(In press, 2019). In C. Xie, F. Yus, & H. Haberland (Eds.

), Internet pragmatics: Theory and

practice. Amsterdam: John Benjamins.

Interpreting Emoji Pragmatics

Ashley R. Dainas Susan C. Herring

[email protected] [email protected]
Department of Information and Library Science
Indiana University, Bloomington

Abstract

This chapter describes the methods and the overall findings of the Understanding Emoji
Survey, which we administered online in early 2018 to determine how social media users
interpret the pragmatic functions of popular emoji types in the discourse context of comments
posted to public Facebook groups. The findings generally validate Herring and Dainas’s
(2017) taxonomy of graphicon functions for emoji, although survey respondents (n=523)
overwhelmingly preferred one function, tone modification, over the others. Moreover,
preferred interpretations of pragmatic function varied according to emoji type. Based on
these findings, we argue for the importance of analyzing emoji meaning from the perspective
of pragmatics.

Introduction

The popular press (at least in the English-speaking world) is currently rife with speculation
that emoji are becoming a new, global “language.”1 However, in order for a set of symbols
to become a language that can be used to communicate effectively with other people, users
of the symbols must agree on their meanings, and a number of studies have shown that even
within the same culture, internet users often disagree in their interpretations of emoji (e.g.
Miller, Thebault-Spieker, et al. 2016; Tigwell and Flatla 2016). Starting from the assumption
that the basic function of emoji is to express emotion, most of these studies have focused on
the emotion, sentiment, or mood conveyed by individual emoji in experiments involving
emoji presented either in isolation (e.g. Jaeger and Ares 2017, Miller, Thebault-Spieker, et
al. 2016) or in the context of Twitter tweets (e.g. Miller, Kluver, et al. 2017). A limitation of
this approach, however, is that it views meaning as residing solely on the semantic level, in
the emoji icons themselves, rather than in emoji-in-use. Emoji do not only express emotions
such as ‘happy’ or ‘sad’, or ‘positive’ or ‘negative’ sentiment; they also have pragmatic
functions whose meanings derive from the contexts in which they are embedded, such as
tone modification, illustration or repetition of accompanying text, and performance of virtual
actions (e.g. Herring and Dainas 2017). Thus research that seeks to understand how emoji
are interpreted in authentic contexts of use – which is where most people encounter them,
rather than as isolated tokens – needs to account for their pragmatic functions.

To address this need, we conducted an online survey, the Understanding Emoji Survey, to
determine how social media users interpret the pragmatic functions of 13 popular types of
emoji (smiles, frowns, winks, etc.) in the discourse context of comments posted to public

1
See e.g. Cohn (2015), Oliveira (2017), and Thompson (2016). Emoji are small, colorful graphical icons
used in text fields in digital communication and that represent facial expressions, objects, actions, and
symbols. The word emoji comes from Japanese e- (‘picture) + moji (‘character’).
 INTERPRETING EMOJI PRAGMATICS

Facebook groups. For each emoji-containing comment, respondents selected from a list of
functions that was adapted from Herring and Dainas’s (2017) taxonomy of graphicon
functions. This chapter describes the survey study methodology and reports on its overall
findings. Drawing on example survey items and the responses they received, we demonstrate
that although the semantics of individual emoji inform their pragmatic uses, understanding
emoji semantics is often insufficient to understand the intended meaning of emoji-containing
messages. Further, different emoji types specialize to varying degrees in expressing
particular pragmatic functions.

Another goal of the survey was to compare our interpretations, as researchers analyzing the
functions of emoji-in-use, with those of ordinary social media users, as a validity check. The
survey included anonymized examples of emoji-containing Facebook comments that we had
found challenging to classify in earlier research.2 The respondents’ interpretations of these
examples turned out to be less nuanced than ours: Tone modification was their default
interpretation, although the other categories from the Herring and Dainas (2017) pragmatic
function taxonomy were selected by a majority of respondents for at least some survey items,
thereby validating the taxonomy. Agreement rates among respondents and between
respondents and the researchers varied according to emoji type and function. Even with this
variance, the agreement levels far exceeded random chance, especially when we accounted
for respondents’ preference for tone modification. From this, we surmise that intersubjective
agreement on other, less challenging emoji uses should be even higher, increasing the
generalizability of the survey findings.

In the last sections of the chapter, we revisit the issue of emoji ambiguity and consider the
effectiveness of providing social media discourse context to clarify emoji users’ intended
meanings. Based on our findings, we also (re)evaluate the status of emoji as a “language”
with shared conventions and meanings.

Background Literature

Emoji as Language

Scholars of language and communication are increasingly joining journalists in advancing

the claim that emoji are developing into an independent, visual language. There is evidence
for this claim on multiple levels. Emoji use in social media is on the rise (Pavalanathan and
Eisenstein 2016; Pohl, Domin, and Rohs 2017), including uses of emoji alone without any
accompanying text. Emoji can serve various structural linguistic functions, such as letter
replacement and word replacement (Cramer, de Juan, and Tetreault 2016; Dürscheid and
Siever 2017), as well as substituting for entire propositions (Herring and Dainas, 2017).
Because they mostly appear at the end of sentences, they can also mark sentence boundaries
(Cramer et al. 2016), functioning like punctuation. Indeed, Pohl et al. (2017) argue that
although emoji lack a phonetic interpretation, they are themselves a form of text:

2
E.g. Herring and Dainas (2017) and subsequent unpublished research.

2
 INTERPRETING EMOJI PRAGMATICS

What makes emoji special as a means of adding visuals to texts is that they are
text. Instead of sending images of smileys or airplanes, characters representing
them are transmitted (they form a logographic writing system). Hence, in contrast
to images, they can be used in places such as URLs, email subjects, or usernames.
(6:5)

In addition, sequences or strings of emoji exhibit grammar-like properties, such as subject-

or stance-first word order (e.g. Danesi 2016; Steinmetz 2014). Danesi (2016) also suggests
that some strings of emoji have an iconic conceptual structure, and he notes the practice of
calquing, whereby emoji are directly mapped onto morphemes, words, or utterances of the
verbal language. The relationship of emoji and emoji sequences to the text they follow can
also be described in syntactic terms (Cramer et al. 2016; Pohl et al. 2017).

Most scholars stop short of considering emoji a fully-functioning language, however. They
point out that emoji mostly denote concrete objects, anthropomorphic facial expressions, and
(occasionally) actions, and that emoji sets lack icons for abstract concepts and grammatical
categories such as tense and number, articles, and conjunctions, which verbal languages
typically possess (e.g. Cohn 2015; Dürscheid and Siever 2017). Moreover, using only emoji,
one could not embed propositions inside other propositions or refer to strings of events other
than in chronological sequence.

Difficulties also exist at the level of meaning. As Miller, Kluver, et al. (2017:152) note, “in
order to avoid miscommunication incidents, people must interpret emoji characters in their
exchanges in the same way (and they must know that they are interpreting them the same
way)”. However, a number of studies have found that people vary in their understanding of
emoji semantics. These studies are discussed in the following section.

Emoji Semantics

Most research on receiver interpretations of emoji has focused on emoji semantics, either in
isolation or in very limited discourse contexts. As an example of the first type, Miller,
Thebault-Spieker, et al. (2016) asked Amazon Mechanical Turkers to rate the sentiment and
also describe the meaning of various isolated emoji renderings. They found within-platform
disagreement on sentiment in 25% of emoji renderings, as well considerable variation in
both within- and across-platform semantic descriptions of emoji. Tigwell and Flatla (2016)
had 70 participants situate eight Android OS emoji renderings and eight Apple OS emoji
renderings on a two-dimensional space, where the vertical axis represented intensity and the
horizontal axis represented a scale of negative to positive sentiment. They found individual
differences along both scales for each emoji. Further, the Apple and Android renderings of
the same emoji displayed distinctly different sentiment and intensity scores.3

Similar studies have been conducted in non-Western contexts. Annamalai and Abdul Salam
(2017) surveyed Malaysian students for their interpretations of isolated WhatsApp emoji
renderings. Not only were there varying levels of agreement among respondents on what a

3
In this chapter, we include sentiment and intensity, along with emotion, under the general domain of
semantics.

3
 INTERPRETING EMOJI PRAGMATICS

given emoji means, but the respondents often did not describe the emoji as intended by the
Unicode Consortium, the organization that approves emoji for inclusion in Unicode and
standardizes their definitions.4 In another study, Jaeger and Ares (2017) surveyed the
dominant consumer interpretations of isolated facial emoji by Mainland Chinese
participants. Out of 33 emoji, some (15) mapped to one emotion, another group (10) mapped
to multiple related emotions, and a final group (8) had multiple unrelated meanings.

