Abstract
It is easy enough to theorize about the impact of literary reading on empathic reactions, but to empirically prove such effects is more difficult. This article provides a concise overview of the theoretical and empirical evidence that literary reading has an effect on empathy (in its multiple forms), distinguishing between the effects of literariness, narrativity and fictionality. Subsequently, it discusses the results of 2 recently conducted studies into reading and empathic understanding. As will become clear, we are still quite a long way from knowing when and how (literary) reading has an effect on empathy for others. Suggestions are offered how future studies can tackle this issue.
Claims that the novel is dead seem to hold equal pace with claims that literary reading is the key to compassion and empathy. Throughout the last few decades, various scholars and authors have come to the defense of literature through emphasizing its supposed effects on fellow feeling (e.g., Booth, 1988; De Botton, 1997; Hakemulder, 2000; Nussbaum, 1995, 1997, 2001, 2010; Oatley, 1999, 2002, 2008, 2016; Pinker, 2011; Scarry, 1999). Often, these claims center on the narrative aspects of literature, sometimes in combination with its ambiguity or polyvalence (the fact that it can be read in multiple ways), or with its beauty or originality. As some have argued, reading literature is a form of role-playing, which helps us get into the minds of others, thereby leading to an increase in our empathic abilities (Booth, 1988; Hakemulder, 2000; Oatley, 1999; Rorty, 1989). Or, as Nussbaum (1995) has phrased it, literary reading triggers a type of imagination that is “an essential ingredient of an ethical stance that asks us to concern ourselves with the good of other people whose lives are distant from our own” (p. xvi).
It is easy enough to theorize about the potential impact of literary reading, especially since bibliophiles may be easily convinced. Yet, to empirically prove such effects is more difficult, not in the least because “literature” and “empathy” can be defined in diverse ways. In this article, I will first give a concise overview of the theoretical and empirical evidence that literary reading has an effect on empathy (in its multiple forms). Where possible, I will distinguish between the effects of literariness (containing aesthetic and/or unconventional features), narrativity (presenting related events happening to characters), and fictionality (depicting imagined events: what could have been or should be instead of what actually was; Koopman & Hakemulder, 2015). Theoretically, this is a crucial distinction, but unfortunately, many empirical studies conflate these different aspects of literary reading. Subsequently, I will discuss the results of two studies into reading and empathy I recently conducted, which tried to differentiate between effects of literariness, narrativity, and fictionality, while taking into account important reader variables (Koopman, 2015, 2016a). As will become clear, we are still quite a long way from knowing when and how (literary) reading has an effect on empathic understanding for others (Koopman, 2016b; Mumper & Gerrig, 2017). Although I will provide suggestions of how future studies can tackle this issue, I also wonder whether this is a question that empirical literary scholars can and want to answer completely.
Effects of Reading on Empathy: Theory
Empathy can be defined broadly as “the notion of responsivity to the experiences of another” (Davis, 1980, p. 3), with a distinction between affective and cognitive aspects. Cognitive empathy is the ability to understand someone else's perspective (Davis, 1980, 1983; see Theory of Mind or ToM: Baron-Cohen, 1991; Premack & Woodruff, 1978) and emotional empathy is about feeling similar emotions to someone else (cf. “emotional contagion”: Hatfield, Cacioppo, & Rapson, 1994). For higher level mental processing like literary reading, the separation between cognitive and affective aspects becomes somewhat artificial, because both types of processes interact (Frijda, 1986; Izard, 2009; Kneepkens & Zwaan, 1994; Nathanson, 2003). I use the overarching term empathic reactions to refer to this complex combination of affective and cognitive empathy.
