Caspar J. van Lissa, Marco Caracciolo, Thom van Duuren, and Bram van Leuveren

Difficult Empathy

The Effect of Narrative Perspective on Readers’ Engagement with a First-Person Narrator

Many claims have been advanced about the effects of specific narrative strategies on readers’ engagement with characters, but the available evidence is still limited. One question in particular stands out in the current debate. Is first-person narrative more or less conducive to empathy and trust for the protagonist than third-person, internally focalized narrative? This essay tackles this question by examining the effect of narrative perspective on readers’ responses to a complex, and potentially unreliable, character. To this end, we conducted an experimental study with 76 Dutch high-school students. Contrary to our predictions, the manipulation of narrative perspective did not affect empathy for the character, but did affect trust. We suggest that the increase in trust in third-person narrative depends on the external narrator’s authority, which validates the perspective of the protagonist. The essay discusses these and other findings, combining experimental research with a qualitative analysis of readers’ comments on the character.


The question of audiences’ attitude towards fictional characters looms large in the study of narrative in literature and other media (Eder / Jannidis / Schneider [Eds.] 2010). Readers and scholars commonly talk about ‘identification’ or use metaphors such as ‘closeness’ and ‘distance’ (Eder 2006) or ‘putting oneself in a character’s shoes’, for the experience of relating to a fictional being. Recently, researchers have given attention to empathic perspective-taking as the main psychological mechanism underlying this experience (Gaut 1999, Coplan 2004, Mellmann 2010, Keen 2013). Fiction, it is claimed, encourages readers to imagine particular aspects of characters’ psychological life in a first-person way, allowing them to take on those aspects through empathic responses: “While reading we find ourselves in the shoes of a wide diversity of people”, argues Frank Hakemulder (2000, 97). On the view defended by Hakemulder and others, empathy is central not only to our imaginative interactions with characters, but to our engagement with fiction as such.

Yet, as highlighted by Suzanne Keen, the evidence suggests that “lowbrow fiction evokes empathy more reliably than treasured classics” (Keen 2007, 84). In fact, literary fiction can be said to both encourage and problematize empathic reading strategies, confronting readers with protagonists who – unlike the heroes of popular genres – call for ambivalent ethical judgments and evaluations. Empathy often depends on a recognition or projection of similarity (cf. Eder 2006, 74-75), while literary texts can foreground protagonists who are distant – socially, mentally, or otherwise – from typical readers. In this sense, there might be more to engaging with literary characters than can be predicted through models focusing exclusively on identification or empathic perspective-taking.

Through the empirical study discussed in the following pages, this article makes a first attempt at coming to grips with readers’ responses to ‘round’ literary characters (Forster 1985), who are multifaceted and may challenge readers rather than straightforwardly invite empathic responses. We seek to map readers’ shifting attitudes towards literary characters by focusing on the interplay between two psychological processes: empathy and trust. There is broad consensus in the psychological literature that empathy is a complex, multidimensional construct, which involves both affective, sympathetic responses to others’ emotions, and cognitive perspective-taking (Davis 1983). Trust, on the other hand, can be used as a measure of a character’s perceived reliability, thus reflecting readers’ evaluation of that character – and tying in with discussions on narratorial unreliability in narratology (on which more in the next section). We will use as a case study a narrative situation that is bound to create ambivalence in the audience: The first chapter of Knut Hamsun’s proto-modernist novel Hunger ([1890] 2001). This text features a first-person narrator who can be suspected of narrative unreliability. While the narrator is not altogether unlikable, his erratic behavior and atypical social status (he is a marginalized, struggling writer) are likely to complicate readers’ attitude towards him. Through experimental manipulation of the narrative perspective, the study investigates how readers’ engagement with the protagonist changes when the text is presented in either first-person form (possibly unreliable character-narrator) or third-person narration with internal focalization. “Internal focalization” (cf. Genette 1980) refers to a narrative situation where the narrator is external to the storyworld – “heterodiegetic” – and talks about the characters in the third person while providing information about the protagonist’s inner, mental life. Our experimental design allows us to examine how character narration affects readers’ attitude towards Hamsun’s protagonist when they cannot depend on a conventionally authoritative, external narrative figure, as in internally focalized texts. Our hypothesis is that first-person narrative brings out the ambivalence of readers’ responses to characters in literature, especially when coupled with a potentially unreliable narrator. While the bulk of the study is experimental, we included two open-ended questions at the end of the questionnaire to explore participants’ engagement with the character from a qualitative perspective.

The study of readers’ attitudes towards complex, literary characters bears on a different, but related, strand of research on the real-world effects of engaging with fiction. Over the last fifteen years scholars have looked at the impact of literary reading on the self-concept (Hakemulder 2000), social cognition (Mar et al. 2006), and theory of mind (Kidd / Castano 2013). These approaches tend to downplay an important distinction between fiction and literary fiction, as two of the authors have argued in past work (Caracciolo / van Duuren 2015). Narrative unreliability is a literary technique par excellence, and understanding the specific effects of literary characters, as opposed to characters found in popular genres, seems crucial to assessing literature’s psychological impact. If, as scholars working in different traditions have suggested (Shklovsky 1965, Iser 1978, Cook 1994), literary fiction tends to challenge – rather than confirm – readers’ beliefs and expectations, then this challenge is at least in part created by morally ambivalent characters. Literary characters tend to overstep readers’ comfort zone, confronting them with perspectives and worldviews dramatically different from their own.

