You are on page 1of 15

Available online at www.sciencedirect.

com

ScienceDirect
Journal of Interactive Marketing 53 (2021) 96 – 110
www.elsevier.com/locate/intmar

Emoji, Playfulness, and Brand Engagement on Twitter


Lindsay McShane a,⁎& Ethan Pancer b & Maxwell Poole c & Qi Deng d
a
Sprott School of Business, Carleton University, 1125 Colonel By Drive, 918 Dunton Tower, Ottawa, ON K1S 5B6, Canada
b
Sobey School of Business, Saint Mary's University, 923 Robie Street, SB 147, Halifax, NS B3H 3C3, Canada
c
Sobey School of Business, Saint Mary's University, 923 Robie Street, Sobey 150, Halifax, NS B3H 3C3, Canada
d
Rowe School of Business, Dalhousie University, 6100 University Ave, PO Box 15000, Halifax, NS B3H 4R2, Canada

Abstract

Brands, both human and corporate, are increasingly communicating with their consumers using emojis. The current work examines if and how
these pictographs shape online brand engagement on Twitter (i.e., likes & retweets). To do so, we first examine datasets generated by scraping the
tweets of the most popular celebrity brands and most popular corporate brands (Study 1). This study demonstrates that emoji presence increases
engagement with tweets, with more emoji leading to more likes and retweets. Two controlled experiments then explore the role of perceived
playfulness in explaining this effect of emojis on engagement (Studies 2 and 3). We find that the effect of emojis on brand engagement varies
depending on the nature of the interplay between emojis and text, and the subsequent effect of this interplay on perceived playfulness. Theoretical
contributions and social media practitioner implications are also addressed.
© 2020 Direct Marketing Educational Foundation, Inc. dba Marketing EDGE. All rights reserved.

Keywords: Emoji; Brand engagement; Social media; Playfulness; Twitter; Consumers

Emojis, which are pictographs that represent facial expres- with both an emoji and a brand name has grown by 49% since
sions, people, places, or things (e.g.,❤), are becoming a 2015 (Agnew, 2017). Despite the prevalence of emojis in
mainstream form of communication. Ninety-two percent of brand communications, empirical research in the domain is
consumers online use some form of emoji in their communica- limited, especially with regard to consumer responses. In the
tions (Emogi Research Team, 2015). It is estimated that there are current work, we examine how emoji use in both human and
700 million emojis used every day on Facebook posts alone corporate brand communications influences brand engagement
(Buchholz, 2019). The Oxford Dictionary went so far as to on social media.
declare the “face with tears of joy” emoji as the word of the year Social media is now a dominant channel of communication
in 2015 (Steinmetz, 2015), and the Telegraph noted emojis as between consumers and brands. These platforms have been
“Britain's fastest-growing language” (Emogi Research Team, noted as an effective way to strengthen the quality of the brand–
2015; Telegraph, 2015). consumer relationships, with many associated benefits such as
This trend is not limited to consumer interactions, but rather is increased satisfaction, loyalty, and word-of-mouth intentions
becoming a dominant form of expression in brand communica- (Clark & Melancon, 2013). However, to date, much of our
tions. Well-known brands such as Sony, Chevron, Coke, Burger understanding of the linguistic content of such communications
King, and Taco Bell have all embedded emojis into their remains both isolated from the digital environment, specifically
marketing communications, with many having developed with regard to social media, and focused on text-based
customized brand emoji as well. In fact, the number of tweets communications. This includes works highlighting how
different types of words (e.g., explanatory, refusal, assertive:
⁎ Corresponding author. Kronrod & Danziger, 2013; Moore, 2012; Patrick & Hagtvedt,
E-mail addresses: lindsay.mcshane@carleton.ca (L. McShane), 2012), word sounds (vowel sounds: Lowrey & Shrum, 2007),
ethan.pancer@smu.ca (E. Pancer), Qi.Deng@dal.ca (Q. Deng).

https://doi.org/10.1016/j.intmar.2020.06.002
1094-9968© 2020 Direct Marketing Educational Foundation, Inc. dba Marketing EDGE. All rights reserved.
L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110 97

and typographics (written brand name: Doyle & Bottomley, Kohler, & Keage, 2014). Emojis have been found to have the
2011) influence the consumption experience. In contrast, by potential to reduce misinterpretations, convey emotion, and
looking at how emoji use influences brand engagement on clarify intention (Kaye, Wall, & Malone, 2016; Riordan,
social media, we aim to extend the brand linguistic analysis to 2017). Yet, Willoughby and Liu (2018) find that messages
the online arena and to broaden the emphasis from basic textual with emoji can be perceived as less credible and elicit lower
content to include emojis as an important visual element of the levels of elaboration than those without emoji, and Miller et
message. al. (2016) identify various misinterpretations of emoji. With
We develop hypotheses and test them across one field study regard to brands more specifically, the research is limited.
and two lab experiments. Study 1 examines consumers' There is, however, speculation that emoji use can help brands
engagement (captured through likes and shares) with both foster perceptions of warmth, build a personal connection with
leading celebrity brands and leading corporate brands. We find consumers, and evoke emotions (Agnew, 2017; Kelly &
that consumers like and share tweets more when an emoji is Watts, 2015; Luangrath et al., 2017).
present, and that this effect strengthens as the volume of emojis Evident here is a gap in our understanding of emoji,
increases. From there, in studies 2 and 3, we run controlled lab particularly with regard to its role in digital brand communi-
experiments to test how engagement with tweets varies cations. Consideration for how these cues influence consumers
depending how emojis are used within a message. We focus is largely absent. The literature that does exist focuses on
specifically on how variations in emoji–text interplay influence general sender–receiver communications rather than brand–
engagement via their effect on perceived playfulness. consumer relationships and has yielded mixed results on emoji
We now turn to review the relevant literature on online effectiveness. The current work takes initial steps to understand
brand communications, non-text-based linguistic content, and the role of emoji in brand–consumer digital communications.
playfulness to develop hypotheses regarding how emojis can We argue that emojis can have a meaningful influence on how
influence brand engagement. We then present the results from consumers engage with a brand's social media content via their
our three studies and discuss the theoretical and managerial effect on perceptions of playfulness.
implications of this work.
Brand–Consumer Relationships, Perceptions of Playfulness
Conceptual Development and Engagement
Brands are increasingly relying on social media as a priority
Emoji in Brand Communications channel for building strong relationships with consumers
(Agnew, 2017). Social media provides a rich network for a
As digital communications evolve, both consumers and variety of relationship-oriented activities such as building
brands are discovering new ways of imbuing online interactions brand–consumer communities (Scarpi, 2010), sharing brand–
with meaningful non-textual cues such as emojis or vocaliza- consumer stories (Gensler, Völckner, Liu-Thompkins, &
tions (e.g., *chuckle*). Recent work by Luangrath, Peck, and Wiertz, 2013), and reinforcing relationships (Labrecque,
Barger (2017, p. 1) refers to such cues as textual paralanguage, 2014). This emphasis on relationship-building goals via social
defined as the “written manifestations of nonverbal audible, media offers a valuable lens for understanding the potential
tactile, and visual elements that supplement or replace written role of emojis in social media messaging. Specifically, we
language.” Luangrath et al. (2017) lay out a framework for draw on extant work establishing playfulness as an important
categorizing the array of cues that online consumers use, where factor in fostering stable and healthy relationships and explore
emojis are categorized as a form of visual textual paralanguage the idea of emojis as a useful relationship-building tool to
that can impact brand perceptions, the consumer experience, and enhance engagement through its positive effects on perceived
the brand–consumer relationship. Building on this work, we aim playfulness.
to further our understanding of the influence of visual textual
paralanguage, in the form of emojis, on brand engagement in the Perceptions of Playfulness and Engagement
social media environment. The concept of playfulness has received scant attention in
To date, there is relatively limited research on emoji use in the marketing literature, and yet has been more widely studied
digital brand communications. Recent work has focused on in areas of social psychology and human–computer interaction.
consumer motives for using emojis as a sender of social media It has been put forth as a potentially central tenet of human
messages. This work highlights that consumers believe that social interactions (Hsieh & Tseng, 2017). It revolves around
using emojis helps them to be better understood, to create a the central idea of play, which we refer to as a behavior or
personal connection, to strengthen their message, and to activity pursued with the goal of amusement, enjoyment, and
express emotion (Agnew, 2017; Derks, Fischer, & Bos, 2008; fun (Bundy, 1993; Van Vleet & Feeney, 2015). Playfulness has
Emogi Research Team, 2015; Willoughby & Liu, 2018). On been characterized as “the extent to which the individual is
the receiver side, research has examined neural responses to curious regarding an interaction and finds an interaction
visual textual paralanguage, demonstrating that our brains enjoyable and interesting” (Moon & Kim, 2001, p. 219). This
react to a face-based emoticon in the same way as they do to approach, in particular, highlights the centrality of interaction to
with face-to-face engagement (Churches, Nicholls, Thiessen, the concept of playfulness. Extending from here, perceptions of
98 L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110

