What I Can Do with the Right Version of You: The Impact of Narrative Perspective on Reader Immersion, and How (in)Formal Address Pronouns Influence Immersion Reports

Abstract

Previous research has shown that readers experience stronger immersion while reading first-person narratives than third-person narratives, but whether this difference in processing is time-sensitive remains unclear. We report on a self-paced reading experiment in which French participants read short stories in first and in third person. Additionally, participants were directly addressed with either formal or informal second-person pronouns in the final sentence of the narrative, as well as in a subsequent (off-line) questionnaire soliciting post-story immersion reports. The suitability of a particular pronoun of address depends on the social context, and misplaced use can impact the way in which people perceive and assess a particular situation. We did not find significant differences between reading times with first- or third-person pronouns, but participants reported higher immersion and emotional engagement after reading first-person stories than third-person stories. Moreover, the effect of story perspective on reported immersion only occurred when participants were addressed with formal second-person pronouns. We take these findings as evidence for an effect of first- versus third-person pronouns on immersion via perspective shifts. Moreover, we argue that the effect of being addressed with an unexpected (in this case, informal) pronoun can overrule such an effect, while being addressed with an expected (in this case, formal) pronoun can be conducive of it. This finding highlights the importance of research into the impact of pronouns of address.

1. Introduction

Readers are said to be immersed in a story when they experience it as if they were part of it. Immersion may be defined as a “state of cognitive, emotional, and imaginative absorption” within the story world (Hartung et al. 2017b, p. 2). The experience of immersion is not identical for everyone; what immersion entails largely depends on someone’s personal preferences (Hartung et al. 2017a) and their disposition to imagination and empathy (Hartung et al. 2016, 2017b; Mak and Willems 2019). Reading habits also contribute to what immersion looks like for each reader: more habitual readers are generally able to experience immersion more deeply and with more ease than those who read sparingly (Mak et al. 2020). Despite the variety of immersive states, researchers agree on what immersion generally looks like: readers mentally re-enact described actions and emotions, create images in their mind, feel transported into the story world, and/or think along with the characters of a story (Green and Brock 2000; Shanton and Goldman 2010; Kuijpers et al. 2014).

Moreover, the experience of immersion varies for an individual depending on the actual reading material at hand (Kuijpers et al. 2014; Hartung et al. 2017b; van Krieken et al. 2017): an instruction manual will probably immerse a reader much less than a fantasy novel. It is generally assumed that the narrative genre promotes immersion with much more ease than other genres because of a perceived similarity with narrative characters, the use of suspense, and view-point techniques. That is, a key feature of narratives that modulates the experience of immersion is the narrative perspective (see Sanders and Redeker 1996).

Narratives are commonly written from a third-person perspective, with an omniscient narrator who is external to the story telling the reader about what the characters feel and do. Narratives may also be written from a first-person perspective, where the narrator is typically one of the characters describing events from his or her own limited perspective. Narrative points of view, and the corresponding linguistic cues, have been shown to impact readers’ perspective-taking: reading in the third person elicits an external or onlooker’s perspective, while reading in the first person prompts the reader to adopt an internal or agent perspective (Brunyé et al. 2009; Ditman et al. 2010). That is, in reading a sentence like “I cut a tomato”, the reader imagines themselves doing the cutting; when the sentence is “He/she cuts a tomato”, the reader feels as if they see someone else cutting it.

Crucially, assuming a point of view while reading is linked to immersion into the story: an experimental study by Hartung et al. (2016) investigating the effect of perspective-taking on immersion reports that reading Dutch short stories in first (versus third) person resulted in readers reporting higher levels of immersion. They also measured the electrodermal activity (EDA) of the participants while reading. This method provides a measure of a person’s arousal through small changes in the glands in the skin. They found increased arousal through EDA while people read third-person stories rather than first-person stories. Hartung et al. interpret these results by assuming that third-person stories involve a higher cognitive demand than first-person stories: in third-person stories, readers may adopt the points of view of multiple characters and, thus, anticipate having to do so even when it is unnecessary for the understanding of the story. The increased processing load of reading third-person perspective results in a less immersive experience for the reader. Moreover, a subsequent study by Hartung et al. (2017b) reports that first-person stories generate higher levels of emotional engagement with the characters than third-person stories. Reading times were measured so as to explore whether the increase in processing load is reflected in reading behavior, but no significant differences between narrative perspectives were found.

In the present study, we reinvestigate the effect of first- versus third-person perspective on immersion in French short stories by means of a self-paced reading experiment and a post-hoc questionnaire collecting reports of the immersive state of participants. We suspect that the measures used in Hartung et al. (2017b) were not sensitive enough to capture an effect of the pronoun type, as they measured the time spent on the full stories, i.e., the time from the onset of the full-text fragment presentation until the button press. In our experiment, we instead measure reading times for the pronouns themselves (and the spillover regions). Thus, we collect a measure that reflects the online processing cost of first- and third-person pronouns (in the short story), as well as an off-line measure that reflects participants’ self-reported post-hoc immersion in the story (in the questionnaire). Our hypotheses are as follows:

H1: 

Third-person pronouns are read more slowly than first-person pronouns.

H2: 

First-person short stories lead to higher reader immersion reports than third-person short stories.

Narratives written from a second-person perspective are traditionally less common (see Fludernik 1994), and their effect on immersion remains understudied. However, we will not include second-person narratives in our study because of the referential complexity of the second person in narratives (it is not easy to determine who the intended addressee is), and the fact that many languages have a bipartite or tripartite second-person pronominal system. An investigation of second-person narratives deserves its own in-depth study.