Bringing into focus the psychological dynamics triggered by literary characters may help scholars and teachers use literature more effectively to prompt reflection on ethical and social issues. This seems particularly important in a developmental context: Stephanie D. Preston and Frans de Waal’s (2002) “Russian doll” model of empathy suggests that empathic abilities become more complex with increasing age, developing from direct affective emotional contagion, to cognitive emotion recognition, to affective empathic concern for the misfortunes of others, and – finally – voluntary cognitive perspective-taking. This theory is supported by empirical research, which shows that both dispositional empathic concern and perspective-taking are not yet fully developed in adolescents, and that empathic concern predicts their development of perspective taking between the ages of 14 and 17 (Van Lissa et al. 2014) – roughly the age range of the participants in our study. Moreover, recent work suggests that reading literary fiction might contribute to the development of affective and cognitive empathic abilities (Kidd / Castano 2013). Given the central (but increasingly contested) role played by literary reading in many educational curricula, the idea of using literature to improve not just linguistic and cultural competencies but also students’ intersubjective skills appears promising. However, the study of the psychological effects of literary reading should not be segregated from the study of the psychological processes underlying it: in order to turn fictional worlds into tools for cultivating real-world skills in educational settings, we need to know more about how we relate to these worlds’ elusive inhabitants. As this article will argue, our responses to characters are often just as multifaceted as the characters themselves.

Readers’ responses to character narrators: Open questions

Several scholars have highlighted the scarcity of empirical work on readers’ engagement with characters in literary studies. Keen, for instance, writing about empathy for characters, notes that most

of the existing empirical research on empathetic effects in narration concerns film (Tan 1996; Zillmann 1991) […]. Novels and stage drama are least studied empirically (though often theorized about), their length and performance conditions being, respectively, at odds with the current modes of empirical verification. (Keen 2013, para. 11)

Two psychologists, Thalia R. Goldstein and Ellen Winner, argue that while “there have been numerous theoretical explorations of sympathetic reactions to fictional characters (e.g., Coplan 2004), no research has examined the psychological components of these reactions” (Goldstein / Winner 2012, 134). There are, of course, a few exceptions. Willie van Peer and Henk Pander Maat (1996) and Hakemulder (2000) examined how empathy and sympathy for characters change as a function of narrative perspective; they found that narrative perspective does have effects on readers’ responses, but the exact nature of these effects depends on the narrative’s subject-matter. Maria Kotovych and colleagues (2011) investigated the textual underpinnings of identification through the lens of the Gricean notion of “implicature” (an inference based on the assumption that the narrator has a cooperative attitude towards the reader). They found that the more is left textually implicit about a narrator’s mental life, the more readers are likely to draw inferences based on their own experiences, which may lead to stronger identification. Finally, Geoff F. Kaufman and Lisa K. Libby’s (2012) study of belief change in response to fictional characters compared the effects of different narrative situations; they focus on what they call “experience-taking”, a particular form of empathic perspective-taking that is “experientially driven […] rather than […] conceptually driven” (Kaufman / Libby 2012, 15). This notion is also supported in the psychological literature; specifically, experiencing empathy-arousing stimuli was found to promote spontaneous perspective-taking (Hawk / Fischer / van Kleef 2011). Kaufman and Libby’s (2012) claim is that experience-taking in relating to characters results in more marked changes in readers’ beliefs and self-concept than modes of engagement that do not involve experience-taking.

Although some of these studies involved first-person narrative, none of them investigated the specific effects of unreliable character narration. Kotovych et al. (2011, 287) do acknowledge that “the reader may abandon the cooperativeness assumption” as soon as there is a suspicion of unreliability, but their experiments didn’t test this possibility. Yet the notion of unreliability is central in narratological discussion of first-person narrative (Shen 2013). Narrative theorists have tended to embrace either a ‘rhetorical’ (Phelan 2005) or a ‘cognitive’ paradigm (Yacobi 1981; Nünning 1999) in studying unreliability. Those in the cognitive camp have argued that unreliability is not a feature of texts and narrators, but an interpretive construct to which audiences may appeal in attempting to solve textual inconsistencies (Yacobi 1981). Character narration confronts readers with a narrator who is distinct from the flesh-and-blood author, and whose perspective on the storyworld may be limited, biased, or otherwise unreliable. Thus, unreliability invites readers to contrast the story as told by the narrator with hypothetical alternative versions of the events (if the narrator misunderstands what happened), or with alternative value systems (if the narrator expresses judgments that clash with what we understand to be the author’s own ethical framework).

Given this sophisticated perspectival play, we may wonder how character narration differs from third-person narrative in terms of readers’ responses. Van Peer and Pander Maat (1996) use a first-person excerpt in one of their empirical studies, but fail to comment on these differences. Hakemulder (2000, chap. 5) compares the effects of two versions of the same third-person text – one with internal focalization, another without – but does not consider a first-person story. Kaufman and Libby do address the effects of character narration, along the following lines:

We expected that first-person narratives, by virtue of creating a more immediate sense of closeness and familiarity to the main character, would be more conducive to experience-taking [i.e., empathy] than would third-person narratives, which explicitly position protagonists as separate entities (and, in our view, are more likely to position readers as spectators). (Kaufman / Libby 2012, 3)

