if a picture was narrative or not, most frequently, when the picture was of lower narrativity.

There were significant differences between the scores of higher, medium and lower narrative street art paintings at the 5 % level using the Huynh-Feldt correction (F_{(1.571,64.415)}= 3.866, p=0.035), among the scores of higher, medium and lower XVIII century paintings at the 1 % level using the Huynh-Feldt correction (F_{(1.672,70.213)}= 111.946, p<0.001, among the scores of higher, medium and lower Baroque paintings at the 1 % level (F_(2,80)= 32.292, p<0.001), among the scores of higher, medium and lower XIV century paintings at the 1 % level (F_(2,78)= 76.697, p<0.001) and among the scores of higher, medium and lower narrative pictures from children’s books at the 1 % level (F_(2,86)= 8.308, p=0.001).

For street art paintings, LSD comparisons revealed significant differences in the scores between pictures of higher and lower narrativity (p=0.022, Cohen’s d=0.239) only. In addition, Bonferroni post-hoc pairwise comparisons revealed significant differences between the scores of lower and higher XVIII century paintings (p<0.001, Cohen’s d=1.24) and the scores of lower and medium paintings of the same style (p<0.001, Cohen’s d=1.26). With regard to Baroque paintings, Bonferroni post-hoc comparisons revealed significant differences between the scores of lower and medium (p<0.001, d=0.819), lower and higher (p<0.001, Cohen’s d=0.527) and medium and higher paintings (p=0.005, Cohen’s d=0.469). In XIV century paintings, Bonferroni post-hoc comparisons revealed significant differences between the scores of lower and medium pictures (p<0.001, Cohen’s d=1.17) and lower and higher pictures (p<0.001, Cohen’s d=1.40). Finally, post-hoc Bonferroni pairwise comparisons revealed significant differences between lower and medium pictures (p=0.004, Cohen’s d=0.414) and between lower and higher pictures from children’s books (p=0.007, Cohen’s d=0.414).

Table 2

Mean narrativity scores across styles

11 Conclusion and future research

It has been argued that at least in the case of ordinary concepts, the validity of definitions should be assessed through intuitions as reflected in categorization dispositions. Definitions as well as expert intuitions and language use were shown to be problematic for categorizations because they are prone to media blindness (Hausken 2004), can be theory-loaded (Hanson 1965: 54–62) and can mesh-up ordinary and theoretical meanings (Fischer 2014: 132–133). Categorizations by non-experts as assessed in large-scale experiments were presented as an alternative. It was argued that in narratology and other fields concerned with describing and analyzing ordinary everyday practices, studying the categorization dispositions of laypeople can and should ground decisions concerning the conditions of subject domain membership. An experiment around pictorial narrative was introduced and its theoretical implications were presented. Using such studies as the basis of our definitions and theories, we can avoid that we inadvertently try to revise how terms like “narrative”, “story” or “game” are used rather than describing what they mean. In the future, it would be interesting to perform similar experiments with other marginal cases, e. g. wordless music and so-called non-mimetic narratives, where abstract figures seem to interact with each other, for instance in comics (see Baetens 2011, 2013). I believe that we would encounter substantial narrativity differences between, say, a rather static Mondrian painting such as Composition with Red, Yellow and Blue (1930) and dynamic Malevich paintings such as Suprematist Painting (1916) or even Red and Black Square (1915). Is it not likely that “the degree of narrativity of an ‘unstable’ constellation (with non-parallel edges, not-centered in the image, etc.) of different shapes (of unequal sizes and colors, of divergent geometric shapes, etc.) generally tends to be higher than that of a stable constellation of identical shapes” (Speidel 2018 b: 120)? The hypothesis remains to be tested. There is thus plenty of room for future experiments in descriptive narratology.