In studies that consider emoji semantics in context, the context is usually a single Twitter
tweet (Barbieri et al. 2016; Miller, Kluver, et al. 2017). For example, Miller and colleagues
(2017) collected public English language tweets containing misinterpretation-prone emoji.
They filtered the tweets to exclude retweets, user mentions, hashtags, URLs, and other
attached media. Amazon Mechanical Turkers rated the sentiment of the emoji in the context
of the tweets, but agreement did not improve compared to Miller, Thebault-Spieker, et al.
(2016). In fact, the tweet context actually decreased the rate of agreement for some emoji.
However, it is not clear what role the context of the tweet itself played in this study’s results,
considering the length restrictions on tweets (120 characters at the time of the study), the
fact that no prior discourse context was included, and the fact that tweets containing retweets
and hashtags – common interactive components of tweets – were excluded.

Explanations for Semantic Ambiguity

Aside from insufficient context, a number of explanations have been proposed for the
semantic ambiguity of emoji. First, some emoji are inherently more ambiguous than others.
Miller, Thebault-Spieker, et al. (2017) found that the grinning face with smiling eyes ( ),
the unamused face ( ), and the smirking face ( ) (as rendered by Apple IOS, like some
of the emoji in the present study) had the most disparate agreement rates, whereas raters
agreed most on the sentiment of the heart eyes ( ), sleeping ( ), and crying face emoji
( ). Similarly, for Jaeger and Ares’s (2017) Chinese social media users, the tears of joy
( ), blushing face ( ), grimacing face ( ), and smirking face mapped to multiple
unrelated meanings, while the throwing (sic) a kiss ( ), loudly crying ( ), winking ( ),
heart eyes, smiley ( ), and the tongue out ( ) faces mapped predominantly to just one
emotion. The varying levels of ambiguity in emoji may be a feature that is inherent to
depictions of facial expressions (Choi, Hyun, and Lee 2017).

Further, as has often been pointed out, emoji render differently across platforms. The
Unicode Consortium creates a code for each emoji but does not specify how to render them.
While the letter A is almost always recognizable as the letter “A” regardless of its rendering,
emoji are more open to interpretation. Research shows that variations in rendering can
significantly alter the perceived meaning of some emoji (e.g. Miller, Thebault-Spieker, et al.
2016; Tigwell and Flatla 2016). This creates opportunities for miscommunication.

Finally, social and cultural factors may affect how users interpret emoji. While studies have
not found strong gender differences in emoji interpretation (Herring and Dainas 2018; Jaeger
et al. 2017), age may (Gullberg 2016; Herring and Dainas under review) or may not (Jaeger

4
http://unicode.org/reports/tr51/

4
 INTERPRETING EMOJI PRAGMATICS

et al. 2017) be a factor. Experience with emoji (Jaeger et al. 2017) and familiarity with one’s
online interlocutors (Tigwell and Flatla 2016) may also make emoji easier to interpret.
Finally, the language and culture of the community of users may influence emoji usage and
meaning (e.g. Barbieri et al. 2016).

In a departure from other researchers, Pohl et al. (2017) suggest that the ambiguity of emoji
may in fact be a strength: “Emoji meaning is fluid and subject to contextual and cultural …
interpretation. It is this malleability that makes emoji attractive from an expressive point of
view” (6:2). This observation underscores the importance of studying contextualized
interpretations of emoji-in-use.

Emoji Pragmatics

Compared to emoji semantics, less research has focused on the pragmatic functions of emoji.
However, numerous studies have identified pragmatic functions of emoticons, the
antecedents of emoji (e.g. Liebman and Gergle 2016; Lo 2008; Yus 2014; Walther and
D’Addario 2001). For example, research has shown that emoticons help clarify a sender’s
intended meaning, tone, emotion, attention, and self-presentation (Lo 2008; Ganster, Eimler,
and Krämer 2012). Yus (2014) created a taxonomy of pragmatic functions of emoticons that
includes mitigating, intensifying, or contradicting the propositional attitude expressed in the
sender’s text. Comparative studies show that emoji and emoticons share pragmatic
functions, especially tone marking (Herring and Dainas 2017), and do not appear to affect
message interpretation differently (Ganster et al. 2012). Despite this, users perceive emoji
as more appealing, familiar, clear, and meaningful (Rodrigues et al. 2018). Emoji also have
a stronger influence on the perceived mood and commitment of the sender (Ganster et al.
2012) than emoticons.

There is evidence that lay users are to some extent aware of the pragmatic functions of emoji
beyond simply expressing emotion. Kelly and Watts (2015)’s interviewees reported using
emoji to maintain or end conversations, to be playful, and to build rapport through shared
idiosyncratic uses. Participants in Gullberg (2016)’s focus group agreed to differing extents
that emoji could serve to enhance emotion, confirm receipt of a message, manage the
conversational climate (as a signal of friendliness, anger, or sincerity), maintain
relationships, and express one’s personal aesthetic.

The above studies were based on qualitative interviews and focus groups. In contrast, only
three studies to our knowledge have attempted a systematic accounting of the pragmatic
functions of emoji. Cramer et al. (2016) collected the most recently sent messages containing
emoji from 228 participants, along with user-reported descriptions of the emoji’s intended
meaning and function. The user descriptions revealed three categories of motivation for
emoji use: to add emotional or situational meaning, to add tone to text, and as a social tool
used to add flair, manage the conversation, and maintain relationships. The researchers also
identified three functions of emoji use: repetition of text, complementary usage, and text
replacement. Na'aman, Provenza, and Montoya (2017) attempted to train a classifier to
identify three high-level categories – Function, Content, and Multimodal – of emoji in tweets
based on annotations by four computational linguistics graduate students. The classifier

5
 INTERPRETING EMOJI PRAGMATICS

performed poorly, particularly with the Multimodal label, which was most akin to our
understanding of pragmatic functions, because there was low agreement among coders on
these variables. Na'aman et al. (2017) concluded that it is difficult to interpret emoji
functions even in context. Herring and Dainas (2017) adopted a more fine-grained, grounded
theory approach in describing pragmatic functions of graphical icons (graphicons5),
including emoji, in comments posted to public Facebook groups. In addition to tone
modification, they found that emoji perform virtual actions, express emotional reactions,
mention (repeat or illustrate) textual content, riff or elaborate playfully on prior messages,
and appear together in narrative sequences.

While Herring and Dainas (2017) were able to reach a high level of agreement in coding
graphicons after discussion, their taxonomy is based on researcher interpretations, which
may not correspond to how the pragmatic functions of emoji are understood by a lay
audience or the intended receivers. For one thing, researchers may spend time scrutinizing
instances of emoji use where most social media users would gloss over them. Yet there has
been, to our knowledge, no systematic comparison of researcher interpretations of emoji to
those of lay users. In the present study, we survey lay users about their interpretations of
emoji functions in context, using a modified version of the Herring and Dainas (2017)
taxonomy of pragmatic functions, and compare them to our researcher interpretations.

Research Questions

The general research question we seek to answer is: How do social media users interpret the
pragmatic functions of emoji in the discourse contexts in which they occur?