Two factors seem to complicate Kaufman and Libby’s hypothesis: first, third-person narrative can be rich in details about a character’s inner life, in the technique known as “internal focalization”. Here the narrative voice remains that of an external narrator, but the text focuses on the experience of a fictional character on the scene. It has often been claimed that this device – typical of modernist fiction – creates an illusion of direct access to characters’ minds (Cohn 1978). A possible explanation for this illusion of transparency is that in internally focalized narrative the external narrator is always, by convention, authoritative (cf. Doležel 1998, 149). Since the external narrator implicitly endorses the textual references to a character’s mental life, readers are likely to take these references at face value, as a faithful reproduction of the character’s thoughts and feelings (for more on this point, cf. Caracciolo 2014). The second factor that potentially undermines Kaufman and Libby’s claim is narrative unreliability – a widely used literary device, and one that readers may see as a possibility inherent in first-person narrative. In this narrative situation, the protagonist is understood to be deliberately telling his or her own story after the fact, and therefore may be manipulating the audience. As an example, consider Humbert Humbert, the pedophile (and famously unreliable) narrator of Vladimir Nabokov’s Lolita ([1955] 1997): it seems reasonable to think that a narrator of this sort, despite relating his first-hand experience, does not create “a more immediate sense of closeness and familiarity”, as Kaufman and Libby would have it. On the contrary, Humbert Humbert is likely to estrange the audience because of the clash between their own moral values and those they attribute to the narrator.

Humbert Humbert is a somewhat extreme case, but many 19th and 20th century novels feature first-person narrators who are complex in the sense of being neither completely likable nor completely dislikable. Without coming across as downright immoral (as Nabokov’s narrator will be perceived by most readers), these characters often seem unpredictable and morally ambivalent. A degree of psychological complexity is widely regarded as a distinctive feature of literary narrative as opposed to more popular genres, in which protagonists tend to conform to social stereotypes and norms. Readers’ engagement with these literary characters is likely to be a multifaceted process that develops over the course of the reading experience. And yet, while scholars have argued that empathy for character – or related phenomena such as identification or perceived ‘closeness’ – are central to audiences’ engagement with fiction, the problematics raised by literary characters remain largely unexamined in empirical studies. First-person narrative is an interesting test bed because it is likely to trigger conflicting expectations and interpretations: on the one hand, a narrator who relates his or her own experience may seem more approachable than an external narrator (as predicted by Kaufman and Libby) and thus evoke notions of spontaneity and authenticity (cf. Korthals Altes 2014, 147-151), which may decrease the imaginative distance between the narrator and readers. On the other hand, if the narrator appears unconventional or distant in social and / or ethical terms, he or she may be suspected of unreliability, and therefore the audience may be encouraged to take a more distanced stance towards him or her. This ambivalent dynamic may problematize straightforwardly empathic responses to the narrator.

The experimental study discussed in the next pages makes a first attempt at addressing these issues. The main research question is whether complex literary characters evoke greater trust and empathy as first-person narrators or as protagonists of internally focalized narratives in the third-person form. Put otherwise: given the same literary character who may arouse suspicions of unreliability, are readers more likely to trust and empathize with him or her if the character is also the narrator of his or her own story, or if that same story is related by an external voice focusing on the character’s experience?

The experimental study


We conducted an experimental study in order to test the effects of narrative perspective on readers’ trust and empathy. Narrative perspective was manipulated (first- vs. third-person narrative), and we measured participants’ age, dispositional empathy (i.e., participants’ self-reported predisposition to empathize with other individuals in the real world); and number of novels read over the previous year (as an indirect measure of reading expertise). Dependent variables were empathic concern for the protagonist (i.e., ‘affective empathy’), perspective-taking for the protagonist (i.e., ‘cognitive empathy’), and trust for the protagonist. We recruited 76 Dutch high-school students (39 female), aged between 14 and 18 (M = 16.42, SD = 0.82). We conducted the study during the normal school hours, making clear that the students wouldn’t be evaluated on the basis of their answers.

The study involved reading a Dutch translation of the first chapter of Hamsun’s Hunger, which offers a prototypical example of a complex, unpredictable literary character (cf., e.g., Wood 2010, 6): a young and famished would-be writer recounts in vivid detail his peregrinations through Oslo. While the protagonist is not completely unsympathetic, in the course of the first chapter he confesses lying, cheating, and not paying the rent, and behaves strangely in a variety of ways (in a scene that caught the attention of many participants, he even follows two young ladies to their home, in what could almost be seen as an early instance of stalking). Here is an example of a passage where the narrator repeatedly lies to his interlocutor:

This was beginning to get interesting. The situation was running away with me, and one lie after another sprang up in my head. I sat down again, forgot about the paper and the remarkable documents, became excited and interrupted him when he spoke. The little dwarf’s gullibility made me reckless, I felt like stuffing him full of lies come what may, driving him from the field in grand style. (Hamsun 2001, 21-22)

Exchanges of this kind may warn the audience about the character’s potential unreliability as the narrator of his own story. After reading the first chapter of Hamsun’s novel (around 8,000 words), the students were asked to complete a questionnaire with both quantitative measures and (at the end) two open-ended questions about their attitude towards the protagonist. We divided the participants into two conditions: one read the original first-person text, whereas the other read a manipulated third-person version of the same chapter. In rewriting the text from the first to the third person, we left everything unchanged, except for one detail: since the narrator of Hamsun’s novel is anonymous, we had to give him a name in the third-person text; we opted for Henrik as a relatively common, and neutral-sounding, Scandinavian name. Here is how the passage quoted above would look in the manipulated version (the texts we used were in Dutch):

This was beginning to get interesting. The situation was running away with Henrik, and one lie after another sprang up in his head. He sat down again, forgot about the paper and the remarkable documents, became excited and interrupted the old fellow when he spoke. The little dwarf’s gullibility made Henrik reckless, he felt like stuffing his interlocutor full of lies come what may, driving him from the field in grand style.

The result is reminiscent of a stream of consciousness novel: as in later modernist classics (e.g., James Joyce’s A Portrait of the Artist as a Young Man, 1916), the narrative perspective is firmly focused on the protagonist’s thoughts and experiences, even as these thoughts and experiences are relayed by an external narrator.