Still, there is recent interest in the effects of second-person pronouns on processing, particularly when a second-person pronoun is used to address the interlocutor directly. Pronouns of address have been found to affect people’s attitudes to, evaluations of, and involvement with real and experimental material. Cruz et al. (2017), for example, investigated social media users’ involvement with online ads in English. They found that when ads contained a form of the second-person pronoun you, users were more likely to interact (i.e., like, share, and comment) with the ad, compared to ads without any pronouns of address. This might be due to the fact that pronouns of address have a self-ascriptive ability (Wechsler 2010; de Hoop and Tarenskeen 2015), where addressees are more ready to interpret the pronoun as a deictic rather than a generic reference.

For languages that use two or more second-person pronouns (e.g., Dutch jij and u, German Sie and du, French tu and vous), additional factors come into play revolving around the particular second-person pronoun. Such languages present speakers with a choice not based on grammatical rules, but on social norms. Traditionally, constructs such as power, solidarity, respect, and distance were applied to explain the variation in the use of these pronouns of address. Pronouns of address are usually categorized as formal (or V, from Latin vos) or informal (or T, from Latin tu) (Brown and Gilman 1960). Several studies looked into how Netherlandic Dutch speakers’ attitudes towards commercial and corporate communications are affected by the use of a T (jij) or a V (u) pronoun of address (van Zalk and Jansen 2004; Leung et al. 2022; de Hoop et al. 2023; Sadowski et al. 2024; Schoenmakers et al. 2024). There is a general preference for T pronouns among speakers of Dutch in The Netherlands (Vismans 2013; Levshina 2017), which led researchers to hypothesize that consumers’ evaluations would be more positive when addressed with T than V pronouns. Indeed, Leung et al. (2022) found that Dutch consumers prefer to be addressed with T forms in commercial communications rather than V forms in general. Schoenmakers et al. (2024) found that T-use in commercial advertising leads to a higher appreciation of product advertisements than V, and Sadowski et al. (2024) found that T-use leads to more willingness to make monetary donations than V-use, particularly for altruistic participants. Van Zalk and Jansen (2004), by contrast, found that Dutch speakers were more positive about an online travel advertisement when it contained V pronouns than T pronouns. However, a post-hoc analysis revealed that, while younger people evaluated both ads with T and V pronouns similarly, older people preferred V pronouns. Schoenmakers et al. (2024) and van Zalk and Jansen (2004) also found that being highly involved with the material (e.g., the scenarios, the ads, or the brands depicted in the experiment are “effectively relevant to the receiver”, Zaichkowsky 1986) results in higher appreciation and more positive evaluations regardless of the pronouns of address received.

German T (du) and V (Sie) pronouns have also been shown to differentially affect people’s attitudes and evaluations (den Hartog et al., forthcoming). In German, there seems to be no clear-cut preference for T or V (Kretzenbacher et al. 2006; den Hartog et al. 2022). Den Hartog et al. (forthcoming) designed an online (mock) job interview, where participants (in the role of candidates) saw a series of videos of a recruiter asking them questions while addressing the candidate with either T or V. German participants evaluated the recruiter more positively when they had used V to address participants. In the case of French, the language under investigation in the present paper, the default pronoun used for public communications is the formal vous ‘you’ (den Hartog et al. 2022). Meanwhile, tu is the marked pronoun, restricted to (perceived) intimate contexts where interlocutors know each other well or at least experience a certain sense of (social) proximity (Warren 2006; Pager-McClymont et al. 2023). The study of den Hartog et al. (forthcoming) also tested French participants taking part in a mock job interview: French speakers were significantly more critical in their evaluation of the recruiter when they received T rather than V. The use of T-pronouns by a stranger in a formal situation was perceived as too intimate, or even disrespectful, and was consequently deemed inappropriate.

The studies discussed above provide evidence that speakers of Dutch, German, and French are sensitive to being directly addressed with T- or V-pronouns, and that this sensitiveness is reflected in their evaluations of real and experimental material, as well as their interlocutor. Whether we choose to address our participants with T or V forms may thus impact their attitude towards the task, the materials, and/or the researcher. The effect also depends on the language under investigation, likely driven by differences in cultural practices.

The social norms in France seem to be quite strict in what is an acceptable and unacceptable use of a T-pronoun, as evidenced by the results of den Hartog et al. (forthcoming). Moreover, research into the choice of T/V by French speakers also highlights that speakers negatively judge a violation of the address practices. French informants in focus groups revealed to Warren (2006) that they felt offended or embarrassed when addressed with tu, or that the speaker was unilaterally trying to establish an undesired relationship by addressing them with the T-form. The informants’ testimonies show that speakers who break the social norms associated with the use of pronouns of address are frowned upon. This is supported by the findings from Pager-McClymont et al. (2023), who conducted a survey in which many participants evaluated T-use by a French news channel on their social media platforms as odd because of the “false familiarity” and forced intimacy.

Our investigation is, therefore, not limited to testing putative effects of narrative perspective (first- versus third-person) on immersion, but extends to putative effects of pronouns of address (T versus V) on readers’ (post-hoc) reports of immersion. To that end, we conducted an experiment consisting of a self-paced reading and a questionnaire component. First, we presented French readers with short stories in first or third person. Next, we asked the participants to fill out a questionnaire about their immersion in the story. The questionnaire was either presented in T or in V. Thus, we collected a measure that reflects the on-line processing cost of first- and third-person pronouns (in the short story), as well as an off-line measure that reflects participants’ self-reported immersion in the story (in the questionnaire).