playfulness reflect rather the perception of whether an to the experience of play. As such, extending this logic to
interaction encapsulates the key elements of play. brand–consumer relationships and integrating it with the earlier
In the field of social psychology, researchers have explored discussion on playfulness, we predict that:
the role of playfulness in adulthood and found its importance to
H1. The presence of an emoji will increase brand engagement
well-being and relationship health (Glynn & Webster, 1992;
(likes, shares).
Van Vleet & Feeney, 2015). This work consistently shows that
playfulness is important to relationships, creating a sense of H2. The presence of an emoji will increase perceived
security and stability that fosters relationship strength (Van playfulness.
Vleet & Feeney, 2015). While this work examines primarily
H3. The positive effect of emojis on engagement will be
person–person relationships, applying these ideas to digital
mediated by perceived playfulness.
brand–consumer relationships highlights a potentially impor-
tant role of playfulness in branded social media content.
We now turn to work on playfulness in the field of human– Emoji Variations, Perceived Playfulness, and Engagement
computer interaction. Although less focused on the role of In hypothesis 3, we predict that the positive effects of emojis
playfulness in relationships, this work addresses the significant on brand engagement will be mediated by perceptions of
influence of playfulness on consumers' acceptance of technology. playfulness. This prediction raises the question then of whether
The emphasis here has been primarily on understanding the certain emojis convey greater playfulness than others, and if so,
influence of playfulness on individuals' reactions to various whether these variations have the predicted corresponding
technological platforms such as websites (Ahn, Ryu, & Han, effect on engagement. Specifically, it opens the door to consider
2007; Lin, Wu, & Tsai, 2005), digital advertising (Gao, Rau, & that consumers may differentially engage with messages that
Salvendy, 2009), social media (Sledgianowski & Kulviwat, include emojis depending on the playfulness conveyed by the
2009). This work finds positive outcomes associated with emoji(s). The way in which emojis are used in social messaging
perceptions of playfulness such as greater likelihood of using a can vary greatly, from the more obvious examples of which
certain social media platforms (Sledgianowski & Kulviwat, emoji is used (there are currently 3,019 different emojis:
2009), of using a certain retailer website (Ahn et al., 2007), and Buchholz, 2019), to more nuanced variations in the way in
of achieving high levels of customer satisfaction (Hsu, Chang, & which the emojis interact with the text. We explore the latter.
Chen, 2012). Perhaps most relevant, Hsieh and Tseng (2017) find Extant research on image–text integration demonstrates that
that emoticons in person–person text messaging enhance social an individual's reaction to messaging varies depending on the
interactions via perceptions of playfulness. Although not way in which the image interacts with the text (Cohn, 2013;
specifically focused on playfulness, related research also shows Cohn, Roijackers, Schaap, & Engelen, 2018). Cohn et al.'s
that the entertainment value of social media brand posts can emerging body of work on emojis specifically as a form of
positively influence engagement (Luarn, Lin, & Chiu, 2015). multimodal communication examines how an individual's
Taken together, this discussion identifies playfulness as an processing and understanding of emojis vary depending on
important construct in relationship-building that can yield factors such as emoji sequencing, emoji location in the
significant positive outcomes. message, and whether the emoji substitutes or reinforces a
Applying these ideas to digital brand–consumer communi- word that is either a noun or verb (Cohn et al., 2018; Cohn,
cations, we expect that perceptions of playfulness in brand– Engelen, & Schilperoord, 2019). His most recent work finds a
consumer social media messaging will lead to greater levels of lack of grammatical complexity in our use of emoji, and so
engagement. The missing piece, then, is how emojis can highlights that individuals tend to rely on text (vs. emoji) to
enhance these perceptions of playfulness and, in turn, drive convey message meaning in this form of multimodal commu-
engagement. We now turn to explore how emojis may engender nications (Cohn et al., 2019).
greater perceptions of playfulness in brand–consumer social Although this extant work focuses primarily on outcomes
media communications. such as processing times and message comprehension, we
believe these variations in emoji–text combinations influence
Emojis, Playfulness, and Engagement playfulness perceptions and, subsequently, engagement. Spe-
Emojis are commonly used by consumers as a relational cifically, we draw on the idea that emojis play a supporting role
tool. Extant research finds that consumers perceived emojis as a in the message whereas written text provides meaning.
means to add personality to their messages, enhance media Extending these ideas to brand–consumer social media
richness, reciprocate emotion, and to convey both emotion and communications, we anticipate that emojis that interplay with
intention (Agnew, 2017; Desta, 2014; Hsieh & Tseng, 2017; text will be perceived as more playful, and so elicit greater
Thompson & Filik, 2016). They are perceived as a way to engagement than those that do not interplay with text. Here,
enhance digital communications by, in some ways, capturing interplay refers to emojis and text co-existing together to
the non-verbal communication cues that we rely on in face-to- convey certain specific meaning. The logic here is that, in such
face communications (Derks, Bos, & Von Grumbkow, 2007). cases of emoji–text interplay, emojis provide visual intrigue for
One of the ways in which they appear to foster relationship the message and the text serves to satisfy the intrigue by
health is through their positive effects on feelings of enjoyment providing concrete meaning. In a sense, we expect that such
(Huang, Yen, & Zhang, 2008; Riordan, 2017), which is central emoji–text interplay temporarily arouses one's curiosity about
L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110 99

message meaning, a key element of playful interactions (Moon scraped the Twitter feeds of top celebrity brands and corporate
& Kim, 2001). Specifically, we predict: brands and then compared messages with emojis to those
without emojis to understand the relationship between emojis
H4a. Emoji–text interplay will lead to greater brand engage-
on engagement. In studies 2 and 3, however, we then ran
ment than when no such interplay is present.
controlled lab experiments to examine how variations in the use
H4b. The effect of H4a will be mediated by perceived of emojis within tweets influenced engagement via the
playfulness. proposed playfulness mechanism.
The previous hypothesis speaks to situations in which
emojis are shown with no corresponding textual support. Study 1
However, pushing the idea of emoji–text interplay further raises
questions about the nature and degree of emoji–text interplay The aim of this study was to examine the relationship
and how this might influence engagement via playfulness. One between emojis and engagement using the actual Twitter feeds
previously noted variation in emoji–text interplay is whether of leading celebrity and corporate brands. To do so, we
the emoji is placed before or after the text (Cohn et al., 2018; conducted a field study where we created a unique dataset
Cohn et al., 2019). Applying the same logic as above, we based on the scraped tweets of these brands. We included both
expect this variation in emoji–text interplay (i.e., the location of celebrity and corporate brands in a single study to help mitigate
the emoji relative to the text) to significantly influence potential motive variances between celebrities and companies.
perceptions of playfulness, and in turn, brand engagement. Celebrity brands were attractive in that they have high levels of
Consider the earlier argument that a key element of emoji–text both emoji use and interaction. Celebrities are active users of
interplay is that emojis provide visual intrigue and text is used emoji in their social media content, to the point where certain
to provide meaning. We thus expect that emojis located prior to celebrities have customized their own emoji (e.g., Kim
the corresponding text will provide arouse greater curiosity and Kardashian and Kimojis). Top celebrities also have high levels
intrigue because the text has not yet provided the specific of active followers (e.g., Katy Perry has the most Twitter
meaning. In cases where emojis follow text, the emojis may be followers with 93 M). Similarly, corporate brands are increas-
considered somewhat redundant because the text has already ingly engaging with consumers via social media, striving to
established meaning. build relationships with consumers (Agnew, 2017). We drew
We do, however, expect this effect to be moderated by the specifically on some of the most followed corporate brands to
extent to which the emojis are directly related to the text. The supply a high volume of tweets and interactions. Combining
logic here is that it is only in cases where the emojis are directly celebrity and corporate brands also helped to ensure that our
related to the text that emojis can serve as an intriguing prelude sample had a high degree of emoji use and brand–consumer
for the message meaning. As such, we expect the effects of interaction.
placing the emoji before to the text to be attenuated in cases TwitterCounter, an analytics service that provides descrip-
where emojis are less directly related to the textual meaning. tive statistics on Twitter, tracks the most followed accounts on
Specifically, we predict: the platform. We used this ranking to determine the top 15
celebrity brand accounts, ranked by number of followers (see
H5a. Emojis located before the text will lead to greater
Table 1 for account descriptives): Katy Perry, Justin Bieber,
engagement than those that are placed after the text.
Taylor Swift, Barack Obama, Rihanna, Lady Gaga, Ellen
H5b. The positive effects of placing emojis prior to the text will DeGeneres, Justin Timberlake, Kim Kardashian West, Britney
be moderated by the extent to which the emojis are directly Spears, Cristiano Ronaldo, Selena Gomez, Jimmy Fallon,
related to the text, where low emoji relatedness will attenuate Ariana Grande, and Shakira. We should note that celebrity
the effects. accounts tend to be the most followed on Twitter (including
humans & organizations). Only three accounts not run by an
H5c. The effect of H5a and H5b will be mediated by perceived
individual would have made the Top 15 (YouTube, Twitter,
playfulness.
and CNN Breaking News).
We then, again using TwitterCounter to sort, identified the
Overview of the Studies most followed corporate brands and added another 13 accounts
to the field study (refer again to Table 1 for account
We now examine these predictions across three studies, descriptives): YouTube, Twitter, Instagram, ESPN, NBA,
focusing on message content from Twitter. Twitter is one of the Real Madrid FC, NFL, FC Barcelona, Vine, Google, Apple
most dominant social media platforms with 330 million active Music, H&M, and Disney Pixar. These organizations operate in
monthly users and is used by consumers and brands alike. different domains (i.e., social media, sports, entertainment, and
According to Statista (2019), Twitter users' average session clothing), which help mitigate concerns that effects are driven
duration is 3.34 minutes. Brandwatch's 2017 Emoji Report by a particular industry.
states that there are on average 250 million emojis used on Examining accounts with the greatest number of followers
Twitter per month (Agnew, 2017). had several benefits. First, mining multiple accounts allowed us
In Study 1, we focused on how emojis influence brand to create a corpus of sufficient size for analysis, this was critical
engagement with the tweets of leading brands. To do so, we considering that emoji effects could be smaller based on the
100 L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110