Because of random assignment, 56% of participants received the first-person condition and 44% received the third-person condition. Of all participants, 79% indicated finishing reading the text within the allotted time, and one indicated having read the text before (this participant was in the original, first-person condition). Reading comprehension was measured on a five-point scale ranging from 1 (‘poor comprehension’) to 5 (‘excellent comprehension’). The results of this self-assessment (M = 3.16, SD = 0.91) did not differ significantly between conditions (all p’s between .26 and .38).


Before reading the experimentally manipulated text, participants indicated their age, sex, and reading expertise (defined as the number of novels read in the previous year). They also completed the Basic Empathy Scale (IRI; Jolliffe / Farrington 2006). This scale uses 11 items to assess dispositional affective empathy (e.g., “I get caught up in other people’s feelings easily”) and had good internal consistency (Cronbach’s α = .83). Dispositional cognitive empathy is measured with 9 items (e.g., “I can often understand how people are feeling even before they tell me”) and also had good internal consistency (Cronbach’s α = .77). Participants responded to all items on five-point Likert scales from “completely disagree” to “completely agree”. After reading the text, participants completed a manipulation check to ensure that they were aware of the narrative perspective (first-person perspective or third-person perspective). All participants answered this question correctly.

Subsequently, participants completed the quantitative outcome measures. Two three-item scales designed for this study were used to measure participants’ self-reported empathic concern for and perspective-taking with the character while reading the text. Empathic concern for the character had good internal consistency (Cronbach’s α = .85, e.g.: “Whilst reading the story I sympathized with the protagonist”). Internal consistency for perspective-taking with the character was negative, which suggests the three items might not form a unidimensional scale (a potential violation of tau-equivalence [cf. Sijtsma 2009]). This problem might be compounded by the small number of items, as alpha tends to increase with the number of items in a scale. Nevertheless, the scale correlated predictably with empathic concern for the character (r = .47, p < .001). The correlation was similar in size to that between dispositional affective and cognitive empathy (r = .40, p < .001), suggesting that the two scales (empathic concern and perspective-taking) measured different but related constructs. Another set of items, based on Larzelere and Huston’s (1980) “dyadic trust scale”, measured readers’ trust for the character and had good internal consistency (Cronbach’s α = .78, e.g.: “I feel that I can trust the narrator completely”). The questionnaire also included two open-ended questions, asking participants 1.) to comment on their attitude towards the character; and 2.) whether or not they trusted him, and why. We will turn to this qualitative part after examining the main findings of the experimental study.


A correlation table for the data is provided in Table 1. We used hierarchical linear regression to analyze the data. The baseline regression model included narrative perspective, dispositional affective and cognitive empathy, demographics (age and gender), and reading expertise. The effect of narrative perspective was analyzed using a dichotomous variable. Its effect size can be interpreted as the difference between the first-person condition and the third-person condition. To allow for the possibility that individual differences might influence the effect of the manipulation of narrative perspective, we explored whether the variance between conditions increased significantly when adding individual interactions of narrative perspective with trait empathy, demographics, or reading expertise, and interactions of trait empathy with demographics. In a second step, interactions that significantly increased this variance were added concurrently. For the sake of parsimony, the resulting model was pruned by removing non-significant effects as long as model fit remained unaffected. We explored significant interactions with a regions of significance approach, which reveals at what levels of a moderating variable the effect size of the predictor is significant (Preacher / Curran / Bauer 2006).

Empathic concern for the character

The best fitting model for empathic concern for the character did not include effects of narrative perspective, indicating that narrative perspective did not predict empathic concern for the character. Instead, the best fitting model included main effects of dispositional affective empathy and reading expertise, as well as an interaction between dispositional cognitive empathy and age (Table 2). Specifically, participants with greater dispositional affective empathy reported greater empathic concern for the protagonist. More experienced readers, on the other hand, reported less empathic concern for the character. Probing the interaction between dispositional cognitive empathy and age revealed that, although the effect of cognitive empathy on empathic concern for the character was stronger for younger participants than for older participants, it was positive and significant within the entire age range of our sample.

Table 1: Correlations between study variables. Correlations with condition (first, third) and sex (female, male) are polyserial.

Table 2: Summary of regression analyses

Perspective-taking for the character

The best fitting model for empathic concern for the character did not involve a main effect of narrative perspective, but did include an interaction between narrative perspective and age. The model further included main effects of dispositional cognitive empathy and reading expertise (Table 2). Participants with greater dispositional cognitive empathy reported greater perspective-taking for the protagonist. Participants who had greater reading expertise, however, reported less perspective-taking for the character. Probing the interaction between condition and age with a regions of significance approach revealed that the third-person perspective engendered less perspective-taking than the first-person perspective for participants older than 17.44, and was not significant for younger participants within the age range of our sample.

Trust for the character

Narrative perspective was the only significant predictor of trust for the character (Table 2). Specifically, reading the text in the third-person perspective significantly increased trust for the character compared to the first-person perspective.