Regarding the T-/V-manipulation in the questionnaire, we expect to observe a difference in the readers’ reported immersion based on the pronoun of address they received (T or V). We expect that French readers have a preference for V-pronouns of address, and judge V-pronouns as appropriate in the context of an experimental setting with unfamiliar researchers. Consequently, receiving a pronoun of address that conforms to readers’ expectations may translate to more appreciation for the task and, by extension, we expect the immersion reports to be higher.1 Conversely, the use of the T-pronoun may be perceived as inappropriate and could result in a lower level of engagement with the experimental material and a diminished post-hoc sense of immersion (i.e., being confronted with an inappropriate address pronoun in the questionnaire following the reading task “kills the vibe” from immersion), leading to reduced immersion reports. Our hypothesis is as follows:

H3: 

T-pronouns will lead to poorer reader immersion reports than V-pronouns.

We see no reason to expect the T-/V-manipulation to yield a different effect based on whether it follows a first- or a third-person story, and so we remain agnostic about a potential interaction effect. Finally, to exploratorily obtain an index of the processing difficulty associated with formal and informal second-person pronouns, participants were directly addressed directly in the final sentence of the story (using the French version of the frame “And you, have you ever …?”). Participants were presented with the same pronoun of address as they would see in the questionnaire. We expect that the T-forms will be read more slowly than the V-forms because of a surprisal effect due to the use of a non-default or even inappropriate form (Warren 2006; Pager-McClymont et al. 2023).

H4: 

T-pronouns are read more slowly than V-pronouns.

2. Materials and Methods

2.1. Participants

We recruited 111 French speakers from the faculty of Legal, Political, and Social Sciences at the University of Lille (42 male, 67 female, 2 non-binary) between the ages of 17 and 24 (M = 19.2, SD = 1.43). Participants were not aware of the purpose of the study. Participation was voluntary and participants received a snack as a reward. All participants were spoken to, and provided informed consent, in English. We discuss the exclusion criteria in Section 2.5.

2.2. Design

This experiment had a 2 (narrative perspective; first or third person) × 2 (pronoun of address; T or V) mixed design. All participants read one story in the first person and one story in the third person (within subjects) but received a single pronoun of address, either T or V (between subjects). For any particular participant, the pronouns used in the final question of the story and in the questionnaire were the same (viz. tu or vous). Two main measures were recorded: reading times at each region of interest (the pronouns) and spillover regions (the two words following each region of interest), and immersion scores (in the questionnaire). Story appreciation, individual reading habits, and a French version of the Author Recognition Test (Stanovich and West 1989) were included as control measures. This study received approval from the Ethics Assessment Committee Humanities of Radboud University (EACH file number 2021-3221).

2.3. Materials

All our stimuli and questionnaires are available via our repository at Supplementary Materials https://doi.org/10.34973/gzx3-yk10 (last accessed on 23 July 2024).

2.3.1. Stories

Two French fictional short stories were selected (see

Table 1. Story information.

Title, Author (Year)	Nr. of Words	Protagonist Gender	Nr. of Target Words	Nr. of Participants	Plot Summary	“And You?” Question (V-Version) (Target Word; Spillover)
Péripéties nocturnes (Barreyre 2019)	517	F	45	96	The protagonist, the tooth fairy (petite souris, lit. ‘little mouse’), is being followed by an owl and must reach the children’s bedroom to succeed at her mission.	Et vous, est-ce que vous avez déjà reçu la visite de la petite souris? (‘And you, have you ever been visited by the tooth fairy?’)
Les grandes découvertes perdues (D. 2014)	415	M	34	86	A geneticist at NASA invented a way to teleport anywhere into the universe.	Et vous, est-ce que vous aimeriez pouvoir te téléporter? (‘And you, would you like to be able to teleport yourself?’)

) from the open-access website of the publisher Short Édition (2011), who granted us permission to use the stories in our experiment. A fragment of one of the stimulus stories can be found in Appendix C. The original stories can be accessed through the Short Édition website and the modified versions through our repository. The stories were chosen based on word length (approximately 500 words) and the gender of the protagonist (one male and one female). Both stories have one protagonist, a single plotline, and do not touch on sensitive subjects. Neither of the stories made use of direct speech, and both originally referred to the protagonist with third-person pronouns. To manipulate the perspective of the stories, we produced a second version of each, referring to the protagonist with first-person pronouns (i.e., modifying the personal pronouns and the corresponding verbs).

Both stories were written in the past tense. This is different from the stimuli in Brunyé et al.’s (2009, 2011, 2016) and Ditman et al.’s (2010) experiments (“I am slicing the tomato”), which were presented in the present tense, even though literary narratives typically use the past tense, which is considered a narrative tense (de Swart 2007; Mulder et al. 2022). The experiments that tested longer story stimuli, rather than single sentences, were not specific about their tense use: the example from a longer text passage used in Brunyé et al. (2011) was in the present tense, and one of two examples in Brunyé et al. (2016) was in the present tense, the other in the past tense.

Furthermore, we added a final sentence to the short stories in which the reader is directly addressed. These sentences took the form of a question (e.g., and you, have you ever…?), beginning with Et vous ‘And you’ (formal) or Et toi ‘And you’ (informal). Earlier experiments used descriptive language to advance the story, although this type of language does not support the self-ascriptive effect of the second person ’you’ as much as interactive language (imperatives, questions) does (de Hoop and Tarenskeen 2015). In descriptive passages, de Hoop and Tarenskeen (2015) find that ’you’ is more often than not interpreted as a generic ’you’ rather than a deictic pointing to the existence of an addressee. Therefore, the final sentence of the short stories we used in our experiment, which directly addressed the reader, was an interrogative clause.

The regions of interest in the main part of the short stories were the first- and third-person pronouns (target words) and the two words following the target words, to control for spillover effects (e.g., elle arrive enfin ‘she finally arrives’).2 Target words and their spillovers were excluded from the analysis when they were the first or last word on the page, and when the target word contracted with the previous or following word (e.g., j’arrive ‘I arrive’). We also excluded a spillover when it was another pronoun (e.g., je me retournai ‘I turned myself’). In the end, one story contained 45 pronouns, while the other contained 34 (see Table 1). We did not control for grammatical function or position in the sentence, as these were the same in the first- and third-person condition.