Table 1
Emoji presence frequencies by Twitter account.
Account Followers Tweets # Tweets with emoji # Tweets without emoji Total
AppleMusic 8,681,807 26,075 604 (46.5%) 696 (46.5%) 1,300
ArianaGrande 42,315,211 39,481 397 (23.5%) 1,293 (76.5%) 1,690
BarackObama 78,089,364 15,372 11 (0.7%) 1,574 (99.3%) 1,585
britneyspears 48,211,881 4,901 487 (30.7%) 1,101 (69.3%) 1,588
Cristiano 47,523,858 2,690 163 (10.6%) 1,372 (89.4%) 1,535
DisneyPixar 7,566,075 4,868 160 (11.0%) 1,294 (89.0%) 1,454
espn 29,818,100 81,137 34 (2.3%) 1,430 (97.7%) 1,464
FCBarcelona 18,753,102 80,603 692 (39.8%) 1,045 (60.2%) 1,737
google 16,044,410 20,785 8 (7.1%) 105 (92.9%) 113
hm 8,156,745 10,089 60 (3.6%) 1,602 (96.4%) 1,662
instagram 40,347,878 9,601 473 (32.3%) 992 (67.7%) 1,465
jimmyfallon 43,092,544 9,290 18 (1.0%) 1,725 (99.0%) 1,743
jtimberlake 56,821,914 3,488 256 (19.9%) 1,028 (80.1%) 1,284
justinbieber 88,910,140 30,570 26 (2.4%) 1,066 (97.6%) 1,092
katyperry 93,532,264 7,397 695 (44.9%) 852 (55.1%) 1,547
KimKardashian 48,526,572 22,470 117 (7.3%) 1,481 (92.7%) 1,598
ladygaga 64,103,475 7,763 382 (30.8%) 860 (69.2%) 1,242
NBA 23,428,527 143,730 81 (12.4%) 570 (87.6%) 651
NFL 20,351,195 127,054 378 (30.2%) 872 (69.8%) 1,250
realmadrid 20,900,202 53,527 1,433 (81.5%) 325 (18.5%) 1,758
rihanna 66,717,251 9,885 74 (4.6%) 1,550 (95.4%) 1,624
selenagomez 45,972,659 4,201 390 (24.7%) 1,191 (75.3%) 1,581
shakira 40,644,258 4,230 57 (4.6%) 1,181 (95.4%) 1,238
taylorswift13 81,272,395 4,146 64 (6.1%) 978 (93.9%) 1,042
TheEllenShow 62,744,308 12,963 21 (1.2%) 1,736 (98.8%) 1,757
twitter 57,211,607 3,285 217 (19.1%) 918 (80.9%) 1,135
vine 16,902,141 3,521 949 (76.2%) 297 (23.8%) 1,246
YouTube 64,418,145 17,797 214 (12.7%) 1,465 (87.3%) 1,679
Total 1,241,058,028 760,919 8,461 (21.7%) 30,599 (78.3%) 39,060
Source: www.twittercounter.com/pages/100

frequency of use. Second, by looking at celebrity and corporate (2 emoji variables and 30 control variables), would only
brands in heterogeneous domains (i.e., musicians, talk show require 1,842 tweets. Tweets were downloaded using
hosts, pop culture icons, a sports star, a sitting U.S. president, Facepager, a tool developed for fetching publicly available
clothing companies, social media, etc.), we could mitigate data from Facebook, Twitter, and other APIs. All data were
concerns that these effects are driven by a particular type of scraped within a three-hour window on August 15, 2016.
celebrity or industry involvement and balance our sampling. Facepager interfaces directly with the public Twitter API,
Third, popular accounts have more followers, which provides which allowed us to extract the full tweet, like and retweet
sufficient variance in engagement to more easily distinguish count, as well as the date and time of post.
liking and sharing behaviors of individual tweets. It is important
to note that, in this study, we explicitly excluded replies from the
dataset. Replies are not visible to followers unless users follow Measures
both parties, and so including replies artificially depresses
audience opportunities to engage with content. Main Variables
Having established our sample, we then scraped the most
recent 1,800 statuses (less retweets) from each of the 15 Emoji Presence and Emoji Counts. We coded the tweets on
celebrity accounts' timelines (22,186 total tweets) as of August several dimensions. First and foremost was our key variable
15, 2016—this was the maximum allowable limit for interest—the presence or absence of an emoji in a tweet. To
downloading tweets using the public Twitter application code for this variable, we adopted the Unicode Full Emoji Data
program interface (API) at the time. Updates to our search v3.0 as our list for coding emoji. Unicode, which is the
practices after collecting the celebrity data allowed us to computing industry consortium responsible for the encoding
scrape up to 3,200 (less retweets) of the most recent tweets and handling of computer text characters, maintains a repertoire
from each of the 13 corporate accounts (26,955 total tweets) as of universal emoji available for use across platforms and
of September 23, 2016 using Facepager. This created a total devices. As of 2016, there were 1,791 unique emoji in the
corpus of 41,141 tweets. Unicode dataset. We should note here that new emojis can be
A power analysis revealed that to detect small-sized effects added every year, with new emoji character candidates
(Cohen's f = 0.02) on likes and retweets, given 32 predictors reviewed and approved by the Unicode Technical Committee.
L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110 101