Empathic reactions can be evoked by seeing the tribulations of real people, but also through reading about (fictional) characters. This second type of empathic response can be called “narrative empathy” (after Keen, 2007) and is part of the broader concept “narrative feelings” (see Kneepkens & Zwaan, 1994; Miall & Kuiken, 2002). Narrative feelings include feeling empathy and sympathy with characters (feeling with and for them; e.g., Coplan, 2004; Keen, 2006), identification with characters (seeing oneself as similar to them, taking on their perspectives; e.g., Mar, Oatley, Djikic, & Mullin, 2011), and absorption (also called “transportation,” feeling immersed in the narrative world; e.g., Busselle & Bilandzic, 2009; Green & Brock, 2000). In the available literature, these distinctions between various narrative feelings have often not been made, or somewhat different definitions have been used. When discussing the effects, I will try to be as clear as possible about which narrative feelings are involved.
How could reading influence empathic reactions? The idea that narrative texts would have stronger effects on empathy than expository texts makes sense theoretically. When reading a story, Oatley (1995, 1999, 2002, 2008) argues, we make mental models of the narrative world, take on the goals and plans of the protagonist (perspective-taking), and subsequently experience emotions according to our evaluation to what extent these goals are accomplished (Hakemulder, 2000). This type of simulation, or “role-taking,” Oatley (1999) suggested, is likely to result in identification, empathy and/or sympathy with these characters (see also Mar & Oatley, 2008). Further theoretical explanation of this perspective is provided by Kotovych, Dixon, Bortolussi, and Holden (2011), who argued that reading is like a conversation between narrator and reader: When we try to understand a character in a book, we make similar inferences about what the other is thinking and feeling as in conversation, and making such inferences would increase our understanding of and identification with the character. Reading narratives could, in this way, function as practice for inferring emotions and taking the perspectives of others in real life. Through the process of imagining others’ experiences, readers could eventually feel more empathy for others outside of the narrative world (Mar & Oatley, 2008; Nussbaum, 1995).
Claims have furthermore been made that fictional narratives would be especially conducive to empathic reactions. Oatley (e.g., 1999, 2002) has argued that fiction creates a beneficial environment to engage with others: As these others are not presented as existing in the real world, we may be less defensive and more inclined to indulge in feeling (Keen, 2006, 2007). We can care for characters without having to worry about our ability to actually come to their aid. This is important for us because, overall, we are social beings who are biologically very well prepared to feel what others feel and to deduce what others think (Breithaupt, 2012, p. 85). Our empathic ability, according to Breithaupt (2012), is something we need to block if we do not want to be overwhelmed by the perspectives of others. When reading or viewing fiction, the “self-loss” caused by empathy is not threatening, because the limited length of the story promises “an end to the empathic engagement when the development has come to an end” (Breithaupt, 2012, p. 85). To some extent, this goes for all narratives, but fictional narrative constitutes a clearer separate realm to “bypass the blocking mechanisms” that people commonly use to protect themselves from an excess of empathy (Breithaupt, 2012, p. 86). Through the fact that fictional characters are not “real” and that a fictional engagement has a time limit, the context of reading fiction may create an optimal aesthetic distance to engage with other people's tribulations (Cupchik, 2002). However, one could also argue that this “distance” to the real world tempers empathic reactions, precisely because fiction resembles play and brackets actual consequences.
Apart from narrativity and fictionality, literariness has been argued to be a factor in causing empathic reactions (e.g., Hakemulder, 2000; Kidd & Castano, 2013; Mar & Oatley, 2008; Oatley, 1999). Although Oatley has pointed to the higher complexity of characters in literary texts (e.g., Mar & Oatley, 2008), others have shown that striking stylistic features in literary texts (i.e., “foregrounding”) can increase the time needed to process the content as well as making the content more vivid (Hunt & Vipond, 1985; Miall & Kuiken, 1994), which could ultimately lead the reader to look at the world in a different way. 1 As Miall (2000) has proposed, reading texts with “special uses of language” (p. 51) helps to upset the stereotypical schemata through which we usually make sense of the world. “Through literary reading we dehabituate, that is, we are enabled to contemplate alternative models for being in the world” (Miall, 2000, p. 50). This can be called the “defamiliarization” hypothesis, based on Shklovsky's (1965) term ostranenie (estrangement or defamiliarization). Miall and Kuiken (1994, 1999, 2002) have discussed such an effect of literature in the context of affect more generally, but they did not relate defamiliarization explicitly to empathy. Still, following Shklovsky's (1965) idea that literariness sensitizes one's perception and felt sense of things, people and events, readers of literary texts might also be more likely to feel empathic understanding. On the other hand, it can be argued that striking stylistic features could make readers focus more on the form than on the content. This can create aesthetic distance between the reader and the narrative world (Cupchik, 2002), which might hamper feelings for characters (Kneepkens & Zwaan, 1994), and thereby possibly for real people in similar situations. We need to turn to empirical research to find out which effects different aspects of literary reading actually has, and when (for which readers and which texts).