As an offline measure of immersion (one of our main dependent variables), and other factors related to immersion, we made use of several questionnaires, which we discuss in the next subsection. When T-/V-pronouns were used in these questionnaires, the pronoun was manipulated as a between-subjects variable (in line with the pronoun read in the self-paced reading component). Hartung et al. (2016, 2017b) showed that psychological aspects, such as the ability to empathize and a disposition to imagination, explained much of the variance in their results. Assuming that our sample would be as varied in this respect, we recorded our participants’ reading habits and preferences as well to make sure that potential patterns observed in our data were not due to uniformity of our sample (university students).

2.3.2. Immersion Questionnaire

An adapted version of the Story World Absorption Scale (SWAS, Kuijpers et al. 2014) and the Narrative Engagement Scale (NES, Busselle and Bilandzic 2009) was used to measure immersion, containing 19 items in total. The full questionnaire can be found in Appendix A.

The SWAS was specifically designed to measure the subjective experience of absorption in the story world of narrative texts. It is divided into four subscales, each with their own set of three-to-five statements: attention (e.g., “When I was reading the story, I was focused on what happened in the story”), transportation (e.g., “The world of the story sometimes felt closer to me than the world around me”), emotional engagement (e.g., “I felt sympathy for the main character”), and mental imagery (e.g., “I could imagine what the world in which the story took place looked like”). The SWAS lacks a dimension about participants’ ease of understanding the story, and so we added a narrative understanding scale from the NES (e.g., “I could easily follow the thread of the story”), since understanding is a prerequisite for immersion (Busselle and Bilandzic 2009; Hartung et al. 2016).

After removing several of the statements in the scales to avoid redundancy, we translated them into French and rephrased them with second-person pronouns (T/V, see Appendix A). Six of the selected nineteen statements were reverse-coded to ensure that participants continued to pay attention to their responses. Participants responded to the statements on 100-point slider scales, with the anchors Je ne suis pas du tout d’accord ‘I don’t agree at all’ and Je suis tout à fait d’accord ‘I agree completely’. The corresponding numerical scores were not visible to the participants so that they had to rely on their intuition. Participants were forced to move the slider before moving to the next question.

2.3.3. Appreciation Questions

Variation in story appreciation was measured through the question Que pensez-vous/penses-tu de cette histoire? ‘What do you [V or T] think of the story?’. Participants responded to this question on five semantic binary questions: mauvaise–brilliante ‘bad–brilliant’, ennuyeuse–captivante ‘boring–captivating’, pas originale–originale ‘unoriginal–original’, nulle–drôle ‘lame–funny’, and mal écrite–bien écrite ‘badly written–well written’. These questions are included because a correlation has frequently been found between appreciation and immersion: readers who experience higher degrees of immersion also like the stories better than readers who experience less immersion (e.g., Kuijpers et al. 2014; Hartung et al. 2016; Mak and Willems 2019).

2.3.4. Reading Habits Questionnaire

To get an indication of participants’ individual reading habits, we compiled a questionnaire based on Hartung et al. (2016) that consisted of five questions: two questions asked about reading frequency, two about genre preferences, and one about the preferred medium for reading (i.e., book or e-reader). With this questionnaire, we can capture variability between participants due to their individual reading habits. The full questionnaire can be found in Appendix B.

2.3.5. Author Recognition Test

To obtain a more objective measure of reading exposure than self-reported habits, we created a French version of the Author Recognition Test (originally developed for American English by Stanovich and West 1989). We conducted a pre-test in which 15 French participants categorized a list of 78 names, of which 60 were existing authors and 12 were foils, in three categories: Je connais cet auteur ‘I know this author’, J’ai entendu parler de cet auteur ‘I’ve heard of this author’, and Je ne connais pas cet auteur ‘I don’t know this author’. Based on the responses, we compiled a list of 42 names. This list consisted of 10 authors from each category (well-known, somewhat known, and unknown names) and 12 foils. The foils were created by combining a first and last name from the top 500 most common names in France according to Forebears (2014). These names were checked to make sure they were not coincidentally the names of real authors. The number of authors participants recognize serves as an indication of how much they read (Stanovich and West 1989; Acheson et al. 2008; Mar and Rain 2015) and as a control for reading frequency, in addition to participants’ self-reported reading habits.

2.3.6. Content Questions

To ensure that the participants paid attention during reading, content questions were added to the questionnaire. For each story, three content statements were included, which participants could answer with vrai ‘true’ or faux ‘false’.

2.4. Procedure

The experiment was conducted on the campus of the faculty of Legal, Political, and Social Sciences of the University of Lille. Information and consent documents were provided to participants in English to avoid exposing them to T-/V-pronouns prior to the experiment. Other instructions provided in French during the experiment avoided the use of T/V. The self-paced reading task was presented on a laptop using the software package PsychoPy (Version 2022.2.2; Peirce et al. 2019). A moving window mechanism was used that allowed the stimuli to be revealed on a word-by-word basis, from left to right, and with masking lines that provided an indication of word length, punctuation, and sentence structure. Participants advanced through the text by pressing the spacebar at their own pace. Reading times for each word were recorded. At the end of the story, and still in self-paced reading mode, the final question Et vous?/Et toi? ‘And you?’ was directed at the reader. The reading times for the words in this sentence were also recorded but stored separately for the purpose of analysis. Participants did not have to answer this question.