Specifically, we aim to address three questions:

RQ1: Which emoji functions are chosen most often, and for which emoji types?
We are interested to know if some pragmatic interpretations of emoji are preferred over
others, and also whether there are associations between the type (or the rendering) of an
emoji and the functions it expresses.

RQ2: To what extent do users agree among themselves on emoji functions?

Addressing this question should shed light on how ambiguous the functions are, and how
subject they are to misconstrual. It should also provide a basis for evaluating the efficacy of
discourse context in resolving pragmatic ambiguity.

RQ3: To what extent do user interpretations of emoji functions agree with the
researchers’ interpretations?
We address this to test the validity of our previous analyses and to validate the taxonomy of
pragmatic functions proposed in Herring and Dainas (2017).

5
Graphicons are graphical icons, including emoji, emoticons, stickers, images, GIFs, and video clips, that
can be used to convey propositional content in message exchanges (Herring and Dainas 2017).

6
 INTERPRETING EMOJI PRAGMATICS

Methods

Survey Design
Survey Items and Discourse Context
To construct the Understanding Emoji Survey, we collected 46 different examples of emoji
and their relevant prior discourse context from 14 public Facebook groups. These groups
were selected for their relatively high density of graphicon content as compared with other
public Facebook groups.6 Each example of emoji-in-use consisted of the message in which
the emoji appeared and the previous message(s) to which it most likely responded, as
determined by the researchers after reading the full comment thread. Due to space limitations
and to reduce possible distractions, we kept this context to a minimum, while preserving the
essential contextual information from the comment thread.

The most relevant discourse context was typically a post to the Facebook group to which the
emoji-containing message responded. This is because public Facebook comment threads are
usually prompt focused, meaning that most users respond directly to the prompt or initial
post of a thread rather than to other users (Herring 2013). Only a few examples (10.9%)
required reference to prior local (non-prompt) messages. The original prompt often
contained a large colorful image, GIF, or video together with text. In such cases, to minimize
distraction, we substituted a brief verbal description of the prompt for the original
multimodal prompt (e.g. Figure 1).

Each survey item included a Facebook comment containing (typically) a single emoji. In the
rare cases where an item included two or more different emoji, the instructions directed
respondents to focus on only one of them. Initially, items were selected based on the
difficulty we had encountered in coding those items in previous research. To these, we added
items to expand the representation of emoji types and pragmatic functions. The selected
messages were anonymized and simplified for the survey. We replaced user IDs with
pseudonyms that preserved the gender and nationality implied by the original names, in case
that information was relevant to the interpretation of the messages. For some items, we
modified the text of the message to make the topic of discussion accessible to a wider variety
of users. For example, we changed a reference to to a lesser-known book series to a better-
known book series.

Figure 1 shows an example of an emoji-containing message as it appeared in the survey,

along with its prior discourse context.

6
The Facebook groups that provided examples were: EmojiXpress, CatGIFs, AnimeGIFs, Nihilist Memes,
Grumpy Cat Memes, Smiley, Stickers, StickersFB, Rise of the Guardians, The Chronicles of Narnia, Star
Wars, Percy Jackson, Jared Padalecki, Selena Gomez.

7
 INTERPRETING EMOJI PRAGMATICS

[Prompt: Image announcing that a new book in a series is now available]

Alyssa Mueller: Either my parents won’t buy it or I’m not getting it until the 28th,
which is my birthday
Cecelia Silva: My baby's birthday is also the 28th. Lol had to throw that out there
happy early birthday

Figure 1. An emoji-containing message and its prior discourse context

Emoji Types
The emoji included in the survey represent 13 of the most common emoji types (smile, big
smile, frown, wink, blush, grimace, tears of joy, heart, heart eyes, blowing a kiss, crying,
tongue out, and “meh”7), which were rendered in the survey to match the emoji that appeared
in the original Facebook messages. For some types, the emoji renderings varied across
examples. We assume this is because the emoji were posted from different platforms (the
Facebook website or the mobile app) or from different mobile platforms (e.g., Apple or
Android), since emoji renderings differ on each of these (Miller et al. 2016). To preserve the
original context, in the survey we used a combination of Apple iOS 10 renderings and
screenshots of the emoji as they appeared on Facebook (Table 1).

Our pool of 46 messages included two to five examples of each emoji type. Four balanced
versions of the survey were created, each consisting of the same example question at the
beginning and 12 items drawn from the 45 remaining messages, with three items being
repeated once. The repeated questions were ones that we determined from our experience
and a pilot study to be most challenging; these were placed toward the end of the survey.
Thus each version of the survey contained at least one example of most of the 13 emoji types
and had a similar progression from easier-to-code items to more difficult ones.

Label Emoji Label Emoji

“meh” Heart Eyes

Big Smile Kiss

Blush Smile

Crying Tears of Joy

Frown Tongue Out

Grimace Wink

Heart

Table 1. Emoji types and emoji included in the survey

7
We use these shortened terms henceforth, rather than the Unicode labels, for the sake of brevity.

8
 INTERPRETING EMOJI PRAGMATICS

Pragmatic Functions
Herring and Dainas (2017) identified eight pragmatic functions of graphicons: Tone
Modification, Reaction, Action, Mention, Riff, Sequence, Ambiguous, and Other. (See
Table 2 for a description of each function.) Emoji constituted the most frequent graphicon
type in that study, and emoji were used in all eight functions. Thus we consider the taxonomy
to be well-suited to the analysis of emoji alone.

The original taxonomy was derived using a rigorous grounded theory approach. We allowed
the eight functions to emerge from our dataset of Facebook comments. Then we successively
refined our observations into a systematic coding scheme consisting of exhaustive, logically-
independent categories. Interrater reliability measures were calculated and proved
acceptable, and we discussed disagreements until consensus was reached; this process led to
further refinements to the coding scheme.

However, the rigor of this process notwithstanding, the pragmatic function taxonomy is
based on observations by researchers, who may tend to perceive more fine-grained
distinctions than ordinary social media users, in that the latter normally do not spend as much
time interpreting each instance of emoji use. Thus one goal of the current study is to
investigate whether or not the identified functions are also recognized by lay users and thus
can be said to have real-world validity. In order to achieve this, we simplified our definitions
of each function and translated them into language that is more accessible to a lay audience.
For example, rather than asking the respondents if an emoji was functioning as a “mention”,
the survey asked if the emoji was “being used to illustrate the text of the message comment”.

We modified the pragmatic function taxonomy in several respects for the purposes of the
Understanding Emoji Survey. First, we separated the Tone Modification code into two
codes: tone modification proper, or the use of an emoji to attribute a manner, attitude, or
emotion to the text it accompanies, and softening, the use of an emoji to hedge the
illocutionary force of the accompanying text (cf. Dresner and Herring 2010) in order to
mitigate, “soften”, or render more polite the act performed by the text. We included this
function because softening is sometimes associated with emoticon and emoji use in the CMC
literature (e.g. Baron and Ling 2011; Eisenchlas 2011).

Second, we added two logically possible categories that we expected might be chosen by
some respondents. Decoration indicates that the emoji is merely being used as decoration
without adding other meaning to the message. Physical action indicates that the message
sender was physically making the facial expression or doing the action depicted by the emoji.
We also added the option “I don’t know”, and renamed the Ambiguous function multiple
functions. Respondents who chose multiple functions or other were asked to write in an
explanation.

Finally, we determined that two of the codes from the original taxonomy were unnecessary,
given the examples included in the survey and our focus on emoji rather than graphicons
more generally. We omitted the Riff function, which we previously found to be associated

9
 INTERPRETING EMOJI PRAGMATICS

more with larger graphicons like GIFs. We also omitted Sequence, since it applies only to
two or more different emoji in a row, and no such examples were included in the survey.8

Function
Survey Options:
Formal Descriptions Additional
The emoji shows that
(Herring and Dainas 2017) Clarification
the user is...