The present study set out to investigate the effects of narrative perspective on readers’ engagement with characters in literary fiction. This engagement was seen here as a function of two factors: empathy (including both self-reported empathic concern and perspective-taking) and trust for the character. Contrary to our predictions, the results indicated that narrative perspective had no main effect on empathic concern and perspective-taking for the character: Only for older participants did narrative perspective influence the amount of perspective-taking for the character; these participants engaged in greater perspective-taking with the character when they read the text from a first-person perspective. This result is consistent with Kaufman and Libby’s (2012) prediction that first-person narratives is more conducive to experience-taking than third-person narratives. The fact that this effect was significant only for older adolescents is in line with developmental psychological research, which shows that mature perspective-taking abilities are still very much under development in adolescence (e.g., van Lissa et al. 2014; Blakemore / Choudhury 2006). Moreover, engaging in perspective-taking with complex literary characters may be even more demanding than perspective-taking in social situations with peers and parents. Perhaps effects of narrative perspective on participants’ perspective-taking started emerging only once participants were old enough to perceive similarities between themselves and the narrator (who is in his early twenties).

We found that narrative perspective does influence trust for the character, as readers in the third-person condition reported greater trust. Narrative unreliability is the prime suspect for this change in trust: readers in the first-person condition might have been more distrustful of the narrator. Because the narrator lies to some of the characters he interacts with in the storyworld, he could be perceived as manipulating his audience as well. Conversely, the narrator’s authoritativeness in the third-person text may have indirectly validated the character’s actions, translating into a higher degree of trust for him and not just for the narrator. This notion of ‘trust transfer’, of course, does not rule out the other explanation, based on the potential unreliability of first-person narrative. Only follow-up research can shed light on the exact mechanisms and causal relationships involved. What is clear is that questions surrounding trust and reliability should rank high in the priority list of empirically minded literary scholars, since these phenomena contribute significantly to the complexity of readers’ responses to characters (and literary narrative more generally).

Together with the interaction between perspective-taking and age in the first-person text, the main effect of narrative perspective on trust highlights the complexity of first-person narrative, which calls for two opposite interpretive strategies: on the one hand, engaging with a narrator who recounts his or her own experiences may create a sense of spontaneity or authenticity in the reading experience, thus potentially encouraging empathic responses; on the other hand, when the narrator as protagonist behaves in puzzling or unconventional ways, readers may be encouraged to question the trustworthiness of his or her narrative.

Another interesting finding was that the number of novels read over the preceding year negatively predicted empathic concern and perspective-taking for the protagonist. The question was phrased as follows: “How many novels have you read over the previous year (12 months)?” Thus, participants included in their answers both literary novels and more popular genres. A possible explanation is that experienced readers are more familiar with fictional characters (including the likable protagonists of popular fiction), and thus have learned to read in more distanced ways, paying attention to aspects of texts different from the protagonist’s predicament. Clearly, any attempt at using literature as a tool for developing real-world empathy in adolescents will have to carefully evaluate the role that literary competence may play in modulating readers’ attitude towards characters.

Finally, the study revealed that dispositional affective and cognitive empathy predicted empathic concern for the character, and dispositional cognitive empathy predicted perspective-taking for the character. This finding contradicts the results of a study by Sklar (2009), according to whom real-world empathic concern is not correlated with empathic concern for characters. According to Sklar, our engagement with characters is – to some extent – ‘cordoned off’ from everyday social interaction: fiction is thought to provide a safe harbor to experiment with worldviews and social stereotypes that we are reluctant to question in our day-to-day experience. As Keen puts it, “the perception of fictionality releases novel-readers from the normal state of alert suspicion of others’ motives that often acts as a barrier to empathy” (Keen 2007, 169). On this view, people would be more likely to feel empathic concern for an outsider like Hamsun’s character than for his many real-world counterparts. Our findings, however, go in the opposite direction. This discrepancy might be explained by the fact that Sklar used the empathic concern subscale of the IRI, which assesses empathic responses in social situations (cf. Davis 1983). The IRI also contains a fantasy subscale, which specifically refers to empathic responses to characters. Therefore, the measurement instrument itself seems to ‘cordon off’ engaging with characters from compassionate responses in daily life. By contrast, the measure of dispositional affective empathy (the IRI’s Basic Empathy Scale) we used was more general: it reflected the extent to which participants reported sharing others’ emotions, while cognitive empathy measured their awareness and understanding of others’ emotions. The tendency to share or understand others’ emotions might reflect more basic empathic predispositions that are not exclusive to actual social situations, and might therefore also generalize to empathy for characters.

Indeed, our study casts doubt on Sklar’s account by showing that participants’ self-reported dispositional empathy is in line with their situational empathy for the character. The text’s fictionality, therefore, does not make a fundamental difference when it comes to feeling empathy or sympathy for the protagonist: real-world intersubjectivity and readers’ responses to characters build on the same background of predispositions and assumptions (cf. Caracciolo 2013, 32-33). Effects of the kind posited by Sklar and Keen are theoretically possible, but they are by no means inevitable or automatic. This points to the importance of contextual factors in framing readers’ encounters with fictional texts (and beings): the perceived separateness of fictional characters would thus be an interpretive framework that can be activated in specific circumstances – for instance, when we are explicitly instructed to put ourselves in a character’s shoes or when the strangeness of Hamsun’s protagonist is embedded in discussions of social and ethical issues.

Qualitative questions

At the end of the questionnaire readers were asked to comment freely on their attitude towards the character, and on whether they trusted him or not (and why). The answers to these questions were fairly diverse in length, from single adjectives (e.g., “curious”, “surprised”, etc.) to paragraph-long descriptions. Three of the authors (Caracciolo, Van Duuren, and Van Leuveren) coded these answers, jointly annotating the corpus on the basis of the study’s goals and working through potential disagreements in an iterative process. This resulted in a set of 21 codes. Overall, the participants appeared to favor what James Phelan (1989) would call a “mimetic” stance towards the character, interpreting him in psychological terms – i.e., as if he were a lifelike being – without referring to style or literary themes. (Of course, this could reflect the fact that the qualitative questions already implied a psychologizing stance towards the protagonist.) The most frequent codes, with over 10 occurrences across all the answers, are listed and exemplified in Table 3. A complete list of codes is provided in Table 4 below.

