Digital technology enables people to work away from the office, at least part of the time, and this shift has intensified due to the Covid-19 pandemic (Bick et al., 2020; Kramer and Kramer, 2020; Nguyen et al., 2020; Ohme et al., 2020), expanding to a new extreme the reliance on communication technology and the need to communicate effectively via computer-mediated communication (CMC). Specifically, more and more communication is being conveyed via text-based media, such as emails (Radicati and Levenstein, 2015; Goodman-Deane et al., 2016), known to be lean in social cues compared to face-to-face (FTF), video, or voice interactions (Daft and Lengel, 1986). While emails enable quick communication and increase productivity (Derks and Bakker, 2010), people often experience work-related email overload (Dabbish and Kraut, 2006), which is no surprise as an enormous amount of work emails is being exchanged daily (app. 130 billion according to Radicati and Levenstein, 2015). Clearly, not all emails are equally important or urgent to the email writer. Cues, such as repeated punctuation (“!!” and “??”), are often used as non-verbal signaling of importance (Oberlander and Gill, 2006). Adding various cues to emails is prevalent (Kalman and Gergle, 2014); however, their impact on message interpretation has received scant attention. When on the receiving end, do we interpret the use of repeated punctuation as a sign of importance, and do we respond accordingly? The present study’s aim is to systematically investigate how the use of repeated punctuation shapes the interpretation and consequent behavioral intentions of the message receiver in work-related email communication.

Email usage as well as other text-based communication forms is expected to continuously grow in coming years, as it warrants sufficient control over the information being shared and the way the message is crafted, allowing the writer to actively influence self-representation, even more so than in FTF interaction or videoconferencing (Wells and Dennis, 2016; Marlow et al., 2017). However, as emails are asynchronous and relatively lean in social information, they leave much room for the reader’s interpretation. Due to this interpretation challenge, readers tend to actively search for additional information embedded within an email text that can help with the interpretation process, including paralinguistic cues (Byron, 2008; Johri, 2012; Tong and Walther, 2015). Moreover, the growing reliance on text-based digital services (Przegalinska et al., 2019; Yang, 2021) has led to the use of programmed chat bots. It is important that these digital tools “understand” the true intention of customers and, as importantly, communicate back in such a manner that would lead to the clearest and most effective message. Thus, understanding the impact these cues have on text-based interaction and communication is paramount.

Paralinguistic cues in written texts, such as repeated punctuation, are defined as written communicational codes (e.g., words and symbols) which supplement written language and signify a socially shared gist (Lea and Spears, 1992; Luangrath et al., 2017). Social information processing theory suggests that paralinguistic cues are utilized in place of non-verbal cues to overcome CMC limitations of lack of social presence (Walther, 2007, 2016). Overall, paralinguistic cues have been argued to enrich the expressiveness of the text, disambiguate it, convey affect and a sense of immediacy, and emphasize intended statements (Aldunate and González-Ibáñez, 2017; Prada et al., 2018; Liu and Sun, 2020). However, some empirical research that examined interpretation of paralinguistic cues in textual communication has challenged this notion. For example, Riordan and Trichtinger (2017) found that paralinguistic cues did not facilitate accurate interpretations of email messages, despite feelings of (over)confidence in perceiving the intended message. Thus, whether paralinguistic cues are accurately interpreted and convey the communicator’s intentions remains an open question (Scott and Fullwood, 2020).

Repeated punctuation marks serve as expressive cues, to communicate affect and clarify context and meaning (Gill and Oberlander, 2002; Oberlander and Gill, 2006). If perceived as such, emails containing repeated punctuation should convey and be interpreted as an indication of high importance or urgency. Yet, we know that people do assess not only a message but also the messenger. There is a plethora of research showing how people interpret and attribute situational as well as dispositional inferences when judging actions of others (e.g., Cramton et al., 2007; Kammrath et al., 2007; van Kleef et al., 2012; Hareli, 2014; Elsbach and Bechky, 2017; Riordan and Glikson, 2020). Two fundamental dimensions on which people predominantly judge others are warmth and competence (Judd et al., 2005; Fiske et al., 2007). Warmth reflects perceived social intentions (e.g., friendliness), while competence stands for perceived capacity to achieve goals (e.g., skill). Notably, these judgments were found to beget important behavioral intentions in social settings, such as deciding whether or not to hire a candidate for a job (Cuddy et al., 2011). Thus, it is not merely interpretations; these interpretations have consequences on future actions.

While email norms are still evolving and change between groups and teams (Cheshin et al., 2013; Glikson and Erez, 2013), research suggests that formal and conservative use of accurate punctuation, spelling, and grammar is the expected etiquette (Pankoke-Babatz and Jeffrey, 2002; Lewin-Jones and Mason, 2014). Politeness (e.g., correct grammar) increases positive views of the message and of the sender (Jessmer and Anderson, 2001), while etiquette violations negatively shape readers’ perceptions of email writers (Vignovic and Thompson, 2010). Thus, repeated punctuation might violate expression norms or might be deemed as inappropriate, thereby leading to negatively biased dispositional attributions and responses. For example, Vareberg and Westerman (2020) found that for first impression of instructors, using paralinguistic cues affected their perceptions by students such that they were perceived to be more caring, yet less competent. Namely, paralinguistic cues are not always helpful and in some cases might even harm readers’ perceptions of the writer, creating a gap between the writer’s intentions and the reader’s interpretation that shapes the reader’s response.