Effects of Reading on Empathy: Empirical Evidence
When it comes to the general impact of reading narratives on empathy, several studies yielded positive effects on various empathic measures, like self-reported cognitive empathy (Djikic, Oatley, & Moldoveanu, 2013), empathic attitudes toward outgroups (e.g., Batson et al., 1997; Hakemulder, 2000, 2008; Johnson, 2013; Litcher & Johnson, 1969; Marlowe & Maycock, 2001; Shapiro, Morrison, & Boker, 2004), and motivation for prosocial behavior (e.g., Bilsky, 1989; Johnson, 2012).
Although these kinds of studies typically did not include a thematically related non-narrative text as control condition, those among them which did use adequate control conditions suggest a stronger empathic effect of narrative than non-narrative texts (e.g., Black & Barnes, 2015; Hakemulder, 2000; Djikic et al., 2013; Kidd & Castano, 2013). Hakemulder (2000), for example, found that readers of a narrative text about a woman in a fundamentalist Islamic country opposing traditional gender roles were more inclined than readers of an expository text on the same subject to believe that women in such countries find it hard to accept their secondary position in society. It may be that readers are more likely to (over)generalize from the experience of one character than from an essay recounting the experiences of many (cf. the “identified victim-effect,” e.g., Kogut & Ritov, 2005). This general effect that imagining how one person feels leads to sympathy for that person (the “empathy-attitude effect”), which leads to more understanding for the group as a whole (the “empathy-attitude-action effect”), has also been shown in social psychological research by the Batson research group (Batson et al., 1997; Batson, Chang, Orr, & Rowland, 2002). 2 In Batson et al.'s (2002) experiment, participants listened to an interview with a heroin addict, after which they completed a scale about their attitudes toward addicts and were asked to allocate funding to an outreach program. Participants who received a perspective-taking instruction (imagining how the addict feels) allocated significantly more funding and had significantly more positive attitudes than those who were instructed to be “objective.” It needs to be noted that Batson et al. (2002) had to instruct their participants to imagine themselves to be in the “character's” position. For readers, fictionality and literariness might play a role in making such engagement with the character occur naturally.
That fiction reading has an effect on—or at least a relationship with—empathic ability, has recently been suggested in a meta-analysis comparing correlations between measures of lifetime reading habits for both fiction and nonfiction with measures of empathy and ToM (Mumper & Gerrig, 2017). As Mumper and Gerrig (2017) indicate, “the results of this meta-analysis generally agree with the existing literature that fiction has a stronger relationship with empathy and ToM as compared to nonfiction” (p. 8). Yet, the average correlation was small, especially the correlations of fiction with (self-report) empathy measures (r = .07); the relation with ToM was somewhat more convincing (r = 22). More importantly, as in most of the research claiming an effect of fiction, the studies reviewed generally looked at the aggregate category “narrative fiction” or “literary fiction,” saying little about a potential effect of fictionality.
This was indeed the case in two of the most notable studies claiming a relation between fiction reading and empathy: Mar, Oatley, Hirsh, dela Paz, and Peterson (2006) and Mar, Oatley, and Peterson (2009). Both of these studies found correlations between people's empathic ability and people's general exposure to narrative fiction (as opposed to nonfiction). Empathic ability was measured by the Reading the Mind in the Eyes Test (RMET: Baron-Cohen, Wheelwright, Hill, Raste, & Plumb, 2001), in which participants view 36 pictures of actors’ eye-regions and each time attribute one out of four varying emotion words to these pictures. Lifetime exposure to narrative fiction was measured by Mar et al.'s (2006) version of the Author Recognition Test (ART: Stanovich & West, 1989; West, Stanovich, & Mitchell, 1993), in which participants have to indicate which author names they recognize. The correlation between the ART and the RMET has been replicated by Djikic et al. (2013) and Panero et al. (2016).