Table 3: Main codes (over 10 occurrences). The frequencies show the number of participants whose answers were assigned each code at least once (so, for instance, if the same participant expressed ethical condemnation in different parts of the response, he or she still counts as one).

While negative judgments (unpredictability, unreliability, etc.) are clearly prevalent, sympathetic responses are not too uncommon. Since these codes exist at the level of individual words and sentences, a participant’s overall response to the qualitative items could contain both positive and negative attitudes. We thus decided to divide the participants into three categories. Most readers (44 out of 76) expressed an attitude towards the character that combines positive and negative elements. In this class the most frequent codes are: unpredictability (frequency 20), ethical condemnation (17), strangeness (15), positive attitude (14), and psychological evaluation (13). 28 out of 76 readers expressed a negative attitude towards the character. The most frequently occurring codes are ethical condemnation (15), strangeness (11), unpredictability (11), unreliability (10), and negative attitude (9). Finally, a small number of readers (4 out of 76) expressed a positive attitude towards the character. Here positive evaluations (4) and sympathy (4) are the most frequent themes.

The fact that over half of the participants had mixed feelings for the protagonist is, of course, an indicator of his complexity, and confirms what we suggested above about the ambivalence of readers’ engagement with characters in literary fiction: in relating to Hamsun’s protagonist, readers tend to oscillate between positive and negative attitudes, with a number of participants showing both in their commentaries. By contrast, feelings of closeness to the character appear only in a small minority of commentaries (with 5 occurrences in total). This dovetails with the idea that empathy for the protagonist was disrupted by other aspects of readers’ stance towards him.

We also compared the distribution of the codes across the two conditions (see Table 4), though these results are not completely reliable because the participants’ responses were not coded blind to the condition. The only statistically significant phenomenon that can be observed here is that the condition predicted reports of pity for the character: participants in the first-person condition were assigned the code pity more frequently (p = .03). Pity is an empathic emotion, in the sense that it refers to the emotional reactions of one person to the experiences of another (Davis 1983), with a negative valence. This observation suggests that the first-person perspective might indeed elicit stronger empathic emotions, albeit not the other-oriented, caring empathic responses (such as empathic concern or sympathy) that we attempted to measure with the quantitative items. This might be related to the character’s potential unlikability, and to the suspicion of narrative unreliability in the first-person context: hence, according to these qualitative results, the participants’ feelings towards the character-narrator were both more empathic and more negatively colored than those participants experienced in the third-person condition. A qualitative analysis of this kind is necessarily limited insofar as it is based on interpretive categories and decisions, but it does suggest hypotheses that may be tested in follow-up research.

Table 4: Frequency of codes per condition.


The main finding of the experimental study discussed in this article is that the narrative situation has an effect on readers’ trust for the protagonist of a literary text. This result suggests that, other things being equal, readers may be less inclined to trust a deviant character when he or she is also the narrator of the story – possibly, because of the awareness that he or she might be lying or deliberately manipulating them. When, in our study, the narrative perspective was shifted to third-person form with internal focalization, readers tended to consider the character more trustworthy even if his actions and thoughts remained the same. This discrepancy can be interpreted in light of the different real-world frames activated by first- and third-person narrative: since character narration is reminiscent of face-to-face conversation, the audience can develop an illusion of direct involvement in the communicative process (in what Monika Fludernik would call a “telling” frame, 1996, 50). This idea ties in with Kotovych et al.’s (2011) thesis that reading – and specifically reading texts with first-person narrators – activates inference-making processes broadly similar to those at work in everyday conversation. However, while Kotovych et al. focus on reliable narrators, this essay has shown that readers of fiction are equally likely to cast doubt on the narrator’s trustworthiness, just as in real life we may question our interlocutor’s reliability and ethical stance. By contrast, in third-person narrative readers’ access to the protagonist’s mental life is mediated (and authenticated) by an anonymous narrator who is not, apparently, part of the storyworld. This set-up has two consequences: first, the trust readers place in the external narrator (who is by convention an authoritative figure) may be transferred to their engagement with the protagonist, in what Kotovych et al. (2011) call an “association” between narrator and character; second, readers are invited to imagine the character from an observer position – which may result in feelings of distance and / or sympathy.

Moreover, our study indicates that first-person narrative and internal focalization are equally likely to trigger (or not trigger) empathic responses – except for the older participants in our sample, who displayed greater perspective-taking for the protagonist in the first-person condition. This result may reflect a closer ‘match’ not just in age but also in existential outlook between the narrator and older participants: students aged between 17 and 18 are about to finish school, and have to face some important life choices. Therefore, these participants may have appreciated the spontaneity of a first-person narrator who is living through a period of emotional turmoil and uncertainty, and may have more easily identified with his narratorial voice than with that of an anonymous, external narrator.

It may be argued that the results obtained here are specific to Hamsun’s text. Yet Hamsun’s protagonist is similar to many of the ‘round’ characters of the 19th and (especially) 20th century novel, and we would expect similar reading strategies in response to such characters. Regardless of the narrative perspective, most readers did not feel especially close to the protagonist of Hunger, as suggested by the qualitative commentaries. This finding is perhaps surprising, given the supposed centrality of empathic concern and perspective-taking in relating to fictional characters. On the other hand, it dovetails with Suzanne Keen’s (2007, 84) intuition that ‘lowbrow’ fiction (which usually features likable protagonists) tends to be more conducive to empathy for characters than literary fiction, in which characters are complex and often ethically ambivalent. Surely, there is more than empathy involved in engaging with a literary protagonist like Hamsun’s: a character’s deviation from societal norms is likely to disrupt empathy, inducing judgments of unpredictability, strangeness, and mental illness – all of which can be seen as a manifestation of his or her unconventional and challenging nature. A principled approach to literary character should account for these distancing effects, and should explore the interpretive strategies through which audiences cope with the character’s perceived strangeness (for a first attempt along these lines, cf. Caracciolo 2013).