In the present paper, we aimed to test whether repeated punctuation increases perceived message importance and thus facilitates positive behavioral intentions, or whether they have a negative impact on dispositional attributions thus hindering positive behavioral intentions. Therefore, this work puts into test two competing hypotheses:

The underlying assumption of the present study, based on CMC literature described above, is that people use repeated punctuation in CMC to convey situational cues. Purposely, they wish to convey through these cues the importance or immediacy of their message and express their emotions and are not intentionally attempting to convey information about their character. To establish this assumption, we conducted a pilot study. Sixty-three undergraduate students from a large American university (52% female, M_age = 25.9) filled the questionnaire on a voluntarily basis. The questionnaire described the use of multiple question marks in CMC and consisted of a variety of filler questions regarding paralinguistic habitual use. The relevant question for the present work regarded the role of repeated question marks and, particularly, what they convey. Items included multiple affective reasoning (“negative affective state”) and contextual reasoning (e.g., “emphasizing the urgency of the issue”), as well as person-related reasoning (e.g., “inability of the person to express themselves with words”) rated on a Likert scale from 1 (“Strongly disagree”) to 7 (“Strongly agree”). As expected, the affective and contextual reasons received significantly higher ratings (M = 4.32, SD = 1.22, Cohen’s d = 0.56, M = 4.63, SD = 1.57, Cohen’s d = 0.68, respectively) than person-related reasons (M = 3.58, SD = 1.54), p’s ≤ 0.006. Moreover, only affective and contextual reasons differed significantly from the scale’s medium value (p’s < 0.001). These results offer empirical support to the theoretical notion that paralinguistic cues in CMC are perceived as a means to convey information about the context the writer is experiencing (i.e., situational), rather than information about the writer’s traits or personality (i.e., dispositional).

We examined the effect of using repeated question marks on situational and dispositional interpretation of an email message and consequent behavioral intentions. A scenario relatable to our sample population was selected (a question regarding employment opportunity of a student). Notably, past research has shown that the use of paralinguistic cues in CMC is gender biased and is associated more with women (Wolf, 2000; Glikson et al., 2018). For instance, Marlow et al. (2017) found that using a salutation with an exclamation mark harmed the perception of competence for female, but not for male writers. Butterworth et al. (2019) found that perceptions of appropriateness and likability of email writers were also malleable to gender and type of emoji used. Thus, the use of repeated punctuation by women might strengthen negative female stereotypes, which are associated with over-expressiveness and are generally related to negative trait assessments (Greenwood and Isbell, 2002; Barrett and Bliss-Moreau, 2009). We therefore examined gender as a potential moderator:

In Experiment 2, we sought to look at the external validity of our findings by examining other types of repeated punctuations (e.g., “!!” and “?!”) in more ecologically valid stimuli. Thus, we collected “real-life” work-related email correspondences containing paralinguistic cues and presented them as the stimuli in our experiment. To further understand the process by which people interpret paralinguistic cues, we aimed to examine whether the clarity of the message content would affect the relationship between using repeated punctuation and interpretation. As mentioned, paralinguistic cues are often used to clarify the message and its context (Houghton et al., 2018). Therefore, the use of paralinguistic cues with ambiguous messages should be more justifiable than such use when the content of the message is clear. This means that for ambiguous messages the use of repeated punctuation could be less harmful with regard to dispositional attributions and intentional behaviors than such use for messages with a clear content. This novel factor was examined in Experiment 2 by asking participants to rate the clarity of the message. As gender did not moderate our findings in Experiment 1, we did not address it in Experiment 2.

Previous research regarding the role of paralinguistic cues in CMC considered them as facilitators in text interpretation, especially when forming first impressions (McAndrew and De Jonge, 2011). Correspondingly, research has found that paralinguistic cues are utilized under the assumption that they will facilitate the understanding of writer’s emotional state or the current situation (Spears and Lea, 1994; Houghton et al., 2018). The pilot survey we conducted offers further support for these findings, indicating that people believe that paralinguistic cues, particularly repeated question marks, are intended to convey situational information rather than reflecting the writer’s traits. However, scarce research has examined the “other side,” namely, what attributions people make when encountering paralinguistic cues.

The present paper addresses this gap by examining a model of situational and dispositional attributions for repeated punctuation in work-related email messages. We used repeated punctuation marks, which are associated with expression of emotional states (Mwangi et al., 2014), but not with people’s personality. Our findings show that people make more trait (i.e., stable) attributions related to paralinguistic cues: In two experiments, the paralinguistic cues were interpreted as representative of the trait of the writer in a negative way, which was related to the reporting of worse outcomes to the (ab)user of those cues.