Using the Kutools plugin for Excel, we were able to strip out Twitter Content Dummies. We added several controls for the
all characters (e.g., alphanumeric, punctuation) from each tweet presence of additional features of the Twitter message,
that were not emoji as classified by the Unicode emoji dataset. including tags and external links. The mere presence of Twitter
We verified that the remaining characters were only emoji by hashtags, website links, and user mentions have been
converting them to their U + hex roots and running a demonstrated to influence tweet liking and sharing (Pancer &
VLOOKUP function in Excel based on the Full Emoji Dataset Poole, 2016). We coded each tweet for the presence of tags and
from Unicode. Each tweet with an emoji was also visually links, including hashtags, user mentions, website links,
confirmed by two members of the research team. Tweets were embedded media (i.e., image or video), and retweets using
then coded for whether they contained an emoji (emoji customized formulas in Excel. Based on Twitter protocol,
presence), as well as how many emoji characters they contained hashtags were coded and counted for each time the (#) symbol
(emoji count). began an alphanumeric string in the tweet. User mentions were
similarly coded based on each time the (@) symbol started an
Brand Engagement (Likes and Retweets). Brand engagement alphanumeric string in the tweet. Facepager extracted whether
was measured using the like and retweet count of each tweet, an external website link was included in the post. Using this
both of which were downloaded with Facepager. These information, we were able to compute if the tweet featured
variables are frequently used as important indicators of brand embedded media.
engagement on social media, and generally positive responses
to brand messaging (Murdough, 2009; Vargo, 2016). Account Dummies. Because the tweets originated from 1 of
28 different accounts, we added dummy controls for the
Control Variables. There are many factors that could account the tweet was sent from. This helps ensure that our
influence the social media audience to like and share tweets, results are not driven by the tastes of particular celebrities
so we controlled for a series of other variables in our whose messages are more inclined to be liked and shared as
hierarchical regression analysis. well as the unique audience profiles for each celebrity brands
(i.e., Justin Bieber vs. Barack Obama). These dummies also
inherently capture the effects of author gender and their domain
Text Positivity. Prior work suggests that positive content is
of cultural influence.
shared more based on its ability to inspire arousing emotions
(Berger & Milkman, 2012). It was thus important to control for
the valence of the message. We analyzed the tweet using Word and Character Counts. We also control for the tweet's
automated sentiment analysis from the Readability Score. The length in both words and characters. Even though Twitter
analysis uses language processing, text analytics, and compu- already constrains posts to 140 characters or less, prior research
tational linguistics to detect the valence associated with specific has found that longer articles are more likely to be shared as
words, entities, and concepts. It then aggregates these valences they contain more information and enough detail to elicit
to form a global polarity score for the entire text. Scores ranged arousal (Berger & Milkman, 2012). Word counts and character
from either negative, moderately negative, slightly negative, counts per tweet were calculated using Excel formulae.
neutral, slightly positive, moderately positive, and positive. All
tweets were able to be categorized into a single polarity. Timing Controls. To control for the possibility that tweets
made on certain days and times of the year could influence the
Text Readability. We also control for writing complexity. The attention they receive, we added five timing controls based on
ease of reading text can also impact how people feel about it, when the post went online. Prior research has identified the
especially when people are processing information quickly as optimal timing for tweet attention to be on the weekend
in a social media context (Oppenheimer, 2006). The preference (Cooper, 2014). We added controls for the day of the week, day
for reading simpler texts, commonly referred to as linguistic of the year, the month, weekend or weekday, and the year of the
fluency, has been demonstrated to lead to more favorable tweet.
evaluations of messages (Alter & Oppenheimer, 2009). We also
analyzed the text of the tweet for readability using both the Results and Discussion
SMOG index (McLaughlin, 1969) and the Dale-Chall read-
ability formula (Chall & Dale, 1995; Dale & Chall, 1948). The Preliminary Analysis
SMOG index estimates the years of education needed to A preliminary look at our dataset revealed that a very small
understand a piece of writing based on the number of number of tweets generated a disproportionately high number
polysyllabic words present in the text. The Dale-Chall formula of likes and retweets, which positively skewed our data. To
also assesses comprehension difficulty but is based on word account for the positive skewness, our first step was to calculate
familiarity instead of word length. SMOG and Dale-Chall are the natural logarithm of both likes and retweets for use as
both commonly used in the linguistics literature to measure dependent variables in our analysis. To explain the dependent
one's ability to understand contents. We include both in our variables (the natural logarithm of the number of likes or
analysis to control for variance associated with word familiarity retweets), we use an Ordinary Least Square method to estimate
and word length as proxies for reading difficulty. the following equation:
102 L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110

ln(dependent variable) = b1 + [b2 × Emoji Presence] + emoji on likes and retweets (H1). We focused first on the
[b3 × Text Characteristics] + [b4 × Tweet Characteristics] + influence of emoji presence on likes. The results indicate that
[b5 × Celebrity Account Dummies] + [b6 × Timing] + cX. the mere presence of having an emoji in a tweet significantly
As we are estimating two related dependent variables, we increased likes (Table 2, Model 1). This was consistent with
could use a seemingly unrelated regression setting. However, H1. Our effect persists even when controlling for a battery of
since we are explaining both dependent variables using the other content characteristics, including both text valence and
same independent variables, using this would not improve complexity (see Table 2, Model 2 for a full list). This is
efficiency beyond OLS. important to our analysis in that controlling for text valence,
Next, we examined the frequency with which emoji was despite finding that positive tweets tend to be liked more,
used for each celebrity account. Based on our sample, 21.8% of enables us to conclude that our findings are not simply driven
all tweets contained emoji. There were significant differences by positive wording. In the case of text complexity, where we
between the various accounts with regard to their emoji use expectedly find that more complex tweets are liked less, we can
(Table 1). An ANOVA revealed a main effect of account on similarly see that our effects cannot be solely explained by
Emoji Presence (F(1, 26,831) = 242.88, p < .05). Several message readability metrics.
accounts used emoji in over 25% of their messages (Britney The emoji result also persists when accounting for other
Spears, Katy Perry, Lady Gaga, AppleMusic, FCBarcelona, features inherent in Twitter posts. This demonstrates that liking
NFL, realmadrid) while others used them less than 2% of the is not simply driven by hashtags, user mentions, images, or
time (Barack Obama, Jimmy Fallon, Justin Bieber, Ellen videos—all features that are designed to increase the attention
DeGeneres, espn). We kept all 28 accounts for inclusion in our paid to and the exposure received by a message. Emoji
regression analysis. influences liking in a way that is not captured by these tagging
tools. The results are also robust to controlling for the brand
Main Analyses itself, indicating that our findings are not an artifact of
messages from a specific brand being liked more than another.
The Effect of Emoji Presence and Emoji Count on Likes. We Similar to Berger and Milkman's (2012) viral content analysis,
then conducted our main analyses—hierarchical regression the inclusion of topic dummy variables allowed us to
analyses (Aiken, West, & Reno, 1991) to examine the effect of conservatively test our hypothesis – the observed relationship

Table 2
The “liking” and “retweeting” of tweets as a function of emoji presence and other content characteristics.
Emoji Presence - Including Controls - Including Emoji Count - Emoji Presence - Including Controls - Including Emoji Count -
ln(Likes) (1) ln(Likes) (2) ln(Likes) (3) ln(Retweets) (4) ln(Retweets) (5) ln(Retweets) (6)
Emoji present 0.067 ⁎⁎⁎ 0.111 ⁎⁎⁎ 0.049 ⁎⁎⁎ 0.023 0.115 ⁎⁎⁎ 0.040 ⁎⁎
(0.000) (0.000) (0.001) (0.277) (0.000) (0.025)
Emoji count 0.037 ⁎⁎⁎ 0.044 ⁎⁎⁎
(0.000) (0.000)

Text positivity 0.015 ⁎⁎⁎ 0.015 ⁎⁎⁎ 0.007 ⁎ 0.008 ⁎


SMOG index − 0.032 ⁎⁎⁎ − 0.031 ⁎⁎⁎ − 0.034 ⁎⁎⁎ − 0.033 ⁎⁎⁎
Dale-Chall score − 0.010 ⁎⁎⁎ − 0.010 ⁎⁎⁎ − 0.008 ⁎⁎⁎ − 0.008 ⁎⁎⁎
Links − 0.244 ⁎⁎⁎ − 0.236 ⁎⁎⁎ − 0.291 ⁎⁎⁎ − 0.282 ⁎⁎⁎
Hashtags 0.035 ⁎⁎⁎ 0.036 ⁎⁎⁎ 0.095 ⁎⁎⁎ 0.097 ⁎⁎⁎
User mentions − 0.015 ⁎ − 0.014 ⁎ − 0.013 − 0.013
Character count − 0.001 ⁎⁎⁎ − 0.002 ⁎⁎⁎ − 0.001 ⁎⁎⁎ − 0.002 ⁎⁎⁎
Word count 0.016 ⁎⁎⁎ 0.019 ⁎⁎⁎ 0.014 ⁎⁎⁎ 0.017 ⁎⁎⁎
Day of year 0.001 ⁎⁎⁎ 0.001 ⁎⁎⁎ 0.001 0.001
Day of week − 0.017 ⁎⁎⁎ − 0.017 ⁎⁎⁎ − 0.015 ⁎⁎⁎ − 0.015 ⁎⁎⁎
Weekend of year 0.082 ⁎⁎⁎ 0.082 ⁎⁎⁎ 0.075 ⁎⁎⁎ 0.075 ⁎⁎⁎
Month of year − 0.037 ⁎⁎⁎ − 0.037 ⁎⁎⁎ − 0.044 ⁎⁎⁎ − 0.043 ⁎⁎⁎
Year 0.495 ⁎⁎⁎ 0.496 ⁎⁎⁎ 0.107 ⁎⁎⁎ 0.108 ⁎⁎⁎
Account X X X X
dummies (28)