That is, the emoji

Modification

Graphicon directly modifies Associating a specific

tells the reader how
text, clarifying how a tone (e.g. happy or
Tone

the comment is
message should be some related tone) with
intended to be
interpreted their comment
interpreted.
For example,
Softening

making the
Softening their
N/A comment less
comment
forceful or more
polite.
(Virtually) expressing
Graphicon used to portray a an emotion in response
Reaction

That is, reacting

specific emotion in response to previous content, not
directly to the
to something that has been necessarily related to
prompt.
posted the text of their
comment
That is, performing
(Virtually) saying [Text
each part of the
Action

Graphicon used to portray a of Message], and then

comment in
specific physical action performing a virtual
sequence, one part
action (e.g. smiling)
after the other.
Mentioning a graphicon That is, the emoji is
Mention

rather than using it, e.g., a graphic

Illustrating the text of
Use: “I’m so excited! :-) illustration of some
their comment
Mention: “That jerk had the word(s) in the
nerve to send me a :-) comment.
Graphicon is a humorous
elaboration on, play on, or
Riff

N/A N/A
parody of a previous
graphicon or comment.

A series of consecutive
graphicons (often of the
Sequence

same type) that convey a

N/A N/A
narrative of some kind as
opposed to a composite
message

8
Several survey items had two or more emoji in a row, but they were the same or semantically similar emoji
repeated for emphasis.

10
 INTERPRETING EMOJI PRAGMATICS

At the time the

Literally (physically) message was typed,

Expression
doing what the emoji the Facebook user

Physical
N/A expresses (e.g. smiling) was actually
while typing their feeling or doing
comment what the emoji
expresses.
The emoji has no
Decoration

function except to
Just using the emoji as
N/A make the text more
decoration.
visually interesting
or appealing.
Ambiguous

More than one function After considering

The graphicon has multiple, is equally plausible all of the options,
distinct meanings (Specify/Explain your you think there is
choices) no one best answer.

None of the above

options captures
Cases that cannot be
Other

how you think the

accounted for by the coding Other (Explain)
emoji functions in
scheme
this comment.
Instead you think...

N/A I have no idea You totally give up

Table 2. Formal and lay descriptions of pragmatic functions of emoji based on Herring and
Dainas (2017)

Table 2 shows the formal definitions of the pragmatic functions identified by Herring and
Dainas (2017), as well as the general function descriptions shown to the survey respondents.
Not shown are the specific versions of the function descriptions we crafted to match each
item in the survey. The last column in Table 2 shows the clarification of the pragmatic
function meanings provided in the example question at the beginning of each version of the
survey.

Multiple-Part Items
In the majority of survey items (n=37), the list of function options appeared directly below
the emoji-containing message. In addition to items of this basic type, we also created nine
multi-part items, in which there was a preliminary question that respondents had to answer
before being asked about the pragmatic function of the emoji in the message. Of these, six
were cases where the emoji was located in the middle of a textual string rather than at the
end or the beginning. Respondents first had to answer which part of the message the emoji
was associated with. Based on their answer, they were then shown a customized version of

11
 INTERPRETING EMOJI PRAGMATICS

the pragmatic function options. If they chose “I have no idea” for the first question, they
were advanced to the next item in the survey (Figure 2).

[Prompt: Announcement of the next Chronicles of Narnia movie – The Silver Chair]
Sara Conti Giordano: Ahhh… The Silver Chair. When is this supposed to show?
The emoji in this comment is most closely associated with:
o The first part (what comes before it)
o The second part (what comes after it
o Both
o Neither
o I have no idea

Figure 2. First part of a multiple-part survey item

The other three multiple-part questions featured the grimace face emoji. Many studies have
found that this emoji is prone to ambiguity and misunderstanding; we have also found it
difficult to interpret in our research. In order to get a sense of how the grimace emoji was
perceived in context, we asked respondents first to identify what emotion/attitude they
thought the emoji was expressing. Respondents were then shown a customized version of
the pragmatic function options based on their answer to that question. If respondents chose
“I have no idea” for the first question, they were advanced to the next item in the survey
(Figure 3).

[Prompt: "New Stickers! Rilakkuma by Sanrio" above an image of a series of stickers

involving bears and chicks]
Alice Williams Bateson: Update Facebook
What best describes the meaning of the above emoji?
o Grimacing/Forced smile
o Happy/Grinning widely
o Angry/Fierce
o I have no idea

Figure 3. First part of a grimace face multiple-part survey item

Pilot Study
As part of the process of developing the survey, we created two pilot versions, each
containing 23 emoji items drawn from the original pool of 46 messages. These versions were
shared with 14 individuals9 in the fall of 2017. The results of the pilot study were used to
refine the survey instrument. For example, based on feedback that the survey took too long,
we shortened it to 13 questions and created four versions of the survey instead of two. Based

9
The participants in the pilot study were graduate students in Information Science and friends and family of
the researchers, ranging in age from 25 to 62.

12
 INTERPRETING EMOJI PRAGMATICS

on confusion that some respondents had about the simplified descriptions of the pragmatic
functions, we fine-tuned the descriptions and added a sample question at the beginning of
the survey containing the expanded definitions of the pragmatic functions in Table 2.

Final Survey Structure

The final survey was created using Qualtrics,10 a cloud-based survey tool. Respondents took
the survey using Qualtrics’ online interface, and the outputted results of the survey were
collected using Qualtrics software.

When a potential survey respondent clicked on the link to the online survey, they were
provided with the study information sheet11 and were asked if they agreed to take the
survey.12 If they selected “I agree”, they were considered to have given informed consent,
and they continued to take the survey. If a respondent selected “I do not agree”, they were
sent to the end of the survey and thanked for their time.

The continuing respondents were next shown a block of demographic and social media usage
questions. Questions asked about the respondent’s gender, age, first language, and country
of residence. The questions about social media practices included whether the respondent
had an active Facebook account at the time. If so, they were asked about their posting
frequency, emoji use, and time spent on Facebook. All respondents were asked “In general,
how confident are you that you understand the intended meaning of emoji (other than
reaction emoji) when you see them in social media?”

Following these questions was the sample emoji item described earlier, presented with
expanded definitions of the pragmatic functions (see Table 2). Each continuing survey
respondent saw this sample item. Next, each respondent was randomly assigned one of the
four versions of the emoji item blocks by the Qualtrics Survey Software. Items were
presented one at a time.

After finishing the block of emoji items, the respondent was asked to rate how difficult it
was to interpret the emoji in the survey and to rate their confidence in their own answers.
They were also asked which graphicons they normally used (i.e. emoji, emoticons, stickers,
images, GIFs, video clips) and what other social media platforms they had accounts on. The
last question in the survey was open ended and asked “Do you have any other comments
about emoji use in social media?”

Distribution

The Understanding Emoji Survey ran between January 11 and February 20, 2018. The link
to the survey was shared with students and colleagues at a large North American university
as well as with friends, family members, and strangers via social media sites (i.e. Facebook,
Tumblr, Reddit, Twitter, and Ravelry). Initial respondents were encouraged to share the link

10
Versions available between August 2017 and February 2018. Copyright © 2005 Qualtrics.
https://www.qualtrics.com.
11
The information sheet stated that respondents should be between 18 and 75 years old.
12
The study was approved by the Indiana University Internal Review Board on August 9, 2017.

13
 INTERPRETING EMOJI PRAGMATICS

on their social media accounts as well as with other people they felt might be interested in
the survey, and in this way, the survey was distributed to a wider audience.

Quantitative Measures

The survey responses were analyzed in Microsoft Excel 2013. The results are presented
using descriptive statistics.

The frequency distributions of the responses to the multiple-choice items, normalized as

percentages, are presented in charts (e.g. Figures 4-14) and/or described in prose. For the
open-ended questions, including the ‘other’ and ‘multiple choice’ options where respondents
were asked to provide further details, the authors jointly conducted thematic content analysis
to group the responses into categories (Bauer 2000). The results of one of these analyses are
presented in Table 6; other results are discussed in prose.