Further research is needed to shed light on these issues, but the evidence presented here does question assumptions about the direct effects of textual strategies on narrative empathy. Indeed, one of the lessons that can be drawn from our experiment is that literary scholars tend to overestimate the effects of textual cues on readers’ responses: we had predicted changes in empathy as a result of narrative strategies (first-person vs. third-person form) while, in fact, our results show that empathy should be conceptualized as an emergent phenomenon depending on the interaction between textual and non-textual factors: readers’ age, reading expertise, and dispositional empathy.

We should keep in mind, however, that our study measured readers’ self-reported empathy as an outcome of reading a certain text. This does not necessarily reflect what happened while reading Hamsun’s chapter. Readers may have imaginatively entertained the protagonist’s viewpoint while reading even as other aspects of their experience eventually led them to distance themselves from the character. Indeed, the ambivalence of most readers’ commentaries suggests that empathy – in the sense of both perspective-taking and empathic concern or sympathy – may have been part of the readerly dynamic, but it was compromised by other factors in the audience’s overall evaluation of the protagonist. In order to explore these factors, we would need to examine more closely the temporal progression of readers’ engagement with literary characters. Obviously, the specifics of this progression will depend on the text used, but the structural tensions and dynamics underlying it may well be more generalizable.

A final point concerns the possible implications of our findings for the study of the real-world effects of literary reading. Our choice of a group of high-school students for our experiment was not coincidental. Perspective-taking is still malleable in teenagers, and it is legitimate to think that literary reading may help shape empathic competencies by exposing young readers to complex and often ambivalent social situations. Future research might therefore address longitudinal effects of reading literary fiction on adolescents’ empathy development. However, in order to fully understand these effects, we have to acknowledge that, in literary reading, empathy for characters is likely to be complicated, and may in some cases be counteracted, by other factors. This study has begun to chart readers’ attitude towards characters as a function of empathy, trust, and (through the qualitative answers) ethical judgments. The psychological impact of literary fiction can be properly explained only against this broader background – that is, by taking into account interactions across multiple aspects of readers’ engagement.


Blakemore, Sarah-Jayne / Choudhury, Suparna (2006): “Development of the Adolescent Brain: Implications for Executive Function and Social Cognition”. In: Journal of Child Psychology and Psychiatry 47 (3-4), pp. 296-312.

Caracciolo, Marco (2013): “Patterns of Cognitive Dissonance in Readers’ Engagement with Characters”. In: Enthymema VIII, pp. 21-37.

Caracciolo, Marco (2014): “Beyond Other Minds: Fictional Characters, Mental Simulation, and ‘Unnatural’ Experiences”. In: Journal of Narrative Theory 44 (1), pp. 29-53.

Caracciolo, Marco / Duuren, Thom van (2015): “Changed by Literature? A Critical Review of Psychological Research on the Effects of Reading Fiction”. In: Interdisciplinary Literary Studies 17 (4), pp. 517-39.

Cohn, Dorrit (1978): Transparent Minds: Narrative Modes for Presenting Consciousness in Fiction. Princeton.

Cook, Guy (1994): The Discourse of Literature: The Interplay of Form and Mind. Oxford.

Coplan, Amy (2004): “Empathic Engagement with Narrative Fictions”. In: The Journal of Aesthetics and Art Criticism 62 (2), pp. 141-52.

Davis, Mark H. (1983): “Measuring Individual Differences in Empathy: Evidence for a Multidimensional Approach”. In: Journal of Personality and Social Psychology 44 (1), pp. 113-26.

Doležel, Lubomír (1998). Heterocosmica: Fiction and Possible Worlds. Baltimore / London.

Eder, Jens (2006): “Ways of Being Close to Characters”. In: Film Studies 8 (1), pp. 68-80.

Eder, Jens / Jannidis, Fotis / Schneider, Ralf (2010) (Eds.), Characters in Fictional Worlds: Understanding Imaginary Beings in Literature, Film, and Other Media. Berlin / New York.

Fludernik, Monika (1996): Towards a “Natural” Narratology. London.

Forster, E. M. (1985): Aspects of the Novel. San Diego.

Gaut, Berys (1999): “Identification and Emotion in Narrative Film”. In: Carl Plantinga / Greg M. Smith (Eds.), Passionate Views: Film, Cognition, and Emotion. Baltimore / London, pp. 200-216.

Goldstein, Thalia R. / Winner, Ellen (2012): “Sympathy for a Character’s Plight: Sex Differences in Response to Theatre”. In: Empirical Studies in the Arts 30 (2), pp. 129-41.

Hakemulder, Frank (2000): The Moral Laboratory: Experiments Examining the Effects of Reading Literature on Social Perception and Moral Self-Concept. Amsterdam / Philadelphia.

Hamsun, Knut (2001): Hunger. Edinburgh. [Translated by Sverre Lyngstad].

Hawk, Skyler T. / Fischer, Agneta H / Kleef, Gerben A. Van (2011): “Taking Your Place or Matching Your Face: Two Paths to Empathic Embarrassment”. In: Emotion 11 (3), pp. 502-13.