Of course, because of the correlational set-up of these studies, we do not know what came first: reading or empathic ability. But even if we could say that reading came first, that it has an effect on empathy, it can be doubted whether this is an effect of “fiction” in a strict sense (i.e., of narratives that do not make truth claims), or an effect of “narrative.” The revised ART by Mar et al. (2006), which distinguishes between “fiction” (i.e., novels) and “non-fiction” (i.e., essays, popular science), may be better characterized as a measure of exposure to narrative texts versus non-narrative texts. Finally, it needs to be stressed that the RMET is a rather basic measure of empathy, measuring the very first stage of ToM: attributing the correct mental state (Baron-Cohen et al., 2001). It is not a direct measure of empathic behavior or attitudes. Other studies suggest that fictionality may not be crucial. In the experiment described above by Batson et al. (2002), similar effects were found when the researchers told the participants beforehand that the interview was a work of fiction, but the effects were not as strong (i.e., positive) as for nonfiction.
For now, we cannot yet rule out a specific effect of fiction. In favor of a fiction effect, there is neurocognitive evidence that when one reads the same texts as either fictional or factual, responses differ (Altmann, Bohrn, Lubrich, Menninghaus, & Jacobs, 2014). In Altmann et al.'s (2014) fiction condition, brain activation patterns of participants suggested that they perceived the events in stories as possibilities of how something might have been, imaginatively constructing these hypothetical events or scenarios (Oatley, 1999). In the nonfiction condition, on the other hand, readers focused on content, on updating their world-knowledge. Altmann et al. (2014) suggested that the simulation processes of reading fiction must involve perspective-taking and generating relational inferences.
Causal evidence for a relation between reading specifically literary narratives and empathic ability can mainly be found in the famous study by Kidd and Castano (2013). In a series of five experiments, Kidd and Castano (2013) compared reactions on multiple measures of ToM after reading literary fiction versus popular fiction, and literary fiction versus expository nonfiction. In both comparisons, they found higher scores for the literary condition on the RMET and on the Yoni test (Shamay-Tsoory & Aharon-Peretz, 2007)—the latter test measures correct inference of emotions when looking at cartoons, using verbal and eye gaze cues.
Again, however, these measures of ToM tell us little about empathy in real-life situations. More importantly, it is unclear to what kind of literary feature or features the results can be attributed, because Kidd and Castano (2013) selected the literary texts on the basis of critical acclaim and the popular texts on the basis of sales figures. They did not match the literary and nonliterary text according to subject and they did not take into account in which ways the literary and nonliterary texts differed. Differences between the popular and literary texts in the Kidd and Castano (2013) study might explain why Djikic et al. (2013) failed to reproduce these results on the RMET when they provided participants with either stories or expository texts of equal interestingness. Even more worrisome is the fact that two recently published studies, which were rather exact replications of the Kidd and Castano study, failed to replicate the findings of short-time literary reading on the RMET (Dijkstra et al., 2015; Panero et al., 2016). Whereas Black and Barnes (2015), on the contrary, claim to have replicated the findings of Kidd and Castano (2013), this study only made the comparison between texts classified as literary fiction and as nonfiction. Just as in Kidd and Castano's (2013) study, the nonfiction texts of Black and Barnes (2015) contained no human protagonists, while the literary fiction texts did.
Still, there is cause for optimism regarding the effects of literariness. A study by Pino and Mazza (2016), in which participants (N = 214) read entire books—either literary, expository nonfiction, or science fiction—also replicated Kidd and Castano's (2013) finding that reading literature enhances mentalizing abilities.