Three kinds of agreement measures were also calculated. The first measured the degree to
which the respondents agreed among themselves on their preferred (top) function codes (see
Table 3). The second measured the degree to which the respondents agreed among
themselves on their top choice of function code (regardless of what it was) for each emoji
type (see Table 4). The final measure assessed the extent to which the respondents’ coding
choices agreed with the researchers’ codes (see Table 5). The details of these calculations
are described in the Agreement section further below.

Findings

Respondent Demographics

In all, 658 surveys were collected. In order to maximize the amount of usable information,
we analyzed the responses from all surveys in which a respondent selected both a gender13
and chose a function code for at least one emoji item beyond the sample question. 523
surveys met these requirements and were used for the analyses reported in this chapter.

The gender breakdown for the 523 surveys was 352 females, 121 males, and 50 ‘other’. The
average age was 28.6 (range: 18 to 70+). Most of the respondents (74.2%) were native
English speakers; the next most common native language was German (5.5%). Three-
quarters (75%) of the participants reported their country of residence as the U.S., while 4.4%
were based in Canada, 4.2% in Germany, and 2.7% in the U.K.

Respondents’ Social Media Usage

The survey respondents were active Facebook users and active emoji users. Most
respondents (n=445; 85.1%) reported having an active Facebook account. Of these, 74.4%
(n=331) said they check Facebook at least once a day, and 67.6% reported posting or
commenting on Facebook at least once a month. The majority of these respondents also
13
The survey findings are analyzed by gender in Herring and Dainas (2018) and by gender and age in
Herring and Dainas (under review).

14
 INTERPRETING EMOJI PRAGMATICS

reported using emoji on Facebook (not including the reaction emoji available after each
Facebook post and comment) ‘sometimes’ (38.9%), ‘often’ (31.2%), or ‘in every message’
(1.6%). Only 19.1% indicated that they use emoji ‘rarely’, and 9.2% said they ‘never’ use
emoji. Respondents reported using emoji most ‘mainly in private chat’ (42.2%), followed
by ‘in any kind of message’ (34.3%), and least often ‘mainly in [non-private] posts and
comments’ (14.8%). Only 8.5% of respondents with a Facebook account indicated that they
did not use emoji on Facebook.

The respondents were confident emoji interpreters. Of all who started the survey, the vast
majority reported being ‘very confident’ (55.4%) or ‘somewhat confident’ (39.2%) of their
ability to understand the intended meaning of emoji (other than reaction emoji) when they
encountered them in social media. Only 5.4% reported being ‘not at all confident’ in their
ability to understand the meaning of emoji. After responding to the survey items, however,
the respondents reported somewhat less confidence. Of the 454 people who finished all of
the emoji items, 34.6% reported being ‘very confident’ in their answers, 62.5% reported
being ‘somewhat confident’, and 2.9% reported being ‘not at all confident’. The lower
degree of (strong) confidence is perhaps not surprising, given that many of the survey items
were included precisely because, in our estimation, they were difficult to interpret.
Nonetheless, 15% of respondents reported that the survey was ‘very easy’, 52% said it was
‘somewhat easy’, and 20.3% found it ‘neither easy nor difficult’. Only 12.3% reported that
the survey was ‘somewhat difficult’, and only 0.4% reported that it was ‘very difficult.’

The survey respondents were active social media users. On average, respondents reported
having accounts on three social media platforms, not including Facebook. Out of the 433
people who responded, the majority reported having an Instagram account (62.1%), a Twitter
account (59.6%), and/or a Tumblr account (57.7%). Smaller numbers reported having a
Snapchat (44.6%), WhatsApp (33.5%), Reddit (20.6%), or Imgur (5.1%) account, and 17.1%
of users reported having at least one account on some other social media platform besides
the ones listed in the survey.

Respondents’ Interpretations of Pragmatic Functions

Overall
The function chosen most often in response to the emoji survey items was tone modification
(52.6%). Tone was the predominant choice for 39 out of the 49 items included in the survey.
Tone was followed by action (13.4%), mention (7.8%), softening (6.3), reaction (5.5%),
multiple functions (4.4%), decorative (3.3%), other (2.7%), “I don’t know” (2.5%), and
physical (1.4%), as shown in Figure 4.

15
 INTERPRETING EMOJI PRAGMATICS

60%
50%
40%
30%
20%
10%
0%
ne on on ng on ns ve er w" cal
To Acti enti fteni eacti nctio orati Oth kno hysi
M So R u c 't P
l eF De d on
i p "I
ult
M

Figure 4. Overall distribution of pragmatic function codes (N=6330) chosen by the survey
respondents

By Emoji Type
When broken down by emoji type, each pragmatic function has a distinctive emoji profile.
Hearts and kisses were especially interpreted as expressing virtual actions (Figure 6); smiles
and winks as softening the force of a message (Figure 7); grimaces and tears of joy as
reactions to a prompt (Figure 8); and kisses as mentions that illustrate message content
(Figure 9). Even tone marking was associated more with certain emoji (tongue out, crying,
frown) and less with others (e.g. grimace, kiss) (Figure 5). As for the additional options that
we included to supplement Herring and Dainas’s (2017) taxonomy, big smiles and hearts
were interpreted as decorative by some respondents (Figure 10), and some respondents
interpreted the heart eyes emoji as describing a physical action (described in the survey as
“looking adoringly” at one’s computer screen) (Figure 12). Finally, the fact that a number
of respondents chose multiple functions, other, or “I don’t know” suggests that for some
emoji items, the respondents were either not satisfied with the options provided in the survey,
or the functions of those emoji were especially difficult to interpret. The tears of joy emoji,
in particular, was said by several respondents to have other functions (e.g. laughing in a
mocking way) (Figure 14), and the grimace emoji received the most “I don’t know”
responses (Figure 11).

Individual Items
The survey did not generate enough replies to conduct meaningful quantitative analysis of
interpretation of function by emoji type at the level of individual emoji items. Nonetheless,
we observed variation among the items within a given emoji type, albeit not always where
variation might be expected. For the five emoji types with different renderings (blush,
crying, frown, tongue out, and wink; see Table 1), respondents choose the same top function
(tone) for each individual item. Conversely, the big smile, grimace, and heart were rendered
exactly the same in all survey items, yet the items elicited different functional interpretations
– for example, the three heart items had action, tone, and mention as their respective top
function choices. We can infer from these examples that the local discourse context, rather
than the emoji themselves, determined the respondents’ interpretation of the emoji’s
function.

16
 INTERPRETING EMOJI PRAGMATICS

Wink
Tongue Out
Tears of Joy
Smile
Kiss
Heart Eyes
Heart
Grimace
Frown
Crying
Blush
Big Smile
"meh"

0% 20% 40% 60% 80% 100% 0% 10% 20% 30% 40%

Figure 5. Tone Figure 6. Action

Wink
Tongue Out
Tears of Joy
Smile
Kiss
Heart Eyes
Heart
Grimace
Frown
Crying
Blush
Big Smile
"meh"

0% 10% 20% 30% 0% 10% 20% 30%

Figure 7. Softening Figure 8. Reaction

Wink
Tongue Out
Tears of Joy
Smile
Kiss
Heart Eyes
Heart
Grimace
Frown
Crying
Blush
Big Smile
"meh"

0% 10% 20% 0% 10% 20% 0% 10% 20%

Figure 9. Mention Figure 10. Decorative Figure 11. “I don’t know”

Wink
Tongue Out
Tears of Joy
Smile
Kiss
Heart Eyes
Heart
Grimace
Frown
Crying
Blush
Big Smile
"meh"

0% 5% 10% 0% 5% 10% 0% 5% 10%

Figure 12. Physical Figure 13. Multiple Figure 14. Other

17
 INTERPRETING EMOJI PRAGMATICS

The importance of context is also highlighted in cases where there is a mismatch between
the semantics of an emoji in isolation and how that emoji functions pragmatically in a
Facebook comment, as in the four examples that follow. Each example starts with a table
that displays the Unicode label for the emoji and the semantic labels applied to that emoji in
isolation by participants in previous studies. This is followed by a survey item containing
that emoji, a bar chart showing our respondents’ choices of pragmatic function for that item
(with the researchers’ interpretation circled), and a bar chart showing respondent choices for
all survey items containing that emoji type.