Immediately after reading the story, participants answered the content questions and filled in the immersion and appreciation questionnaires. The procedure was the same for the second story. Each participant saw both stories: one in the first person and the other in the third person. The order of the story versions was randomized and counterbalanced across conditions. In the last part of the experiment, participants filled out a questionnaire concerning demographics, general reading habits, and French ART, which was constructed and presented in Qualtrics (2022).

2.5. Data Exclusion

Data from two participants were discarded because they did not finish the full experiment. Data of two more participants were discarded because they were exposed to the wrong demographic and reading habits questions. Data from 11 participants were excluded because they were non-native speakers of French. This left us with a total of 96 participants (38 male, 56 female, and 2 non-binary) between the ages of 17 and 24 (M = 19.1, SD = 1.39).

Each participant read two stories and answered three content questions about each story. These content questions were designed to check if participants paid attention when reading. If more than one out of the three questions per story was answered incorrectly, all data from the participants for that story were excluded from the analysis. This resulted in the deletion of 7.1% of the participant data.

We removed reading times below 50 ms or above 3000 ms as these outliers were implausibly low or high. Datapoints were also removed when the reading time differed over 2.5 SD from the participant or item mean. Following these deletions, with regard to the main part of the story, 7.04% of the data points for the target word (first- or third-person pronoun), 6.04% of the data points for the first spillover, and 6.40% of the data points for the second spillover were removed. With regard to the direct address question data, 3.91% of the data points for the target word (T- or V-pronoun), 3.91% of the data points for the first spillover, and 3.35% of the data points for the second spillover were removed.

2.6. Data Analysis

All data were processed and analyzed in R (Version 4.2.2; R Core Team 2022). We calculated Cronbach’s alpha values as a measure of reliability for each construct in our immersion questionnaire. At α = 0.447, the mental imagery scale proved insufficient (generally an alpha of 0.7 or higher is considered acceptable; Tavakol and Dennick 2011), thus mental imagery was excluded from the analyses. The immersion score was consequently based on the constructs attention, transportation, emotional engagement, and narrative understanding.

Our dependent variables were the total immersion scores and the reading times of the regions of interest (averaged over all instances of the pronouns and their spillover). The reading time data were log-transformed before analysis. For a more detailed understanding of which components of immersion are affected by the use of narrative perspective and pronouns of address, in addition to the total score analyses, we also performed separate analyses for each individual dimension of immersion. The processed answers on the general reading habits questionnaire, the Author Recognition Test, and the appreciation questions were used as measures of individual differences.

We used the R-package lme4 (Bates et al. 2015) to construct multiple linear mixed-effects regression analyses for (total) immersion, reading times of the narrative perspective condition, and reading times of T/V. For the immersion scores, each base model included the fixed effects of perspective and T/V, as well as the interaction between them. By-participant, by-story, and by-question varying intercepts were included as random effects. For the reading times of perspective, each base model included a perspective as a fixed effect, and by-participant, by-story, and by-item (the pronoun in the story) varying intercepts as random effects. For the reading times of T/V, each base model included the fixed effects of perspective, T/V, and mean immersion, and the interaction between perspective and T/V. By-participant and by-story varying intercepts were included as random effects. To each of these base models, other potentially relevant factors (covariates) were added one by one. A covariate was only included in the model if it improved the model fit. Models were compared using ANOVA comparisons in the R-package lmerTest (Kuznetsova et al. 2017) using Satterthwaite’s method. We discuss the final models in the sections below.

We contrast-coded the variables perspective, T/V, and the question on preferred materials in the reading habits questionnaire (Question 5) as [–0.5, +0.5]. We used the R-package emmeans (Lenth 2022) for the comparison of the conditions in the interaction if it was significant. p-values were then Bonferroni-corrected for multiple comparisons. We used the R-package ggplot2 (Wickham 2016) to visualize the data.

3. Results

3.1. Reading Times

Three analyses were conducted on the reading times for perspective (first versus third person): one for the target words (the pronouns), one for the first words after the pronouns (first spillover), and one for the second words after the pronouns (second spillover). Figure 1 shows the mean reading times and standard errors per perspective. Differences can be observed between first-person pronouns and third-person pronouns in the first spillover region (target word: M₁ = 442.6, SD₁ = 196.7; M₃ = 444.4, SD₃ = 191.7; first spillover M₁ = 409.8, SD₁ = 136.6; M₃ = 402.7, SD₃ = 128.0; second spillover M₁ = 405.4, SD₁ = 143.8; M₃ = 407.4, SD₃ = 150.9). Stories with third-person pronouns appear to be read faster than stories with first-person pronouns. However, we did not find significant main effects of perspective on the reading times of the target word (β = 0.006, SE = 0.021, t = 0.257, p = 0.797), first spillover (β = −0.012, SE = 0.017, t = −0.744, p = 0.458), nor second spillover (β = −0.002, SE = 0.0017, t = −0.53, p = 0.913). Further, we do not observe effects of the covariate appreciation on reading times in any regions of interest.

Next, we constructed six models for the analysis of T-/V-effects on reading times: one for each of the two target words (the pronouns) in the question (viz. Et tu/vous, est-ce que tu/vous as/avez…? ‘And you, have you…’) and one for each of the pronouns’ first and second spillovers. We constructed separate models for the first and second occurrence of a pronoun because the question starts by directly addressing the participant directly after the short story, which was written in first or third person. The surprisal effect may thus be larger for the first occurrence of a pronoun of address. Figure 2 shows the mean reading times and standard errors of the two target words in the question and their two spillovers. The reading times are lower for V (M = 347.3, SD = 100.8) than for T (M = 354.1, SD = 82.8), except in the first spillover of the first pronoun.

None of the effects entered into the models were significant, see

Table 2. Model specifications of the reading time data for the two target words (pronouns) and their two spillovers in the Et tu/vous, est-ce que tu/vous…? ‘And you, have you…?’-question per T/V. Interaction effects are marked with *.