The studies discussed above thus provide limited and mixed empirical evidence for an effect of literariness on empathy. The evidence for an effect of fiction on empathy is also not convincing, since these studies were mostly correlational and generally conflated narrativity and fictionality. What we have found so far may mostly be effects of narrativity, that is to say, of imagining the feelings, thoughts and actions of a character (perspective-taking).
The Importance of the Reader-Text Interaction
To determine whether and under which conditions which texts lead to empathic reactions, first of all, more systematic comparisons between different types of texts need to be made. Such comparisons would preferably control quite carefully which elements of the text vary, so that conclusions can be drawn about what it is in the text that causes empathic reactions. Second, to grasp the mechanisms through which reading can lead to empathic reactions, it is relevant to take into account personal background characteristics of the reader as well as the reader's experience during reading: how this reader responds to characteristics of the text.
Of course, various personal characteristics of readers could influence empathic reactions; I will only mention a few likely candidates that I have included in the empirical studies that I will discuss shortly. First of all, while reading about others could temporarily increase (state) empathy, readers also already have a certain empathic sensitivity to others before starting to read a text. This disposition, trait empathy, is a personality variable that would logically influence one's empathic reactions to real-life others directly, but perhaps also through making one empathize more with characters. As some have argued (e.g., Mar et al., 2006, 2009), trait empathy could have been developed more strongly for some people because of their exposure to literature. Following that argument, both trait empathy and one's previous exposure to literature are personal factors to be taken into account when experimentally exploring the relation between (literary) reading and empathy.
Furthermore, personal experience can be an important factor. As Loewenstein and Small (2007) argued in their overview of factors that are likely to contribute to sympathy and prosocial behavior, having been through an experience ourselves makes it easier to feel what this is like for another. There is some empirical evidence for the importance of readers’ personal experiences in the context of narrative persuasion (Green, 2004), but not as straightforwardly for empathy with others (see, however, Koopman, 2015).
As suggested above, such personal factors could directly influence empathic reactions after reading, but they could also influence narrative feelings toward characters and the narrative world, which can in turn be expected to influence empathic reactions (mediation). A couple of studies have pointed to the role of narrative feelings in determining empathic outcomes. For example, Bal and Veltkamp (2013) found that reading a fiction text (as opposed to a nonfiction newspaper article) only caused an increase in self-reported affective empathy after reading (measured by Davis's Interpersonal Reactivity Index) when participants felt absorbed in the story (measured by the narrative engagement scale by Busselle & Bilandzic, 2009).
Johnson (2012, 2013) provides further empirical evidence for the importance of narrative feelings. Johnson (2012) assumed that the degree of being absorbed into a narrative indicates the extent to which one simulates the social experiences depicted, and that this is the main way in which reading leads to empathic responses (see also Mar & Oatley, 2008). Indeed, participants who reported higher “transportation” also reported higher sympathy with the character and were more likely to engage in prosocial behavior (picking up a pen the experimenter had dropped). This effect was independent of trait empathy (Johnson, 2012). In addition, Johnson (2013) found supporting evidence for the notion that absorption is responsible for changes in beliefs and attitudes toward outgroups. Although it is unclear whether transportation occurs before sympathy and empathy with characters (or vice versa) or whether these aspects of narrative engagement occur simultaneously, the Johnson studies do suggest that transportation and sympathy/empathy with characters work together in influencing attitudes toward others who are like the characters.
With regard to aesthetic feelings and empathic reactions, very few empirical studies have been conducted (see Koopman, 2015, 2016a). There are suggestions that aesthetic features and aesthetic feelings are connected to an overall more emotional response (e.g., Miall & Kuiken, 1994; Van Peer, 1986). Finding a text more beautiful can help to find its characters more sympathetic or vice versa: aesthetic feelings and narrative feelings may reinforce one another, in a process of oscillation (Cupchik, 2001). Indeed, empirical evidence suggests that aesthetic and narrative feelings go together, although correlations between perceived foregrounding and narrative feelings tend to be lower than those between aesthetic appreciation and narrative feelings (Andringa, 1996; Koopman, 2011; Koopman, Hilscher, & Cupchik, 2012).