A potential explanation for this effect stems from applying the Fundamental Attribution Error (Ross, 1977; Berry and Frederickson, 2015) principle, exemplifying the relevance of social biases to CMC. Indeed, research has shown that text-based communication is prone to stereotypes and heuristic decision making (Walther, 1996; Johri, 2012). For example, Vignovic and Thompson (2010) demonstrated that readers form negative perceptions of the email writer when emails contained technical language violations. To the best of our knowledge, the present study is the first to demonstrate the powerful effect of social biases in interpretation of repeated punctuation for behavioral intentions. Future research should explore the relations between other paralinguistic cues and social biases in text-based communication.

Despite expectations (Wolf, 2000; Glikson et al., 2018), we found no evidence for gender moderation. This finding joins others who maintain that gender does not play a role when discussing paralinguistic cues and their effects (Fullwood et al., 2013). This gender-blind phenomenon can be attributed to the strength of communication norms, violation of which drives negative attributions for both genders. In contrast, we did find a moderating role of message clarity. When messages were perceived as ambiguous, repeated punctuation cues did not have an adverse effect. Only when the content of message was clear did these added cues have a negative impact. This highlights the role of message clarity as a novel potential moderator for interpretation of paralinguistic cues in CMC, suggesting that when people do not understand the written message, they try to use additional cues as information and do not penalize the punctuation (ab)user.

Our findings allow inferring about expected norms in text-based CMC, adding insight to this growing literary domain (Marlow et al., 2017). In contrast to Houghton et al. (2018) who demonstrated a negative effect of over-formality in text-messaging among acquaintances, we found a negative effect of informality in email correspondence among strangers. The use of these additional textual cues, which attempt to signal situational or emotional elements to the lean text message, might be assessed as an inappropriate emotional display. Thus, the negative consequences we find in our experiments might be related to interpretation of these cues as an inappropriate emotional display (Cheshin, 2020). Specifically, the consequential negative attributions for (ab)using repeated punctuation across our experiments imply that formality is expected in various types of email correspondence, when interacting with strangers, as previously found in academic situations between students and professors (Stephens et al., 2009). Future research should include these different findings to investigate the exact boundaries of each condition, and the role of communication media and mode, as well as different types of communicators’ relationships, in readers’ interpretation of text-based communications.

There are some limitations to this work. First, the interaction scenarios used in both experiments were one-sided and asynchronous, namely, we did not examine reciprocal or synchronous communications. While the focus of the present study was impression formation of novel situations, it will be interesting to further examine how reciprocal interactions play out when repeated paralinguistic cues are used. In addition, the effect of repeated paralinguistic cues could be influenced by the hierarchical roles of the correspondents. Future research should explore whether the unequal social roles between the writer and the reader moderate the repeated marks impact. Second, these were all lab-based studies; despite the use of real emails in Experiment 2, the evaluations given were requests from researchers, and we can only assume that these also take place in real-life settings. Third, we ran our experiments in two different countries with different cultural norms and values. The first experiment was run in the Netherlands, in English, while the second experiment was run in Israel in Hebrew. Since the findings are consistent, we have confidence that these findings are robust; however, future research should test the effect in additional cultures and languages where the use of paralinguistic cues has different grammatical rules, as well as in cross-cultural communication.

The results of this work offer important practical implications for communicating via text-based electronic means. People use paralinguistic cues in CMC assuming that they will facilitate message interpretation (e.g., Riordan and Kreuz, 2010). Our findings suggest that these cues should be used with caution. Particularly, when abused, they might communicate inferior cognitive skills and result in a decreased willingness to assist the writer. Uncovering the biases in text interpretation may be helpful for text data analyses and creation of automated assistance for people using text-based service but also reply services, such as chat bots. The ability of an automated machine not only to correctly interpret the use of paralinguistic cues but also to implement their meaning in their own communication could be highly valuable for their effectiveness.

In conclusion, the lean text medium, which is becoming more and more prominent in our communication life, includes a paradox. Due to its leanness in non-verbal cues, writers add various paralinguistic cues to bring the text more to life, highlighting elements and adding more intonation and affect to the text. Yet, this attempt is futile, at least when it comes to (ab)using repeated punctuation such as question marks and exclamation points—and it may even lead to adverse consequences. Repeated punctuation did not lead to situational interpretation but rather to stable dispositional ones. When it comes to text-based communication, the gap between the intention of the writer and the interpretation should be minded.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Mark Rotteveel, University of Amsterdam and Nira Munichor, Bar Ilan University. The patients/participants provided their written informed consent to participate in this study.

All authors have contributed equally to this work. YS, EG, and AC equally contributed to the conception and design of the study and wrote sections of the manuscript. EG and AC performed the experiments. EG performed the statistical analyses. AC created the figures. YS wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.