(1)

Internet sources
Rendering Unicode Jäger and Ares (2017)
(from Jäger and Ares 2017)
Happy (53%)
Slightly Content/satisfied (31%)
smiling Pleasure (28%) Smile, joy, happy, grinning
face Good (25%)
Fun (20%)

[Prompt: Image of a sticker store page featuring a few examples of the sticker set]
Margaret Blakey: I still can’t get that one. Any idea when?
Stickers for Facebook: Just wait, pls! it’ll slowly appear in your Sticker store! It’s
system is similar to fb interface Update, some people’s fb is updated, and others still
not. please be patient!

Stickers for Facebook All Smile Items

70% 50%
60% 40%
50%
40% 30%
30% 20%
20%
10% 10%
0% 0%
fte e

"
do O s

do O ns
n c al

n c al
ac n

le hy e

ac n
M tion

M tion
kn er

kn er
n

or g

n
fte e
or g

"
le hy e
ow
n
tip P ativ

ow
S o on

tip P ativ
So Ton
tio

tio
Re tio

Re tio
ec in

ec in
Fu sic

Fu sic
n't th

n't th
tio

tio
D n

D n
T
c

c
en

en
A

A
"I

"I
ul

ul
M

Previous research (Jäger and Ares 2017) found that the slightly smiling face in isolation has
meanings such as “happy”, “content”, “good”, and “joy”. However, the smiling emoji as
used by Stickers for Facebook in example 1 cannot be interpreted as expressing positive
emotion. The text of the message expresses the commenter’s irritation and perhaps
frustration with Margaret Blakey asking (potentially repeatedly, given the word “still”) when
a new sticker set will become available. The emoji serves to soften or mitigate Stickers for
Facebook’s irritated response; accordingly, softening was the preferred interpretation of both
the survey respondents and the researchers. Tone was selected more often for smile-

18
 INTERPRETING EMOJI PRAGMATICS

containing survey items overall, though, suggesting that the contexts of the other smile items
were different from that for Example 1.

(2)

Rendering Unicode Jäger and Ares (2017) Annamalai and Abdul

Salam (2017)
Flushed Surprised (40%)
Surprised/shocked (37%) Flushed (32.9%)
face
Shocked (27.1%)

[Prompt: Video of bloopers from the television show Supernatural]

Kaylin Durand: Are my eyes vibrating??

Kaylin Durand All Blush Items

50% 60%
40% 50%
30% 40%
30%
20% 20%
10% 10%
0% 0%
kn er

kn er
fte e
or g

"

fte e
or g

"
do O s

do O ns
n c al

n c al
ac n

le y e

ac n

le hy e
M tion

M tion
n

n
ow

ow
n
tip P ativ

tip P ativ
So Ton

So Ton
ec in

ec in
tio

tio
Re tio

Re tio
Fu sic

Fu sic
n't th

n't th
tio

tio
D n

D n
c

c
en

en
A

A
h

"I

"I
ul

ul
M

In previous research on emoji interpretation (Annamalai and Abdul Salam 2017; Jäger and
Ares 2017), the flushed face presented in isolation was ascribed meanings such as surprised,
shocked, and flushed. While it is possible that Kaylin Durand was shocked or embarrassed
by the video of bloopers (humorous out-takes) from the American television show
Supernatural, it seemed more likely (to us, and to many of our survey respondents) that
Kaylin chose this emoji to humorously illustrate vibrating eyes, since the eyes of the emoji
appear to be vibrating. This would be an example of the mention function. For this and the
other blush emoji examples, though, tone was the preferred respondent interpretation.

(3)

Rendering Unicode Jäger and Ares (2017) Annamalai and Abdul

Salam (2017)
Face with tears Happy (30%) Tears of joy (87.1%)
of joy Excited (21%) Funny (12.9%)

[Prompt: Image of a new book (play script), Harry Potter and Cursed Child]
畅畅: I want to buy~ But, my country haven't translation. And my English isn't good.

19
 INTERPRETING EMOJI PRAGMATICS

畅畅 All Tears of Joy Items

60%
50% 50%
40% 40%
30% 30%
20% 20%
10% 10%
0% 0%

kn r

kn r
fte e
or g

"

f e
or g

"
do O s

do O s
n c al

n c al
Re tion

tip P tive

Re tion

tip P tive
M tion

M tion
n

n
n't the

n't the
ow

ow
n

n
S o on

So Ton
ec n

ec n
tio

tio
Fu ic

Fu ic
tio

tio
D ni

D teni
T

le ys

le hys
a

a
c

c
ac

ac
en

en
A

A
h

"I

"I
ul

ul
M

M
The tears of joy emoji, in isolation, expresses joy, happiness, excitement, and amusement
(Annamalai and Abdul Salam 2017; Jäger and Ares 2017). However, none of these meanings
is clearly present in 畅畅’s use of the emoji in example 3; rather, the commenter is describing
negative circumstances: lack of a translation of a book they want to read and inability to read
it due to their poor language skills. We (the researchers) interpreted this emoji usage as
softening or mitigating what could otherwise be construed as a complaining or whining
comment. The survey respondents preferred tone (“associating a highly amused tone with
their comment”), which was the preferred function for the tears of joy emoji items overall,
and nearly 20% of respondents chose other; softening was their third choice. Cultural
differences in interpretation may be at work here. When the second author presented this
example in a talk recently, a Chinese woman in the audience recognized the softening use
of the tears of joy emoji and said it is not uncommon among Chinese social media users.

(4)

Rendering Unicode Jäger and Ares (2017) Internet sources (from

Jäger and Ares 2017)
Grimacing Surprised, awkward,
Surprised/shocked (18%)
face nervous, grimace

[Prompt: Image of a malevolent looking cat sitting in a box that says "Sour Puss"]
Amanda Jäger: Abdul Rahal

20
 INTERPRETING EMOJI PRAGMATICS

Amanda Jäger All Grimace Items

30% 30%
25% 25%
20% 20%
15% 15%
10% 10%
5% 5%
0% 0%

kn r

kn r
fte e
or g

"

f e
or g

"
do O s

do O s
n c al

n c al
Re tion

tip P tive

Re tion

tip P tive
M tion

M tion
n

n
n't the

n't the
ow

ow
n

n
S o on

So Ton
ec n

ec n
tio

tio
Fu ic

Fu ic
tio

tio
D ni

D teni
T

le hys

le hys
a

a
c

c
ac

ac
en

en
A

A
"I

"I
ul

ul
M

M
The grimacing face emoji, considered in isolation, is thought to express surprise,
awkwardness, or nervousness (Jäger and Ares 2017). Survey respondents who agreed among
themselves that ‘grimacing/forced smile’ was the meaning of the emoji in this item did not,
however, agree on the pragmatic function of the emoji. A narrow majority chose reaction
(to the prompt), as did the researchers, but action, tone, and “I don’t know” were also
popular choices. A similar lack of consensus is evident for all grimace face items. Thus this
emoji is ambiguous functionally as well as semantically.

Agreement

The charts in examples 1-4 show varying levels of respondent agreement, both among
themselves and with the researchers’ interpretations. To assess agreement levels overall, we
first calculated the degree to which the respondents agreed among themselves on their top
choice of function code for all the survey items. This was done by first counting the number
of items where each function was the predominant choice. The total number of respondents
who agreed with the predominant code was divided by the total number of respondents who
selected a code for those items; the results are shown in Table 3.