N 39,059 38,970 38,970 39,059 38,970 38,970


r2 0.000 0.777 0.778 0.000 0.741 0.741
p values in parentheses.
⁎ p < .10.
⁎⁎ p < .05.
⁎⁎⁎ p < .01.
L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110 103

between emoji and likes holds not only across topics but within season. In actual fact, this was a fictional tweet sent by a
them. Lastly, our results hold when accounting other robustness fictional coffeehouse for whom we created a Twitter account.
checks, including word and character counts and release timing In the Emoji–Text Interplay condition, participants read the
variables. following “Friends + coffee + chocolate + Christmas = tears of
Table 2, Model 3 is the same as Model 2, but it adds the joy. Come join us at the #RoastedBean for all of your favorite
emoji count variable to the regression analysis. We find that the things. #CoffeeShopHappiness.” These participants also saw
more emoji in a tweet, the more popular it was. The inclusion of five emojis (friends, coffee, chocolate, Christmas, and tears of
the emoji count variable also decreased the effect of the emoji joy), each one corresponding to one of the first five words in
presence variable. the written text, above the written text (see Fig. 1). The basic
premise here is that the emojis serve as intriguing visual prelude
The Effect of Emoji Presence and Emoji Count on Retweets. of which the specific meaning is then clarified by the text. The
We performed the same hierarchical regression analyses on No Emoji–Text Interplay condition was identical except that it
retweets as another form of engagement (Table 2; Models 4–6). replaced the first five words (friends, coffee, chocolate,
Retweets provide more exposure than likes because they share Christmas, and tears of joy) with the corresponding emojis. In
another user's story with their own followers, leading to a this case, there is no textual meaning supporting the emojis.
higher number of observations for retweeted content. Consis- The control condition included the same text, but no emojis
tent with the models on liking, the presence of emoji increased were shown to the participants in the control condition (see
the sharing of a tweet (consistent with H1) (Table 2, Model 5). Fig. 1). Here, the Emoji–Text Interplay condition is expected to
Notably, this effect was significant only when controls were elicit greater levels of perceived playfulness than when the
included in the analysis. When we added the emoji count emojis simply substitute for the words in the No Emoji–Text
variable to the analysis, the presence effect disappeared Interplay conditions.
(Table 2, Model 6). This is not surprising considering that the Following exposure to the tweet, all participants responded
presence variable is fundamentally rolled into the emoji count, to a series of related questions. These items were designed to
which manages to explain more variance. The more emoji capture perceived playfulness and intentions to engage. We
contained in a tweet, the more it was shared and retweeted. also included several downstream measures including happi-
The results of this first study provide robust support for the ness, perceptions of brand warmth, perceptions of brand
idea that emoji is positively related to brand engagement, and competence, and perceptions of brand playfulness. Although
that more emoji can further enhance engagement. The study, in the primary downstream variable of interest is engagement
using real tweets, offers the advantage of examining the actual across studies, we were also interested in measuring feelings of
relationship between emoji use and brand engagement and happiness and certain brand-level perceptions. These variables
offers clear, immediate, and applicable implications for have been previously associated with emoji use (Agnew, 2017;
managers. However, because it focuses on examining actual Kelly & Watts, 2015; Luangrath et al., 2017), yet not explicitly
tweets, we are both limited to only correlational data and tested and so we have included them to examine if they, like
constrained in our ability to examine some of the nuanced engagement, are influenced by emoji use via perceptions of
effects of emoji on engagement, particularly with regard to playfulness.
perceptions of playfulness. To address this gap, we now turn to
conduct two lab studies to better examine the proposed Measures
underlying mechanism. To do so, we focus on tweets that
include emojis and look at variations of emoji use within those All items were measured using 7-point scales. To capture
messages. playfulness, participants were asked to indicate the extent to
which they found the post playful (not at all/very much so), as
Study 2 well as to indicate the extent to which they agreed or disagreed
with the following statements: “I enjoyed reading this tweet”;
This study aims to take initial steps to examine the proposed “It was fun to read this tweet”; “I found this tweet entertaining”
role of playfulness in driving the effect of emojis on (α = 0.91). To capture intentions to engage, participants were
engagement, focusing specifically on the basic presence of asked how likely they would be to engage with and like the
emoji-text interplay (H2, H3, H4a, H4b). In doing so, it aims to tweet (r = 0.78). We also asked participants the extent to which
take a more nuanced look at how the way in which emojis are the tweet made them feel happy. Finally, we also measured
used within message impact engagement. perceptions of brand warmth, competence, and playfulness.
Perceptions of brand warmth were captured using a 3-item
Method scale asking participants the extent to which they perceived the
Roasted Bean Coffeehouse as likeable, approachable, and
One hundred and four (54% male) undergraduate students warm (α = 0.90). Perceptions of brand competence were
participated in a three-level (Emoji-Text Interplay, No Emoji- captured by asking participants the extent to which they
Text Interplay, Text Only) between-subjects experiment. All perceived the Roasted Bean Coffeehouse as competent, expert,
participants were shown a tweet purported to be sent by The professional, and capable (α = 0.89). Perceptions of brand
Roasted Bean Coffeehouse leading up to the December holiday playfulness were captured by asking participants the extent to
104 L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110

Condition Stimuli

Emoji-Text Interplay

No Emoji-Text Interplay

Text Only

Fig. 1. Stimuli for Study 2.

which they perceived the coffeehouse as playful, clever, smart, p < .05, ƞp2 = 0.07). (See Table 3 for summary representation
fun, and creative (α = 0.94) of means across conditions.)
Next, we compared the Emoji–Text Interplay condition to
the Text-Only condition. Consistent with predictions, we found
Results and Discussion a similar pattern of results. Least significant difference post-hoc
analyses showed that the Emoji–Text Interplay condition led to
Main Analyses greater perceived playfulness, greater intentions to engage,
ANOVAs on each dependent variable (playfulness, engage- greater happiness, and more positive brand perceptions than
ment, happiness, brand warmth, brand competence, brand Text Only messages (Playfulness: Ms = 5.41 vs. 4.01, p < .01,
playfulness) showed significant effects of the experimental ƞp2 = 0.03; Intentions to Engage: Ms = 3.79 vs. 2.43, p < .01,
manipulation (F(2,101) = 8.70, 3.35, 8.56, 5.41, 4.89, 9.23, ƞp2 = 0.01; Happiness: Ms = 5.59 vs. 4.00, p < .001, ƞp2 =
ps < 0.05, ƞ p2 = 0.13, 0.04, 0.15, 0.10, 0.09, 0.14). 0.01; Brand Playfulness: Ms = 5.46 vs. 4.17, p < .01, ƞp2 =
First, we engaged in a comparison of the Emoji–Text 0.01; Brand Warmth: Ms = 5.58 vs. 4.62, p < .001, ƞp2 =
Interplay condition to the No Emoji–Text Interplay condition. 0.004; Brand Competence: Ms = 4.37 vs. 3.63, p < .01, ƞp2 =
Least significant difference post-hoc analyses showed that the 0.01). Finally, we compared the No Emoji–Text Interplay to the
Emoji–Text Interplay condition led to significantly greater Text-Only conditions, and there were no significant differences
perceptions of playfulness, greater intentions to engage, greater in participant responses.
happiness, and more positive brand perceptions than the No
Emoji–Text Interplay condition (Playfulness: Ms = 5.41 vs.
4.55, p < .01, ƞp2 = 0.07; Intentions to Engage: Ms = 3.79 vs. Mediation Analysis
3.01, p < .05, ƞp2 = 0.02; Happiness: Ms = 5.59 vs. 4.35, We conducted a mediation analysis to test the role of emoji
p < .001, ƞp2 = 0.11; Brand Playfulness: Ms = 5.46 vs. 4.35, on engagement via perceived playfulness. Specifically, we
p < .001, ƞp2 = 0.12; Brand Warmth: Ms = 5.28 vs. 4.25, simultaneously tested: (i) whether the path from emoji (both
p < .001, ƞp2 = 0.11; Brand Competence: Ms = 5.36 vs. 4.71, Emoji–Text Interplay and No Emoji–Text Interplay) versus no
L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110 105

Table 3
Study 2 means.
Conditions Perceived Intentions to Happiness Perceived brand Perceived brand Perceived brand
playfulness engage playfulness warmth competence
Emoji-text interplay 5.41ab 3.79ab 5.59ab 5.46ab 5.28ab 5.36ab
No emoji-text 4.55a 3.01a 4.35a 4.35a 4.25a 4.71a
Interplay
Text only 4.01b 2.43b 4.00b 4.17b 4.45b 4.57b
Different subscripts indicate significant differences in means across conditions with regard to that particular variable at p < .05. For instance, for the perceived
playfulness variable, the subscript ‘a’ indicates a significant difference between the Emoji-Text Interplay condition and the No Emoji-Text Interplay condition and the
subscript ‘b’ indicates a significant difference between the Emoji-Text Interplay condition and the Text Only condition. The lack of a common subscript across No
Emoji-Text Interplay and Text Only conditions indicates that there is no significant difference.