Pronoun	Region	Effect	β	SE	t	p
First pronoun	Pronoun	T/V	−0.04	0.04	−1.04	0.303
		Perspective	0.03	0.03	1.11	0.272
		Perspective * T/V	0.07	0.05	1.25	0.214
		Mean Immersion	−0.00	0.00	−0.37	0.710
	First spillover	T/V	−0.01	0.06	−0.24	0.812
		Perspective	−0.03	0.03	−0.91	0.364
		Perspective * T/V	0.00	0.07	0.05	0.961
		Mean Immersion	0.00	0.00	−0.33	0.742
	Second spillover	T/V	−0.06	0.05	−1.15	0.253
		Perspective	0.03	0.03	1.02	0.311
		Perspective * T/V	0.04	0.07	0.55	0.586
		Mean Immersion	−0.00	0.00	−0.70	0.488
Second pronoun	Pronoun	T/V	−0.05	0.05	−1.15	0.266
		Perspective	−0.04	0.03	−1.61	0.110
		Perspective * T/V	−0.03	0.05	−0.51	0.612
		Mean Immersion	−0.00	0.00	−0.70	0.484
	First spillover	T/V	−0.03	0.04	−0.66	0.513
		Perspective	−0.01	0.03	−0.45	0.652
		Perspective * T/V	−0.07	0.06	1.73	0.240
		Mean Immersion	−0.00	0.00	−0.42	0.676
	Second spillover	T/V	0.00	0.05	0.02	0.986
		Perspective	0.00	0.03	0.03	0.979
		Perspective * T/V	−0.04	0.05	−0.72	0.476
		Mean Immersion	0.00	0.00	0.57	0.570

. These models show that the T/V address form participants received in the final sentence of the short story was not reflected in their reading times. The full models for reading times per narrative perspective and per pronoun of address can be accessed via our repository.

3.2. Immersion Scores

For the measure of immersion, we analyzed the total scores, as well as the scores of each individual subscale (attention, emotional engagement, transportation, and narrative understanding). We start with the findings of the immersion questionnaire in its entirety. Then, the results for each of the subscales are reported. A complete report of the models used for every subscale can be accessed in our repository at https://doi.org/10.34973/gzx3-yk10.

Figure 3 illustrates the total immersion scores per perspective and T/V. Immersion was higher for stories with first-person pronouns (M = 64.3, SD = 28.3) than for stories with third-person pronouns (M = 58.8, SD = 28.8). In addition, for participants addressed with V, first-person stories scored higher on immersion (M = 66.4, SD = 28.9) than third-person stories (M = 57.2, SD = 31.4). For participants addressed with T, there seems to be no great difference between first-person stories (M = 62.3, SD = 27.6) and third-person stories (M = 60.4, SD = 25.7). The violin plots show great variation between participants.

The best model fit for the mixed model for immersion was achieved when including the reading habits question about frequency of reading (Questions 1 and 2), genre preferences (Question 3), preferred reading materials (Question 5), and appreciation. Random intercepts were included for the participant, story, and question.

The analysis showed a significant main effect of perspective on immersion (β = −3.89, SE = 0.90, t = −4.33, p < 0.001): first-person stories scored higher on immersion than third-person stories. No significant main effect was found for T/V (β = 2.00, SE = 2.62, t = 0.76, p = 0.446). The interaction effect between perspective and T/V was significant (β = −5.76, SE = 1.81, t = −3.18, p = 0.002). Thus, whether the reported immersion scores differ between perspectives is dependent on the pronoun of address. We performed an additional analysis to compare the different conditions. This analysis showed no significant difference between the immersion scores of first- and third-person stories when T was used (β = 1.01, SE = 1.27, t = 0.79, p = 1.000). However, a significant effect was found between first- and third-person stories when V was used (β = 6.77, SE = 1.29, t = 5.24, p < 0.001). That is, immersion scores for first-person stories were higher than for third-person stories when participants had been addressed with V.

Figure 4 illustrates the scores for each of the four subscales of immersion, per condition (perspective and T/V). For all subscales, the distribution of the scores shows a similar general pattern as in the aggregated total immersion scores: first-person stories score higher than third-person stories in the V condition, but this difference is not observed in the T condition. The violin plots show great variation between participants, especially for emotional engagement.

For the subscales emotional engagement, transportation, and narrative understanding, the best model fit was achieved when including the reading habits question about frequency of reading (Questions 1 and 2), genre preferences (Question 3), preferred reading materials (Question 5), and appreciation as covariates. For the attention subscale, the best model was included the ART responses. Random intercepts were included for the participant, story, and question. In what follows, we describe the outcome of the models per subscale.

3.2.1. Attention

No significant main effects of perspective (β = −2.32, SE = 1.96, t = −1.18, p = 0.237) nor T/V (β = 3.75, SE = 3.24, t = 1.16, p = 0.248) were found on attention. The interaction between perspective and T/V was not significant (β = −3.39, SE = 3.96, t = −0.86, p = 0.392).

3.2.2. Emotional Engagement

A significant effect was found for perspective on emotional engagement (β = −6.07, SE = 1.55, t = −3.92, p < 0.001). First-person pronoun stories scored higher on this subscale than third-person pronoun stories. No main effect of the pronoun of addressed was observed (β = 1.26, SE = 2.72, t = 0.46, p = 0.643). The interaction between perspective and T/V was significant (β = −9.65, SE = 3.09, t = −3.13, p = 0.002). Thus, whether the immersion scores differ between perspectives is dependent on the pronoun of address. An additional analysis showed no significant effect on perspective when T was used in the questionnaire (β = 1.24, SE = 2.17, t = 0.57, p = 1.000). However, a significant effect was found for perspective when V was used in the questionnaire (β = 10.89, SE = 2.22, t = 4.91, p < 0.001). The emotional engagement scores for first-person stories were higher than the scores for third-person stories when addressed with V.