Excluding multiple functions and other, there are eight non-overlapping function options.
Five of these were selected as top choices for at least one item. Unsurprisingly, respondents
most often chose tone modification as their top choice, and they agreed most on that choice
at 59.7%, whereas they agreed least on reaction (only 25.9% of respondents chose it for the
two items where it was the top choice). Thus there was considerable disagreement on the
assignment of the five top pragmatic functions.

Nonetheless, all of the percentages are well above the level of chance, given the number of
possible code options provided in the survey. If the answers had been evenly distributed
across the eight non-overlapping function codes, we would expect each option to be selected
12.5% of the time. Or, since we know from Figure 4 that function codes were not evenly
distributed, if we take that distribution as a baseline and adjust it to exclude multiple
functions and other, the expected percentages would be as shown in the right-most column
of Table 3. For all functions except tone, which is only slightly more preferred as a top choice
than it was selected overall, the actual percent agreement on top choice functions is between

21
 INTERPRETING EMOJI PRAGMATICS

two and seven times greater than would be expected based on the adjusted overall
distribution of the function codes. From this we may conclude that respondents tended to
converge on functional interpretations of the emoji items, even though they did not approach
100% agreement.

Codes
Items on
Assigned Total Expected Expected
which Resp.
to Codes % % %
Agreed
Agreed- Assigned (Actual) (random (adjusted
with
Upon Top to Items dist.) dist.)
Themselves
Functions
Action 3 186 369 50.4% 12.5% 14.5%
Mention 3 139 360 38.6% 12.5% 8.4%
Reaction 2 59 228 25.9% 12.5% 6.0%
Softening 3 168 375 44.8% 12.5% 6.8%
Tone 39 3058 5124 59.7% 12.5% 56.7%
Total 50* 3610 6456 n/a 62.5%** 91.4%**
Average 55.9%
* There was a tie between softening and tone for one item, and each code was counted.
** Excluded from these totals are the percentages for decorative, physical, and “I don’t know”, which were
not top choices for any item.

Table 3. Inter-respondent agreement on top choice of pragmatic function

The degree of consensus among respondents varied by emoji type. Respondents agreed most
on the function of the tongue out emoji (82.5%), followed by the crying (74.9%) and frown
(73%) emoji. These are emoji for which tone was the most common top function choice.
The respondents had the lowest level of intersubject agreement on the functions of the
grimace (32%) and big smile emoji (38%), for which the top choices included reaction,
softening, and mention (Table 4). These results were calculated by dividing the total number
of respondents who chose the predominant choice for each item for each emoji type by the
total number of function codes that were assigned to each item for each emoji type.

Next we assessed the extent to which the respondents’ coding choices agreed with ours. This
involved calculating the number of items for which the respondents’ predominant function
choice and our choice were the same. When only their first or top choice was considered,
the respondents agreed with our tone modification codes in 100% of cases, although the
respondents coded more examples as tone than we did. The agreement rates for the other
functions were between 33.3% and 43.3% (Table 5). Examples 1 and 4 above illustrate items
where the respondents’ first choice agreed with our choice.

22
 INTERPRETING EMOJI PRAGMATICS

Survey Codes for All possible Top

%
Items Top Choice Codes Functions

Tongue
4 400 485 82.5% tone
Out
Crying 4 658 878 74.9% tone

Frown 4 356 488 73.0% tone

"meh" 3 230 351 65.5% tone

Tears of
4 278 490 56.7% tone
Joy
tone,
Smile 5 318 598 53.2%
softening
Wink 3 194 377 51.5% tone

Heart Eyes 2 127 248 51.2% tone

Blush 5 296 606 48.8% tone

action, tone,
Kiss 5 267 607 44.0%
mention
action, tone,
Heart 3 156 362 43.1%
mention
tone,
Big Smile 3 142 374 38.0% softening,
mention
tone,
Grimace 4 149 466 32.0%
reaction
Total 49 3571 6330
56.4% n/a
(Avg.) (3.8) (274.7) (486.9)

Table 4. Inter-respondent agreement on top choice of pragmatic function by emoji type

(with number of survey items for each emoji type)

However, the distribution of respondents’ function codes is skewed in favor of tone at the
expense of other functions. To adjust for this, we calculated a second agreement measure in
which we considered whether either the first or second choice of the respondents agreed with
our choices. This raised the agreement rate dramatically on virtual action, for example, from
33.3% to 83.3%, and it raised the overall level of agreement between researchers and
respondents from 67.3% to 95.9% (Table 5). Example 2 above illustrates an item for which
the respondents agreed with us on their second choice. On only four items (8.2%) did the
respondents not agree with our interpretations in either their first or second choice; example
3 above is one such case.

23
 INTERPRETING EMOJI PRAGMATICS

Researcher Agreed w/ % Agreed w/

Choices Researchers Researchers
%
on 1st on 1st or
choice 2nd choice
Action 6 2 33.3% 5 83.3%
Mention 7 3 42.9% 7 100.0%
Reaction 6 2 33.3% 4 66.7%
Softening 7 3 42.9% 6 85.7%
Tone 23 23 100.0% 24 100.0%
Total 49 33 67.3% 45† 91.8%
† This value is one less than the sum of the numbers in the column, because the respondents had tone as their
top choice one more time than the researchers did, and that was not counted as an agreement.

Table 5. Number and percentage of questions for which respondents agreed in their first-
or second-most-frequent choice with the researchers’ top code assignments

Multi-Part Questions
The survey included two kinds of multiple part questions. In the four out of six cases where
an emoji was located in the middle of a textual string rather than at the end or the beginning,
80% or more of respondents felt that the emoji was most closely associated with the first
part of the textual string. This is consistent with the tendency of emoticons and emoji to
appear at the end of textual strings (Cramer et al. 2016; Provine, Spencer, and Mandell 2007).
Respondents associated a variety of pragmatic functions with these emoji, except for
reaction, which was described in the survey as an emotional reaction unrelated to the text of
the message. For the other two multi-part items of this type, respondents slightly associated
the emoji with the first part of the text in one, and slightly associated it with the second part
in the other. In these examples, the favored pragmatic functions differed depending on how
respondents interpreted the positioning and scope of the emoji, although there were not
enough data to identify any recurring patterns.14

For the other kind of multi-part question, where the first part concerns the emotion/attitude
expressed by the grimace face, there was less consensus. The interpretation
“grimacing/forced smile” was preferred in all three examples, but only between 48% and
72% of respondents agreed on this categorization. Moreover, in two items, the second choice
interpretation was “happy/grinning widely”, and in the third it was “angry/fierce”. Overall,
12.3% of respondents said that they did not know what the emotion/attitude of the grimace
face was. There was also little consistency in pragmatic functions associated with each
emotion/attitude. This was particularly evident in one grimace item that was repeated in two
of the survey versions; interpretations of pragmatic function differed in the two iterations of
the same example. (One multi-part positioning item was also repeated, but its responses did
not show much variation.) These findings support previous research on the ambiguity of the
grimace face emoji.

14
The emoji differed in each of the six cases.

24
 INTERPRETING EMOJI PRAGMATICS

Open-Ended Responses
If respondents chose multiple functions, as they did in 257 cases, they were asked to list
which functions the emoji expressed. The most common combination of functions indicated
was tone and virtual action (18.7%). This confirmed our intuition that these two functions
are closely related, because we also sometimes had difficulty deciding between them in our
previous research. The next most common combination mentioned was tone and softening
(14.8%), which also makes sense, given that softening can be considered a subtype of tone
modification (and was not distinguished from it in the original Herring and Dainas
taxonomy). Most responses invoked one or more of the functions provided in the survey.
Respondents were most likely to include tone modification (79%) as one of the multiple
functions in a given example, followed by virtual action (33.9%) and mention (22.6%); this
distribution mirrors the overall distribution of function choices in Figure 4. Other
combinations of functions were idiosyncratic and did not fall into consistent categories, and
11% of combinations were indicated only once. Finally, some portion of the respondents
appeared not to have understood the functions provided, because they used the Multiple
Functions free response box to write in their own description of a single function that
matched one of those already provided (10.1%).