emoji (Text Only) to brand engagement was mediated by Interplay should still yield more positive responses than the
perceived playfulness, and (ii) whether the effect of Emoji– control condition, simply not as great an effect as one with
Text Interplay (vs. No Emoji–Text Interplay and Text Only) on emoji–text interplay. This result raises an interesting possibil-
engagement is mediated by playfulness. To do so, we used ity that only emojis that elicit sufficient levels of perceived
Process v3, Model 4 with the three conditions captured using a playfulness will generate greater engagement and other
multicategorical predictor variable. Consistent with Hayes positive responses. In the absence of sufficient playfulness,
(2017), we set participants in the Emoji–Text Interplay emojis may not engender positive results beyond a text-only
condition to 1 and others to 0 for our first dummy variable message. At a more general level it indicates that the effects of
(D1). For our second dummy variable (D2) we had participants emoji on engagement are quite nuanced. We endeavor to
in the No Emoji–Text Interplay condition set to 1 and others to explore these nuances further in Study 3.
0, thus making the Text-Only condition the reference group.
We first examined the confidence intervals around the Study 3
indirect path from D1 (Emoji–Text Interplay vs. No Emoji–
Text Interplay and Text-Only) to engagement via playfulness The primary purpose of Study 3 was to take initial steps
(using a bootstrap analysis with 5,000 resamples) (Preacher & to examine how a specific type of emoji–text interplay
Hayes, 2008; Zhao, Lynch, & Chen, 2010). The confidence (i.e., location of the emoji relative to text) influences
interval did not span zero (CI95%: 0.332–1.666), consistent engagement. We focus specifically on: (i) testing the prediction
with mediation. The regression coefficient of the direct effect of that emojis placed before the text are expected to enhance
the independent variable on engagement (i.e., the unmediated engagement via playfulness more than those after (H5a, H5c)
effect of the independent variable) was not significant (b = and (ii) testing the proposed moderating effect of emoji–text
− 0.059, t(98) = − 0.139, p > .80), indicating indirect-only relatedness (H5b). To do so, we designed a fully crossed,
mediation (Zhao et al., 2010). These results suggest that the between-subjects experiment varying both the emojis' location
difference in engagement between the Emoji–Text Interplay relative to the text and the extent to which it directly related to
and the non-interplay conditions is entirely attributable to the text. Third, in this study we also sought to extend the
perceived playfulness. In contrast, the confidence intervals generalizability of the results by using Amazon Mechanical
surrounding the path from D2 (both emoji conditions vs. the no Turk to recruit from the general population rather than a sample
emoji condition) to engagement via playfulness did span zero of university students as in the previous study.
(CI95%: − 0.818 to 0.563).
Method
Discussion
Taken together, the results of this study are consistent with Two-hundred and eighteen (61% male) participants were
the prediction that emoji–text interplay enhances engagement recruited from Amazon Mechanical Turk to participate in this 2
significantly more than emojis that do not interplay with (Location: Before vs. After) × 2 (Emoji–Text Relatedness: High
text (H4a). Consistent with expectations, this difference in vs. Low) between-subjects experiment. All participants were
Emoji–Text Interplay also led to significant positive effects on shown a tweet supposedly from the Roasted Bean Coffeehouse
feelings of happiness, and brand perceptions. The results also (a fictional coffeehouse) promoting their seasonal products, as in
support the prediction that the positive effects of emoji–text the previous study. Across all conditions, participants read the
interplay on engagement are mediated by perceptions of following text “Celebrating the season with our chocolate cherry
playfulness (H4b) cookies available now through February. What can be more
It is worth noting, however, that it was somewhat unexpected welcoming than a warmed-up mouthful of chocolate cherry
to see no significant differences in participant responses across cookie. #ChocolateCherryGoodness.” Across conditions, we
the No Emoji–Text Interplay and Text-Only conditions (hence varied both Location and Emoji–Text Relatedness. Across
only partial support for H2 and H3). Our predictions would Location conditions, participants either saw the emoji prior to
suggest that the presence of an emoji in the No Emoji–Text or after the relevant text. Across Relatedness conditions, we
106 L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110

varied whether the emojis were directly related to the text or (Preacher & Hayes, 2008). Results showed that the confidence
not. In the High Relatedness condition, all tweets included the intervals surrounding the conditional indirect effects of
same three emoji—a chocolate bar, cherries, and a cookie— Location on engagement intentions via playfulness varied
that mapped on exactly to the stated text (e.g., chocolate across levels of Relatedness. In High Relatedness conditions,
cherry cookie). In the Low Relatedness conditions, the tweets the confidence intervals did not span zero (CI95% = − 0.848 to
included three emojis related to sentiment of the tweet but not − 0.058). In contrast, in Low Relatedness conditions, the
directly linked to the text—a snowman, star, and happy face confidence intervals did span zero (CI95% = − 0.254 to
(see Fig. 2). 0.600). Consistent with expectations, these results support the
In this study, we expected that Location (Before vs. After) idea that playfulness mediates the effect of Location on
and Relatedness (High vs. Low) would interact to differentially engagement only when the emoji is highly related to the text
influence message playfulness and, subsequently, engagement. (H5a, H5b, H5c). Notably, there was no significant direct effect
Within Before conditions, we expected low relatedness emojis of Location on engagement intentions once the effect of
to attenuate perceptions of playfulness and engagement Relatedness had been taken into account (b = − 0.226, t
relative. In contrast, we did not expect such effects within the (214) = − 1.00, p > .30), consistent with indirect-only media-
After conditions given that the text establishes meaning prior to tion (Zhao et al., 2010).
the emoji being seen in both cases, and so the relatedness of the
emoji is less relevant. Discussion
After reading the tweet, all participants responded to a series Overall, these results showed the predicted path from
of questions designed to capture the primary variables of Location to engagement via perceived playfulness. Tweets
interest. These included playfulness measures to flesh out the with emojis placed before the relevant text were perceived as
proposed mediating process and engagement intentions mea- more playful, and so elicited greater engagement, than those
sures to capture the predicted downstream response. Having placed after the relevant text but only in cases where the emojis
established the link to brand-level perceptions in the prior were directly related to the text.
study, we do not focus on these downstream measures in this These results also help to address the somewhat surprising
study. To capture playfulness, we used the same measure as in result in Study 2 where we found no significant difference in
the previous study (α = 0.83). To capture engagement inten- engagement, and other downstream effects, between the No
tions more comprehensively, we adjusted the measure to Emoji–Text Interplay and Text-Only condition. Specifically, it
include four-items. Specifically, we asked participants to suggests that simply including an emoji in the message does not
indicate the extent to which they were likely to engage with necessarily result in greater engagement. Rather, emojis may
the tweet, like the tweet, share the tweet, and comment on the lead to differential effects on playfulness and so yield varying
tweet. The resulting measure was more robust than the one used levels of engagement. More generally, it speaks to the need for
in study 2 (α = 0.94). brands to strategically integrate emojis into their messaging.
Finally, the design of this study also serves to address a
Results and Discussion potential limitation in the previous study. In Study 2, we find
that emojis that are directly reinforced by text generate greater
Main Analyses engagement and that this occurs via perceptions of playfulness.
Initial analysis revealed an interaction effect of Location and One alternative explanation, however, is that the message is
Relatedness and on both playfulness and engagement intentions simply stronger when the emoji is reinforced by the text versus
(Fs(1, 214) = 4.48 and 4.83, ps < 0.05, ƞ p2 = 0.021 and when it is not. Our measurement of playfulness and corre-
0.022). No main effects were significant. Follow-up analyses sponding mediation analysis enhances our confidence in the
showed that, within the Before conditions, participants proposed playfulness mechanism, but it does not rule out a
indicated greater intentions to engage with the tweet when the corresponding message strength process. The design of Study
emojis were highly related to the text (F(1, 214) = 6.38, Ms = 3, however, does address this issue as the manipulations lead to
3.58 vs. 2.69 p < .05, ƞ p2 = 0.053), consistent with H5a. A different predictions regarding these two variables (i.e.,
similar pattern followed with playfulness. Specifically, the message strength vs. playfulness). A playfulness lens predicts
tweet was perceived as more playful when the emoji was more the interaction effect that we find—a direct emoji that precedes
directly related to the emoji (F(1, 214) = 4.45, M = 5.21 vs. the text will lead to greatest engagement. A message strength
4.77 p < .05, ƞ p2 = 0.038). There were no significant approach, however, would simply predict a main effect of
differences within After conditions across levels of Relatedness Relatedness regardless of the level of emoji–text interplay as in
(ps > 0.25). See Table 4 for summary representation of means both cases the emoji and text work together to enhance message
across conditions. strength.