3.2.3. Transportation

No significant main effects of perspective (β = −2.35, SE = 1.58, t = −1.49, p = 0.137) nor T/V (β = 4.80, SE = 3.37, t = 1.42, p = 0.155) were found on transportation. The interaction between perspective and T/V was not significant (β = −3.15, SE = 3.20, t = −0.99, p = 0.324).

3.2.4. Narrative Understanding

A significant main effect was found for perspective on narrative understanding (β = −3.90, SE = 1.65, t = −2.37, p = 0.018). First-person pronoun stories scored higher on this subscale than third-person pronoun stories. No significant main effect of T/V was observed for narrative understanding (β = −0.75, SE = 3.84, t = −0.20, p = 0.845) nor an interaction effect of perspective and T/V (β = −5.23, SE = 3.28, t = −1.59, p = 0.112).

3.2.5. Individual Differences

In addition to the effects reported above, several covariates showed significant effects. On all scales, a significant effect was found for appreciation: immersion scores were higher when appreciation was higher (cf. Kuijpers et al. 2014; Hartung et al. 2016; Mak and Willems 2019). Moreover, on all scales except narrative understanding, a significant effect was found for the question about frequency of reading: participants who read more frequently scored higher on these scales than participants who never read (Mak and Willems 2019). However, rather surprisingly, the French ART results did not contribute to the models except for the attention subscale model. For the full immersion scale, as well as the subscales attention and narrative understanding, a significant effect was found for the question about reading materials: scores were higher for readers who prefer physical books over e-readers. These findings corroborate the claim that immersion is prone to individual differences (Mak et al. 2020) and warrant additional research into these factors. The details of the mixed-effect models for each subscale can be found in our repository.

4. Discussion

We investigated the influence of first- and third-person narrative perspectives in French short stories on reading times and reported immersion, and whether being addressed with a formal (vous) or informal (tu) pronoun had any effect on readers’ reports about immersion in the story and on their reading times (of the address pronouns). We asked participants to read two short stories on a laptop, word by word, at their own pace. Each story was narrated in either the first or third person. Then, participants were directly addressed with either V- or T-pronouns in a question at the end of each story and throughout the post-hoc immersion questionnaires. Thus, readers’ immersion was gauged by means of a questionnaire administered after each story, and by measuring their reading times of the pronouns (first or third person) at the word level. Other variables indicating individual characteristics were also recorded, viz. reading habits and preferences.

First, we investigated if perspective (first versus third person) had an impact on reading times due to potential time-sensitive cognitive effects. We expected third-person pronouns to be read more slowly than first-person pronouns (H1), but we did not find evidence for this hypothesis. However, we note that previous research has not found conclusive evidence for the claim that narrative perspective leads to differences in reading times either (Hartung et al. 2017b), and so our self-paced reading results rather add to the argument against the notion that one particular perspective poses more processing difficulties than the other (as claimed by Hartung et al. 2016). In a similar vein, Hartung et al. (2017a) suggest that readers simply engage in different modes of perspective-taking regardless of the narrative perspective as a strategy to understand the text, which is supported by different neural networks being activated during their experimental task.

Our results did confirm our second hypothesis (H2): Readers reported being significantly more immersed after reading first-person than third-person stories. In our study, we observed the effect of perspective on the dimensions emotional engagement and narrative understanding. These results differ slightly from what Hartung et al. (2016) observed in their experiment: They found effects of perspective in the dimensions transportation and mental imagery. However, the fact that an identical general immersion pattern was found suggests that further discrepancies in how participants experience immersion may be due to content-based differences. Different types of reading materials might elicit immersion in different dimensions, as the contents and forms of the text evoke diverse immersive states (Kuijpers et al. 2014; Hartung et al. 2017b). So, Hartung et al.’s stories might have contained more visually stimulating language to enhance transportation into the story world, as well as to create mental images of depicted events and physical spaces. Our participants, in turn, experienced heightened emotional engagement, which might have been caused by more relatable emotional states and situations presented in the stories. This emotional resonance can contribute to a more profound understanding of the characters’ motivations, struggles, and the overall storyline, enhancing the reader’s overall comprehension. Either way, perspective (first versus third person) affects the rate of reported immersion, such that first-person pronouns stimulate immersion.

Further, the finding that appreciation and immersion are positively correlated indicates that, by simply enjoying a story less, readers are less prone to report immersion (cf. Kuijpers et al. 2014; Hartung et al. 2016; Mak and Willems 2019). This finding relates to the variability of the immersion results due to individual differences (see Mak and Willems 2019 for more discussion). Indeed, we find significant influences on reading habits, such as reading frequency, but also a preference for physical books over e-books. Surprisingly, our French ART measures did not contribute to explaining any of the differences observed on immersion. Further research into triggers of each individual dimension of the immersion spectrum is warranted to unravel this issue.

We also investigated the impact of being addressed with formal or informal second-person pronouns on the rate of reported immersion. We predicted that the two types of address pronouns would distinctly influence the reports of immersion (H3) based on the strict address norms in France (Warren 2006; Pager-McClymont et al. 2023). We did not find a main effect caused by the pronoun of address. However, we did find that readers who had been addressed with vous after reading first-person stories reported being more immersed than readers who had received vous after reading third-person stories. The informal pronoun of address did not produce any discernible difference between perspectives in the reported rates of immersion. Address pronouns thus affect how readers reflect on and evaluate their reading experience. This difference may be attributed to cultural expectations and social norms. The narrator and the experimenters are strangers to the participants; the task being conducted is perceived as formal. The expectation and the appropriate address form is V. Adhering to these social expectations can influence the reader’s affinity for the task. Therefore, we find a difference when the social norms are followed, and V is used. However, deviating from social expectations by using T-pronouns can cause a surprisal effect that dampens the effect caused by the story’s perspective. This effect may differ across languages and cultures, contingent on the social norms associated with the use of pronouns of address.