Two Recent Studies: Method
In two recent experimental studies within my dissertation research project, Reading Suffering: An Empirical Inquiry Into Empathic and Reflective Responses to Literary Narratives (Koopman, 2016b), I tried to take into account both reader and text characteristics, and the interaction between these. Both studies presented readers with a text and a questionnaire. The questionnaire asked about trait empathy, exposure to literature (via a Dutch adaptation of the Author Recognition Test), personal experience with the subject matter (grief and depression), narrative feelings and aesthetic feelings during reading, and empathic understanding for those in a similar situation to the character (people grieving and people suffering from depression). 3 Participants were also given the option to donate (part of) the money they received for participation to a charity connected to grief or depression.
In the first empirical study, which I label the “genre study” (Koopman, 2015), trait empathy was measured through the Toronto Empathy Questionnaire (Spreng, McKinnon, Mar, & Levine, 2009), which aims to be a unidimensional measure of emotional empathy, based on earlier empathy measures such as Davis's. In the second study, the “foregrounding study” (Koopman, 2016a), I used two subscales of Davis's (1980, 1983) Interpersonal Reactivity Index: empathic concern and perspective-taking. In each case, half of the respondents received the trait empathy questions beforehand and the other half afterward, so I could make sure there was no order effect—in that case, namely, the measures would not have been valid as trait empathy. There were indeed no order effects.
The studies also differed. The genre study compared reactions to expository, simple narrative, and literary narrative texts. Participants (N = 210) were randomly divided into these “genre” conditions. These participants were all students. In addition, for the two narrative conditions, the genre study included a fiction versus a nonfiction instruction as a between-subjects variable. From a computer screen, participants read two texts in each genre, one about depression and one about grief, with a week between the two texts.
The “foregrounding study” did not compare between “genres” (which was still a rather broad distinction), but between three versions of one literary text about grief (Anna Enquist's 2008/2010 Counterpoint). These versions only differed in the amount of original stylistic features: the level of foregrounding was manipulated. Examples of this manipulation can be found in Table 1. Participants (N = 142) read from paper instead of from a screen, and this study included a qualitative component: readers were asked to underline sentences which evoked feelings and thoughts. After reading the entire text, they were asked to explain further what these feelings and/or thoughts had been. Participants were both students and their parents.
Examples of Manipulations Foregrounding Study (Koopman, 2016a)
Note. The English translation is a combination of the official translation by J. K. Ringold (Enquist, 2008/2010), and my own translation, as a few instances of foregrounding got lost in translation.
Two Recent Studies: Mixed Findings Explained
The two experimental studies had different outcomes on the empathy measures. The genre study did not show any significant difference between the genre conditions on the attitudinal measure of empathy with others (empathic understanding). The fiction instruction in this study also did not matter. The foregrounding study, however, did show the expected difference: those readers who had read the most original text (containing most foregrounding) had the highest empathic understanding afterward (see Table 2). Post hoc analyses indicated that the significant difference lay between the original condition and the version without foregrounding (Fisher's least significant difference). This effect was upheld when controlling for the personal factors gender, student/parent, trait empathy, exposure to literature, and personal experience.
Analysis of Variance Mean Differences of Affective Responses and Empathic Understanding per Foregrounding Condition (Including Post-Hoc Test Significances)
p < .05
p < .01 (2-tailed): Indicates significant mean difference from the original condition, based on Fisher's LSD post-hoc tests. The exact p values in the table are the p values of the F-test.
There are multiple explanations for this varying outcome. First of all, the two studies made different comparisons: It could be the case that the (more sentimental and straightforward) simple narrative texts from the “genre study” formed a more directly affective pathway to empathic understanding than the literary texts, which might work through a combined affective-cognitive sensitization to others’ experiences (cf. Nussbaum, 1990; Shklovsky, 1965). These different pathways could have led to similar outcomes on empathic understanding.