Mediation Analysis General Discussion


We then conducted a mediation analysis to test whether
perceived playfulness could explain the differential effects of Brands frequently use emoji in their social media commu-
emoji across conditions. To do so, we used the moderated- nications. Yet, perhaps because of their rapid integration into
mediation design of Model 7 in PROCESS (5,000 bootstrap) computer-mediated communication platforms, we have little
L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110 107

Condition Stimuli

High Relatedness Emoji Before Text

High Relatedness Emoji After Text

Low Relatedness Emoji Before Text

Low Relatedness Emoji After Text

Fig. 2. Stimuli for Study 3.

understanding of how emojis influence brand–consumer highlight nuances in this relationship between emojis and
relationships. The current work takes steps to address an initial engagement, showing that emojis enhance engagement primar-
key piece: how does emoji use influence brand engagement? ily when they precede highly related text in that they convey
Our field study results demonstrate that the mere presence of an greater playfulness. This mixed-method approach, analyzing
emoji in brand tweets increases the extent to which consumers both actual tweets and lab experiment data, enabled us to
like and share content and that more emoji within a message generate unique insights into emoji use in social media
amplify this effect further. Our two subsequent lab studies messaging.
108 L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110

Table 4 role of playfulness more generally in shaping the relationships


Study 3: Differential effects of emojis on engagement intentions and between consumers and brands.
playfulness as a function of location and relatedness.
Emoji location Emoji relatedness
Engagement intentions High relatedness Low relatedness
Practitioner Implications
Before 3.58abc 2.69b
Brands are continuously searching to optimize their brand
After 2.62a 2.89c
communication strategies in the face of ongoing challenges
Perceived playfulness (e.g., new media, changing behaviors, divided consumer
Before 5.21a 4.77 attention: Batra & Keller, 2016). Our research identifies emojis
After 4.66a 4.97
as one important and easy-to-implement tool that managers can
Notes. Different subscripts indicate significant differences in means across strategically use in face of these challenges. Specifically, to the
conditions with regard to that particular variable at p < .05. For instance, for the
extent that emojis can enable brands to convey playfulness,
engagement intentions variable, the subscript ‘a’ indicates a significant
difference between the Before with High Relatedness condition and the After brands stand to gain from greater levels of engagement as well
with High Relatedness condition. The subscript ‘b’ indicates a significant as several other benefits of playful content. We find that one
difference between the Before with High Relatedness condition and the Before specific way in which brands can convey playfulness is to place
with Low Relatedness condition. And, so on for the ‘c’ subscript. The lack of a emojis before the focal text and to ensure that the emojis are
common subscript, for instance in the case of comparing the means of After
highly related to the text that immediately follows.
with High Relatedness and After with Low Relatedness, indicates that the
difference between means is not significant.
Limitations and Future Research

Theoretical Contributions The current work by no means claims to be an exhaustive


examination of the influence of emoji on brand communica-
The current work contributes to the social media messaging tions. Rather, it aims to take important first steps in empirically
literature by identifying emojis as an important factor in digital demonstrating the promise of emoji, as a useful non-textual cue
brand engagement, and by highlighting the nuanced role of in digital brand communications. As such, the research has
emojis in such messaging. It also provides an initial empirical certain limitations, particularly with regard to generalizability.
examination of the phenomenon of textual paralanguage Specifically, all studies focus on Twitter content and both lab
(Luangrath et al., 2017). Luangrath et al. (2017) develop a studies focus on tweets from a fictional coffeehouse.
conceptual framework for examining the mediating role of With regard to social media platforms, it is possible that the
textual paralanguage in shaping the relationship between the differences across the content and user profiles of various
antecedents of brand factors, platform factors and target platforms yield some meaningful differences in brand engage-
audience factors, and key outcome variables (i.e., brand ment. For instance, users might respond to the same emojis in
perceptions, brand–consumer relationships, consumer effects). different ways given research showing variations in individuals'
While their proposed framework will require examination responses to emojis (Chen & Siu, 2017; Marengo, Giannotta, &
across a wide variety of studies, the current research is, to our Settanni, 2017; Tossell et al., 2012; Wall, Kaye, & Malone,
knowledge, the first to empirically test the proposed link 2016).
between textual paralanguage (in this case, visual textual Similarly, it is possible that certain brands are better
paralanguage in the form of emoji) and one of the proposed positioned to use emojis to enhance engagement via playful-
outcomes (engagement). ness than others. B2B and B2C brands might, for instance, have
The current work also contributes to the debate on optimal been shown to receive different consumer reactions when
diffusion strategies. This word-of-mouth literature highlights incorporating emoji in their posts on social media (Deng,
that an ongoing challenge with forecasting social media 2019). Furthermore, it is possible that even across B2C brands,
interactions, and more specifically the virality of online content, consumers reactions to emojis will vary depending on the
is the lack of evidence-based policy regarding message content brand's personality, the nature of the relationship, (e.g.,
(Berger & Milkman, 2012). Our playfulness-based conceptu- exchange vs. communal) or perhaps even depending on the
alization of how emoji might influence engagement speaks to goal of the consumer in consuming the content. With regard to
this issue. It suggests, for instance, that emojis might have the the latter, Barcelos, Dantas, and Sénécal (2018) show that a
capacity to influence the virality of online content and, more more human voice engenders more positive outcomes when the
generally, that playfulness may enhance the virality of content. receiver holds more hedonic goals (vs. utilitarian). Applying
This research also, perhaps most notably, identifies playful- this work to the current research, it is possible that the found
ness as an important construct in brand messaging via social effects of emoji on engagement via playfulness are contextu-
media. This works highlights the potential importance of ally-bound to social media messaging that is consumer for
playfulness in fostering rich online interactions and, more hedonic purposes. Taken together, this discussion identifies
broadly, developing strong brand–consumer relationships. This rich avenues for future research examining the effects of both
is a ripe area for future research that might further extend emoji and playfulness on brand–consumer relationships in
beyond computer-mediated communications to examine the different contexts.
L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110 109