Iser, Wolfgang (1978): The Act of Reading: A Theory of Aesthetic Response. Baltimore / London.

Jolliffe, Darrick / Farrington, David P. (2006): “Examining the Relationship Between Low Empathy and Bullying”. In: Aggressive Behavior 32 (6), pp. 540-50.

Kaufman, Geoff F. / Libby, Lisa K. (2012): “Changing Beliefs and Behavior through Experience-Taking”. In: Journal of Personality and Social Psychology 103 (1), pp. 1-19.

Keen, Suzanne (2007): Empathy and the Novel. Oxford / New York.

Keen, Suzanne (2013): “Narrative Empathy”. In: Peter Hühn (Ed.), The Living Handbook of Narratology. Hamburg. URL: (20.04.2016).

Kidd, David Comer / Castano, Emanuele (2013): “Reading Literary Fiction Improves Theory of Mind”. In: Science 342 (6156), pp. 377-80.

Korthals Altes, Liesbeth (2014): Ethos and Narrative Interpretation: The Negotiation of Values in Narrative Fiction. Lincoln / London.

Kotovych, Maria / Dixon, Peter / Bortolussi, Marisa / Holden, Mark (2011): “Textual Determinants of a Component of Literary Identification”. In: Scientific Study of Literature 1 (2), pp. 260-91.

Larzelere, Robert E. / Huston, Ted L. (1980): “The Dyadic Trust Scale: Toward Understanding Interpersonal Trust in Close Relationships”. In: Journal of Marriage and the Family 42 (3), p. 595-604.

Lissa, Caspar J. van et al. (2014): “The Longitudinal Interplay of Affective and Cognitive Empathy Within and Between Adolescents and Mothers”. In: Developmental Psychology 50 (4), pp. 1219-25.

Mar, Raymond A. et al. (2006): “Bookworms Versus Nerds: Exposure to Fiction Versus Non-Fiction, Divergent Associations with Social Ability, and the Simulation of Fictional Social Worlds”. In: Journal of Research in Personality 40 (5), pp. 694-712.

Mellmann, Katja (2010): “Objects of ‘Empathy’: Characters (and Other Such Things) as Psycho-Poetic Effects”. In: Jens Eder / Fotis Jannidis / Ralf Schneider (Eds.), Characters in Fictional Worlds: Understanding Imaginary Beings in Literature, Film, and Other Media. Berlin / New York. pp. 416-41.

Nabokov, Vladimir (1997): Lolita. New York.

Nünning, Ansgar (1999): “Unreliable, Compared to What? Towards a Cognitive Theory of Unreliable Narration: Prolegomena and Hypotheses”. In: Walter Grünzweig / Andreas Solbach (Eds.) Grenzüberschreitungen: Narratologie Im Kontext / Transcending Boundaries: Narratology in Context. Tübingen. pp. 53-73.

Peer, Willie van / Maat, Henk Pander (1996). “Perspectivation and Sympathy: Effects of Narrative Point of View”. In: Roger J. Kreuz / Mary Sue MacNealy (Eds.), Empirical Approaches to Literature and Aesthetics. Norwood, NJ. pp. 143-56.

Phelan, James (1989): Reading People, Reading Plots: Character, Progression, and the Interpretation of Narrative. Chicago.

Phelan, James (2005): Living to Tell About It: A Rhetoric and Ethics of Character Narration. Ithaca.

Preacher, Kristopher J. / Curran, Patrick J / Bauer, Daniel J. (2006): “Computational Tools for Probing Interactions in Multiple Linear Regression, Multilevel Modeling, and Latent Curve Analysis”. In: Journal of Educational and Behavioral Statistics 31 (4), pp. 437-48.

Preston, Stephanie D. / Waal, Frans B. M. de (2002): “Empathy: Its Ultimate and Proximate Bases”. In: Behavioral and Brain Sciences 25 (1), pp. 1-20.

Shen, Dan (2013): “Unreliability”. In: Peter Hühn (Ed.), The Living Handbook of Narratology. Hamburg. URL: (20.04.2016).

Shklovsky, Viktor (1965): “Art as Technique”. In: Lee T. Lemon / Marion J. Reis (Eds.), Russian Formalist Criticism: Four Essays. Lincoln and London, pp. 3-24.

Sijtsma, Klaas (2009): “On the Use, the Misuse, and the Very Limited Usefulness of Cronbach’s Alpha”. In: Psychometrika 74 (1), pp. 107-20.

Sklar, H. (2009): “Narrative Structuring of Sympathetic Response: Theoretical and Empirical Approaches to Toni Cade Bambara’s ‘The Hammer Man’”. In: Poetics Today 30 (3), pp. 561-607.

Tan, Ed S. (1996): Emotion and the Structure of Narrative Film: Film as an Emotion Machine. Mahwah, NJ.

Wood, James (2010): How Fiction Works. New York.

Yacobi, Tamar (1981): “Fictional Reliability as a Communicative Problem”. In: Poetics Today 2 (2), pp. 113-26.

Zillmann, Dolf. (1991): “Empathy: Affect from Bearing Witness to the Emotions of Others”. In: Jennings Bryant / Dolf Zillmann (Eds.): Responding to the Screen: Reception and Reaction Processes. Mahwah, NJ. pp. 135–67.

Caspar J. van Lissa
Erasmus University Rotterdam
The Netherlands

Marco Caracciolo
University of Freiburg

Thom van Duuren
University of Groningen
The Netherlands

Bram van Leuveren
University of St Andrews
United Kingdom

Please do not cite the HTML version but only the PDF file:

URN: urn:nbn:de:hbz:468-20160607-151127-7

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.