Second, the fact that an effect was found in the foregrounding study while it was not found in the genre study could be due to the conditions of the experiments: The foregrounding study used a more elaborate instruction, asking participants to underline passages evoking thoughts or feelings. In addition, in the genre study, readers read from a screen, whereas in the foregrounding study they read from paper, which might have caused a more shallow type of reading in the genre study (Ackerman & Goldsmith, 2011; Mangen & Kuiken, 2014). 4 Both of these factors may have caused readers in the foregrounding study to pay more attention to the texts, and attention may be a necessary precondition for literary texts to have their effects (see Hakemulder, 2004; Van Peer, 1986). The responses to the open questions in the foregrounding study indeed suggested that many people in this study read this text with their full attention, causing incredibly absorbed reactions like “I could almost touch the dead girl, and she feels cold.” Even though not all readers had such strong experiences, the fact that these did occur after reading only 1,500 words within the context of an experimental study is indicative of the potential power of literature.
Another difference between the genre study and the foregrounding study was the fact that in the foregrounding study, only the foregrounding in the text had an effect on empathic understanding, personal factors did not. The genre study, on the other hand, found effects of both people's trait empathy and of their lifetime exposure to literature on empathy. This latter finding, that one's lifetime exposure to literature has an effect on empathic understanding, is in line with earlier research using different measures of empathy (Mar et al., 2006, 2009). In the foregrounding study, the effect of the original text condition overruled the potential effects of personal factors.
Although the effect of lifetime exposure to literature on empathic understanding was not replicated in the foregrounding study, the foregrounding study did find that one's exposure to literature had an impact on the other empathic outcome measure: donating to a related charity. This effect remained in place when controlling for whether one was a parent or student and education, which is relevant, because people who read tend to be older and highly educated (Witte & Scholtz, 2015). In the genre study, it appeared to be mainly the emotional effect of the simple narrative text about depression that influenced whether one donated. Prosocial behavior as specific as donating might, then, be positively influenced by reading both in the short run (when reading emotional texts) and in the longer run (when reading diverse authors), but the current experiments do not yet provide enough evidence for a specific long-term effect of literariness (let alone fictionality) on such concrete actions. Also, of course, for actual prosocial behavior to occur, many other factors can play a role, like one's awareness of need, one's income (Wiepking & Bekkers, 2012), one's values and one's calculation of the costs and benefits (Bekkers & Wiepking, 2011).
Still, these are promising results for reading narratives in general, particularly given the fact that in addition to these findings, a significant correlation was found between lifetime exposure to literature and self-reported cognitive empathy (or: perspective-taking). There was no relation between lifetime exposure to literature and self-reported emotional empathy. This suggests that Oatley's role-taking theory and Nussbaum's ideas about the importance of practicing one's imagination could be correct: reading may indeed train our ability to understand what it is like to walk a mile in someone's shoes, to take in other perspectives. It remains to be further researched whether literary texts are more conducive to training this ability than simple narrative texts.
The mechanisms through which reading can impact empathy were not completely clarified in my project, although there is some support for both role-taking and defamiliarization. In the genre study, sympathy/empathy with a character appeared to be a clear predictor of empathic understanding for others (role-taking). In the foregrounding study, responses to the open questions pointed toward a mixed emotional state as a potential explanation for increased empathic understanding. For example, as one participant said: “For me this expresses a certain sad beauty. Despite the situation being sorrowful, this part is an expression of a sort of gray joy” (italics added for emphasis).
Although it could not statistically be checked whether mixed emotions affected empathic understanding, such an explanation would be in line with Shklovsky's (1965) defamiliarization theory, which posits that art can make us feel the essence of experiences and things—the stoniness of the stone. More research into these types of mixed feeling states within aesthetic experience and the potential “transformative” power of beautiful and original language could lead to further insight into the relation between foregrounding, feeling, and understanding. Recently, Hanich, Wagner, Shah, Jacobson, and Menninghaus (2014), and Menninghaus et al. (2015) have looked into the mixed emotional state of “feeling moved.” “Feeling moved” as they conceptualize it, as a combination of sadness and joy, might be similar to the mixed feeling states reported in the foregrounding study, but I would hypothesize that beauty and/or originality form an additional aspect to sadness and joy. The joy that was experienced by the readers in the foregrounding study may partly have been joy in the beauty of the language.