Finally, in an exploratory effort, we tested the reading times of T- and V-pronouns of address in a question directed at the reader after the short story, expecting that T-pronouns would be read more slowly than V-pronouns (H4). We did not observe an effect of receiving T or V on reading times: no surprisal effect or processing difficulty was registered. We believe that this may be due to the fact that we only tested a single question directly addressing the participants, and only after the participants were already (supposedly) immersed in the first- or third-person story. The question itself, and the sudden addressing, may have been a surprise, overshadowing the putative effect of the T-/V-pronouns. However, we stress that this distinction was evident in the subsequent questionnaire (as per H3), and so we conclude that address pronouns may trigger surprisal effects in French. Future research may experiment with longer fragments with pronouns of address (T/V) to shed more light on this matter.3

Our study only captured differential effects of perspective on immersion indirectly through readers’ reports. Our instruments were adapted from scales validated in multiple studies (SWAS, Kuijpers et al. 2014; NES, Busselle and Bilandzic 2009). Our results suggest that these scales might be sensitive to the type of language used in them, in particular, to pronouns of address. Our online measure of immersion, reading times at the word level, is hypothesized to be capable of reflecting differences between cognitive processing of each perspective, but no such effect surfaced in our analyses. Future research could implement more sensitive online measures to detect processing differences driven by perspective, such as EDA as in Hartung et al. (2016), eye-tracking, or EEG. These data would provide a more accurate index of real-time immersion.

While our study captures the effects of second-person pronouns in a post-test questionnaire, we did not include a condition with a second-person pronoun in the self-paced reading component of the experiment. Thus, we did not investigate direct effects of second-person pronouns while immersion is happening, as in narratives written in the second person. The second person is not a common narrative perspective, but it is more complex due to the multiplicity of potential referents (see e.g., Sorlin 2022) and the fact that the second-person pronoun is allowed a generic reading quite often (de Hoop and Tarenskeen 2015). Yet research suggests that second-person pronouns may facilitate emotion processing and sense of participation even more so than first- and third-person pronouns, which could translate into higher immersion in the story. Specifically, Brunyé et al. (2011, 2016) found that readers easily develop a sense of agency while reading short narratives written in second person and experience higher emotional engagement than the other perspectives through self-ascriptive devices. Child et al. (2018) similarly find that readers experience text passages describing positive emotions in a more immersed way when addressed in second person (versus third person) in English, and that the second-person pronoun elicits a higher sensitivity to mismatching emotional information.

Since pronouns of address have been repeatedly shown to affect attitudes towards other people and tasks, it may be worthwhile to explore the use of formal-versus-informal second-person pronouns in narratives and their immersive outcomes. Initial endeavors to study such effects are pursued by den Hartog et al. (2024). They investigated the use of T-/V-pronouns in Dutch short stories and found that V has an initial processing cost compared to T (i.e., V is read more slowly at first), which they attribute to their participants (university students) not being used to V-address. They also report that readers experience emotion in narratives differently based on whether the narratives contained V or T pronouns, and this effect is dependent on gender: while male participants respond more negatively to negative texts and more positively to positive texts when V is used, female participants show this pattern when T is used. We leave the question how French T/V pronouns drive immersion in stories as compared to other perspectives to future research.

5. Conclusions

We conducted an experiment with a self-paced reading component and a questionnaire component to study the effect of narrative perspective (first versus third person) and pronoun of address (T versus V) on processing and immersion in French. In the experiment, we collected reading time data (for first- versus third-person pronouns and T versus V pronouns) and self-reported post-story immersion rates. Based on suggestions from the literature, we hypothesized that first-person pronouns and formal pronouns would be read faster than third-person pronouns and informal pronouns, respectively. The rationale for these hypotheses was that (i) third-person pronouns activate the viewpoints of multiple referents and are, therefore, harder to process, and (ii) T-pronouns are non-default and perhaps inappropriate in French and, therefore, trigger a surprisal effect. We did not find evidence for these claims. That we did not find the expected effects may be due to readers employing a reading strategy regardless of individual pronouns, so as to understand the narrative, and for the pronouns of address because of an overall surprisal effect of address after a story.

Furthermore, the post-story questionnaire presented statements using T- or V-pronouns. We expected stories in the first person to yield higher immersion scores than stories in the third person since this effect has been reported before. This prediction was borne out. We also expected T-pronouns to lower the reported immersion scores because of their inappropriateness for the particular situation (at least in French). Although we did not find the hypothesized effect in our data, we did find that T-use nullified the effect of story perspective (or that V-use stimulated it). The interaction between perspective and T/V may thus be explained as follows: we observe that first-person stories trigger a higher reported rate of immersion than third-person stories (due to their higher immersive potential), but this effect in the post-story reports is canceled when an inappropriate pronoun is used to address the participant. Participants are taken aback by the use of the inappropriate T-pronoun and are pulled out of their immersed state, so to speak. This is a novel finding as, to the best of our knowledge, any effects of pronouns of address on the addressees’ attitudes and evaluations were observed solely when the material itself contained pronouns of address (i.e., advertisements).

Pronouns of address may thus have a certain power to affect language users in a psychologically relevant way that is dependent on both individual and cultural factors. We conclude that the processing and the impact of formal and informal pronouns of address deserve further attention.