References Gao, Q., Rau, P.-L. P., & Salvendy, G. (2009). Perception of interactivity:
Affects of four key variables in mobile advertising. International Journal of
Agnew, P. (2017). Emoji Report. Brighton, England, UK: Brandwatch. Human-Computer Interaction, 25(6), 479–505.
Ahn, T., Ryu, S., & Han, I. (2007). The impact of web quality and playfulness Gensler, S., Völckner, F., Liu-Thompkins, Y., & Wiertz, C. (2013). Managing
brands in the social media environment. Journal of Interactive Marketing,
on user acceptance of online retailing. Information & Management, 44(3),
263–275. 27(4), 242–256.
Aiken, L. S., West, S. G., & Reno, R. R. (1991). Multiple regression: Testing Glynn, M. A., & Webster, J. (1992). The adult playfulness scale: An initial
assessment. Psychological Reports, 71(1), 83–103.
and interpreting interactions. Thousand Oaks, CA: Sage.
Alter, A. L., & Oppenheimer, D. M. (2009). Uniting the tribes of fluency to Hayes, A. F. (2017). Introduction to Mediation, Moderation, and Conditional
form a metacognitive nation. Personality and Social Psychology Review, 13 Process Analysis: A Regression-Based Approach. New York, NY: Guilford
(3), 219–235. publications.
Hsieh, S. H., & Tseng, T. H. (2017). Playfulness in mobile instant messaging:
Barcelos, R. H., Dantas, D. C., & Sénécal, S. (2018). Watch your tone: How a
brand's tone of voice on social media influences consumer responses. Examining the influence of emoticons and text messaging on social
Journal of Interactive Marketing, 41, 60–80. interaction. Computers in Human Behavior, 69, 405–414.
Batra, R., & Keller, K. L. (2016). Integrating marketing communications: New Hsu, C.-L., Chang, K.-C., & Chen, M.-C. (2012). The impact of website quality
on customer satisfaction and purchase intention: Perceived playfulness and
findings, new lessons, and new ideas. Journal of Marketing, 80(6),
122–145. perceived flow as mediators. Information Systems and e-Business
Berger, J., & Milkman, K. L. (2012). What makes online content viral? Journal Management, 10(4), 549–570.
Huang, A. H., Yen, D. C., & Zhang, X. (2008). Exploring the potential effects
of Marketing Research, 49(2), 192–205.
Buchholz, K. (2019). In 2019, Global Emoji Count is Growing to More Than of emoticons. Information & Management, 45(7), 466–473.
3000. Retrieved from https://www.statista.com/chart/17275/number-of- Kaye, L. K., Wall, H. J., & Malone, S. A. (2016). Turn that frown upside-down:
emojis-from-1995-bis-2019/. A contextual account of emoticon usage on different virtual platforms.
Computers in Human Behavior, 60, 463–467.
Bundy, A. C. (1993). Assessment of play and leisure: Delineation of the
problem. American Journal of Occupational Therapy, 47(3), 217–222. Kelly, R., & Watts, L. (2015). Characterising the inventive appropriation of
Chall, J. S., & Dale, E. (1995). Readability Revisited: The New Dale-Chall emoji as relationally meaningful in mediated close personal relationships.
Readability Formula. Brookline Books. Paper presented at Experiences of Technology Appropriation: Unantici-
pated Users, Usage, Circumstances, and Design, Oslo, Norway.
Chen, X., & Siu, K. W. M. (2017). Exploring user behaviour of emoticon use
among Chinese youth. Behaviour & Information Technology, 36(6), Kronrod, A., & Danziger, S. (2013). ‘Wii will rock you!’ The use and effect of
637–649. figurative language in consumer reviews of hedonic and utilitarian
Churches, O., Nicholls, M., Thiessen, M., Kohler, M., & Keage, H. (2014). consumption. Journal of Consumer Research, 40(4), 726–739.
Labrecque, L. I. (2014). Fostering consumer–brand relationships in social
Emoticons in mind: An event-related potential study. Social Neuroscience,
9(2), 196–202. media environments: The role of parasocial interaction. Journal of
Clark, M., & Melancon, J. (2013). The influence of social media investment on Interactive Marketing, 28(2), 134–148.
Lin, C. S., Wu, S., & Tsai, R. J. (2005). Integrating perceived playfulness into
relational outcomes: A relationship marketing perspective. International
Journal of Marketing Studies, 5(4), 132–142. expectation-confirmation model for web portal context. Information &
Cohn, N. (2013). Beyond speech balloons and thought bubbles: The integration Management, 42(5), 683–693.
of text and image. Semiotica(197), 35–63. Lowrey, T. M., & Shrum, L. J. (2007). Phonetic symbolism and brand name
preference. Journal of Consumer Research, 34(3), 406–414.
Cohn, N., Engelen, J., & Schilperoord, J. (2019). The grammar of emoji?
Constraints on communicative pictorial sequencing. Cognitive Research: Luangrath, A. W., Peck, J., & Barger, V. A. (2017). Textual paralanguage and
Principles and Implications, 4(1) Article 33. its implications for marketing communications. Journal of Consumer
Cohn, N., Roijackers, T., Schaap, R., & Engelen, J. (2018). Are emoji a poor Psychology, 27(1), 98–107.
Luarn, P., Lin, Y.-F., & Chiu, Y.-P. (2015). Influence of Facebook brand-page
substitute for words? Sentence processing with emoji substitutions.
Proceedings of the 40th Annual Conference of the Cognitive Science posts on online engagement. Online Information Review, 39(4), 505–519.
Society, 1524–1529. Marengo, D., Giannotta, F., & Settanni, M. (2017). Assessing personality using
emoji: An exploratory study. Personality and Individual Differences, 112,
Cooper, B. B. (2014). A Scientific Guide to Posting Tweets, Facebook Posts,
Emails and Blog Posts at the Best Time. Retrieved from https://buffer.com/ 74–78.
resources/best-time-to-tweet-post-to-facebook-send-emails-publish- McLaughlin, G. H. (1969). SMOG grading - A new readability formula.
blogposts. Journal of Reading, 12(8), 639–646.
Miller, H. J., Thebault-Spieker, J., Chang, S., Johnson, I., Terveen, L., & Hecht,
Dale, E., & Chall, J. S. (1948). A formula for predicting readability:
Instructions. Educational Research Bulletin, 37–54. B. (2016). ‘Blissfully happy’ or ‘ready to fight’: Varying interpretations of
Deng, Q. (2019). Three essays on understanding consumer engagement with emoji. Proceedings of Tenth International AAAI Conference on Web and
brand posts on social media. doctoral dissertation . Sprott School of Social Media.Retrieved from https://www.aaai.org/ocs/index.php/ICWSM/
ICWSM16/paper/viewFile/13167/12746.
Business, Carleton University.
Derks, D., Bos, A. E. R., & Von Grumbkow, J. (2007). Emoticons and social Moon, J.-W., & Kim, Y.-G. (2001). Extending the TAM for a world-wide-web
interaction on the internet: The importance of social context. Computers in context. Information & Management, 38(4), 217–230.
Moore, S. G. (2012). Some things are better left unsaid: How word of mouth
Human Behavior, 23(1), 842–849.
influences the storyteller. Journal of Consumer Research, 38(6),
Derks, D., Fischer, A. H., & Bos, A. E. R. (2008). The role of emotion in
computer-mediated communication: A review. Computers in Human 1140–1154.
Behavior, 24(3), 766–785. Murdough, C. (2009). Social media measurement: It's not impossible. Journal
of Interactive Advertising, 10(1), 94–99.
Desta, Y. (2014). Why Is Everyone so Obsessed With Emojis? Retrieved from
https://mashable.com/2014/06/23/emoji-psychology/. Oppenheimer, D. M. (2006). Consequences of erudite vernacular utilized
Doyle, J. R., & Bottomley, P. A. (2011). Mixed messages in brand names: irrespective of necessity: Problems with using long words needlessly.
Separating the impacts of letter shape from sound symbolism. Psychology & Applied Cognitive Psychology: The Official Journal of the Society for
Applied Research in Memory and Cognition, 20(2), 139–156.
Marketing, 28(7), 749–762.
Emogi Research Team (2015). 2015 Emoji Report. Retrieved from https://cdn. Pancer, E., & Poole, M. (2016). The popularity and virality of political social
emogi.com/docs/reports/2015_emoji_report.pdf. media: Hashtags, mentions, and links predict likes and retweets of 2016 US
presidential nominees' tweets. Social Influence, 11(4), 259–270.
110 L. McShane et al. / Journal of Interactive Marketing 53 (2021) 96–110

Patrick, V. M., & Hagtvedt, H. (2012). ‘I don't’ versus ‘I can't’: When news/newstopics/howaboutthat/11614804/Emoji-is-Britains-fastest-
empowered refusal motivates goal-directed behavior. Journal of Consumer growing-language-as-most-popular-symbol-revealed.html.
Research, 39(2), 371–381. Thompson, D., & Filik, R. (2016). Sarcasm in written communication:
Preacher, K., & Hayes, A. (2008). Asymptotic and resampling strategies for Emoticons are efficient markers of intention. Journal of Computer-
assessing and comparing indirect effects in multiple mediator models. Mediated Communication, 21(2), 105–120.
Behavior Research Methods, 40(3), 879–891. Tossell, C. C., Kortum, P., Shepard, C., Barg-Walkow, L. H., Rahmati, A., &
Riordan, M. A. (2017). The communicative role of non-face emojis: Affect and Zhong, L. (2012). A longitudinal study of emoticon use in text messaging
disambiguation. Computers in Human Behavior, 76, 75–86. from smartphones. Computers in Human Behavior, 28(2), 659–663.
Scarpi, D. (2010). Does size matter? An examination of small and large web- Van Vleet, M., & Feeney, B. C. (2015). Play behavior and playfulness in
based brand communities. Journal of Interactive Marketing, 24(1), 14–21. adulthood. Social and Personality Psychology Compass, 9(11), 630–643.
Sledgianowski, D., & Kulviwat, S. (2009). Using social network sites: The Vargo, C. J. (2016). Toward a tweet typology: Contributory consumer
effects of playfulness, critical mass and trust in a hedonic context. Journal engagement with brand messages by content type. Journal of Interactive
of Computer Information Systems, 49(4), 74–83. Advertising, 16(2), 157–168.
Statista (2019). Most popular mobile social networking apps in the United Wall, H. J., Kaye, L. K., & Malone, S. A. (2016). An exploration of
States as of June 2019, by average session duration (in minutes). Retrieved psychological factors on emoticon usage and implications for judgement
from https://www.statista.com/statistics/579411/top-us-social-networking- accuracy. Computers in Human Behavior, 62, 70–78.
apps-ranked-by-session-length/. Willoughby, J. F., & Liu, S. (2018). Do pictures help tell the story? An
Steinmetz, K. (2015). Oxford's 2015 word of the year is this emoji. TIME. experimental test of narrative and emojis in a health text message
Retrieved from https://time.com/4114886/oxford-word-of-the-year-2015- intervention. Computers in Human Behavior, 79, 75–82.
emoji/. Zhao, X., Lynch, J. G., & Chen, Q. (2010). Reconsidering baron and Kenny:
Telegraph (2015). Emoji is Britain's Fastest Growing Language as Most Myths and truths about mediation analysis. Journal of Consumer Research,
Popular Symbol Revealed. Retrieved from https://www.telegraph.co.uk/ 37(2), 197–206.

You might also like