Discussion and Critique: What is Literature For?
As the diverse studies discussed above, and their mixed findings, point out, we have only just begun to scratch the surface of the issue whether and how literary reading leads to empathy. Many more experimental comparisons between texts varying in literariness, narrativity, and fictionality need to be made in order to (dis)confirm the current findings. To deduce what type of reactions are evoked by fictionality and by narrativity, future studies could include manipulations of mixed genres like literary nonfiction and journalistic prose which combine narrative and expository aspects, and experiment with different instructions. In addition, to see whether there is something specific about reading (e.g., the aspect of having to form images oneself), future studies could make comparisons between different media, varying in the level of artistry (literariness).
To determine the role of literariness in evoking empathic reactions, different aspects associated with this construct would need to be looked at, for example: foregrounding on a broader level (e.g., narrative perspective), polyphony, and the extent to which a text contains gaps (Iser, 1988). The contribution of gaps in fostering perspective-taking and empathy is a relevant avenue for further research, because it can theoretically be expected that gaps are conducive to readers making mental inferences (Kotovych et al., 2011). However, the current findings suggest that it would also certainly be worthwhile to further investigate to what extent and when semantic, phonetic and grammatical foregrounding evoke mixed emotional states (Miall & Kuiken, 1994, 2002; Shklovsky, 1965), and when and how such states lead to empathy for others.
If a wider audience would be exposed to literature, would humanity then benefit? It is my position that we should be careful in claiming great prosocial effects of literary reading. Not just because we as of yet do not have that much empirical evidence to go on, but mainly because it would reduce literature to one potential function, while it has value in and of itself. This value lies in beauty, in craftsmanship, in expression, and in the engagement with a human cultural practice, a tradition—whether by substantiating or breaking with that tradition. It remains valuable to see what literature can do to whom, in order to further our knowledge in general as well as for practical purposes such as reading programs in schools and prisons, but we should be wary of judging an art form primarily on the basis of its prosocial uses. Literature does not simply lead us to become better people, literary works can also be disturbing, and it is vital that they remain a free playing space for human thought.
To some extent, then, I hope that we will never completely find out which effects which literary features have on empathic understanding and prosocial behavior. That, namely, would be like finding out exactly why we fall in love with whom.
Footnotes
1
Because of the limited scope of the current article, I cannot fully go into the issue what constitutes “literature”, while this is, of course, contested. Generally, I find a partly text-immanent, partly subjective definition of literature most fruitful. Such a definition considers literary language to deviate from “everyday” language, focusing attention on the language itself (“foregrounding”: Mukařovský, 1976), but this originality also needs to be realized by the reader (Miall & Kuiken, 1994, 
). “Literariness” can be seen as being present in a certain text to a lesser or greater extent by identifying such stylistic features as metaphors and similes, but if the reader browses over these features, is unimpressed by them, then they are unlikely to make a difference.
2
The term empathy-attitude effect that Batson et al. (1997, 2002) use is somewhat misleading, since the induced empathy with an individual that they are referring to is actually a composite construct empathic feelings, which leans largely on what others would label sympathy, namely the adjectives “sympathetic, compassionate, softhearted, warm, tender, and moved” (Batson et al., 2002, p. 1660, italics in original).
3
The outcome variable empathic understanding was measured by statements such as “I feel understanding for people who are depressed,” “It must be horrible to be depressed,” and “People with depression are being dramatic” (reversed).
4
The experiment by Ackerman and Goldsmith (2011) in which participants read expository texts, either from paper or from screen, demonstrated “more erratic study-time regulation on screen than on paper” (p. 18), which, they argued, may be due to people's perception that “the electronic medium is better suited for fast and shallow reading of short texts” (p. 29). In the experiment by , participants reading from screen reported higher “dislocation” in a text (i.e., a feeling for the length of the texts) than those reading from paper, and when adding a non-fiction instruction, participants reading from screen experienced less narrative engagement than those reading from paper.