Respondents who chose other (n=157) were also asked to describe the function of the emoji.
The largest portion of other responses (51.6%) repeated the function categories provided in
the survey, such as tone modification or virtual action, with slightly different wording or
different descriptions of the emoji. The next most common other function (12%) mainly
described the emotion conveyed by the emoji. The remaining other function descriptions fell
into three main groups: 1) meta-pragmatic emoji functions such as signaling friendliness,
playfulness, or sarcasm (cf. Dresner and Herring 2010), 2) descriptions of the content of the
message, rather than the function of the emoji, and 3) item-specific observations (e.g. word
replacement, emoji misuse, evaluation, apology).

Of the 433 people who arrived at the end of the survey, approximately 25% responded to the
open-ended question asking if they had additional comments about emoji use in social
media. The results of a rough content analysis of those comments revealed eight basic
categories. The frequencies of these categories are presented in Table 6.

Comment Categories # %
I Love and Use Emoji (Like This) 28 27.5%
Emoji as a New Language 16 15.7%
Emoji are Annoying 14 13.7%
Ways Emoji Could be Better 14 13.7%
Comments on the Survey 9 8.8%
Emoji Rendering Problems 8 7.8%
Age Differences in Emoji Use 7 6.9%
Other Responses 6 5.9%
Total 102 100.0%

25
 INTERPRETING EMOJI PRAGMATICS

Table 6. Classification of comments in response to the question, “Do you have any other
comments about emoji use in social media?”

The comments were also revealing about respondents’ folk understandings of emoji
meanings and use. Some respondents asserted that emoji always or mainly express emotions:

Always used to express a mood. Emojis are used to express emotion, mood, and
represent items. When you tag a person and put a loving emoji next to it, it is the
feeling of those emoji that are connected to the person. Emojis are great for their
purpose of expressing feelings.

Other respondents focused on the tone modification function of emoji, e.g. I think it helps
convey tone, whereas plain text alone can make that difficult and Sometimes words can't
convey the proper tone; thats why emojis are so important.

Still other comments seemed to acknowledge that emoji fulfill other pragmatic functions,
e.g. It's an interesting way of conveying intent without words, and it's interesting to see how
a certain emoji can convey different meanings. (ie. sarcastic use vs. genuine use). Finally, a
few respondents highlighted the importance of the wider context to emoji interpretation, e.g.
Like any other form of communication, context matters, and the age, educational and
cultural background of the person using the emoji are significant to its interpretation.

Discussion

Research Questions Revisited

The overarching research question in this study was: How do social media users interpret
the pragmatic functions of emoji in their naturally-occurring discourse contexts?
Specifically, we first asked: Which emoji functions are chosen as interpretations most often,
and for which emoji types? The most common function chosen by our survey respondents
was overwhelmingly tone modification. At least 20% of respondents thought that the emoji
was functioning as a tone modifier in every item, with very few exceptions. Thus tone
modification appears to be the basic ‘meaning’ of emoji. Indeed, one could interpret almost
every item in the survey as tone marking, and the message would still be interpretable. We
might even go so far as to suggest, based on these results, that popular face-representing
emoji add tone by default, and that the other functions are in addition to tone marking. This
would capture the intuitions of the respondents who selected multiple functions and specified
tone in addition to another function. Virtual action was the second most common overall
function selected by the survey respondents, consistent with previous research on the use of
emoticons and emoji to represent nonverbal behavior (e.g. Derks, Bos, and Von Grumbkow
2007; Novak et al. 2015), followed by mention, softening, and reaction. With the exception
of softening, none of the options that we added to the original Herring and Dainas (2017)
taxonomy of pragmatic functions were chosen as top functions for any survey item.

The pragmatic interpretations preferred by the respondents varied by emoji type, as shown
in Figures 5-14. About 38% of the emoji types in the survey (big smile, grimace, heart, kiss,

26
 INTERPRETING EMOJI PRAGMATICS

and smile) included items for which the top choice of pragmatic function was something
other than tone modification. However, inter-respondent agreement rates were generally
higher for emoji types that favored tone (e.g. “meh”, blush, crying, frown, heart eyes, tears
of joy, tongue out, wink) (see Table 4). If inter-respondent agreement is taken as a measure
of the ambiguity of an emoji, emoji types that mainly express tone tend to be less ambiguous
than emoji that express other functions.

These findings correspond to some extent with the findings of studies of emoji semantic
ambiguity. Emoji with high agreement rates for pragmatic function include the crying and
tongue out faces, consistent with Miller, Thebault-Spieker, et al. (2016)’s and Jaeger and
Ares (2017)’s findings that these are some of the least ambiguous emoji semantically. The
blush and grimace emoji, which we found to have low inter-respondent agreement, were
similarly identified by Jaeger and Ares (2017) as especially ambiguous. However, we found
the tears of joy emoji to be less ambiguous pragmatically than Jaeger and Ares (2017) found
it to be ambiguous semantically (it marks tone, independent of how one interprets that tone).
Moreover, the kiss emoji is pragmatically somewhat ambiguous (is it performing a virtual
action? Imbuing the text with a loving tone? Illustrating the word ‘kiss’ in the message?),
whereas semantically it is unambiguous (Jaeger and Ares 2017). Interestingly, variations in
rendering did not noticeably impact how the emoji in the survey were interpreted, despite
the fact that some individual emoji of each type appear quite distinct (see Table 1).

The previous paragraph partially answers our second research question, which asked: To
what extent do users agree among themselves on emoji functions? The survey respondents
agreed on their top interpretations at a rate higher than chance, although agreement varied
according to emoji type, as noted above. The respondents agreed most on tone modification
and least on reaction and mention. Most lack of agreement resulted from some respondents
choosing tone as the default while other respondents chose less common functions.
However, even those lower rates of agreement were higher than chance. Thus, although the
survey respondents were probably unfamiliar with many of the pragmatic function options
they were asked to discriminate among, they were able to achieve a significant level of
agreement on their interpretations.

At the same time, overall agreement rates on functions did not exceed 60% (see Table 3).
While differences in methods mean that this number cannot be compared directly with the
numeric results of previous semantic studies of emoji ambiguity, this percentage shows that
there is considerable overall variability in the interpretation of emoji functions, even when
local discourse context is provided, leaving room for misconstrual and ambiguity.

Finally, in response to the third research question – To what extent do user interpretations
of emoji functions agree with the researchers’ interpretations? – the respondents agreed with
some of our interpretations for each of the five most commonly-selected functions: tone
modification, virtual action, softening, mention, and reaction, even when we interpreted
agreement strictly and considered only first choices, and when both their first and second
choices were considered, respondents agreed with most of our interpretations (see Table 5).
They agreed most with us on tone and least on action and reaction. These findings validate
the distinctions proposed in the taxonomy of pragmatic functions (Herring and Dainas 2017),

27
 INTERPRETING EMOJI PRAGMATICS

while also revealing that the distinctions are not all equally robust. Tone clearly outweighs
the others and, as suggested above, has a special status.

Emoji Ambiguity: Pros and Cons

Previous research has found that individual emoji tend to be semantically ambiguous (e.g.
Miller, Thebault-Spieker, et al. 2016). Our findings show that emoji tend to be pragmatically
ambiguous, as well, and that some emoji are more functionally ambiguous than others. That
said, it is unclear how much of an impediment misconstrued emoji pragmatics are to
successful communication. The flexibility of emoji use could be an advantage, rather than a
disadvantage, as suggested by Pohl et al. (2017). It allows users to be suggestive and to leave
their meanings open to interpretation, which might be desirable in some contexts. However,
it could also be a disadvantage if the message sender believes that they have communicated
clearly, but the recipient interprets the message in a different way.

Interlocutors may not realize that they have not understood a communication as it was
intended. Even though we found mixed levels of agreement, the respondents reported
generally high levels of confidence in their responses. Kruger, Epley, Parker, and Ng (2005)
identified a tendency for people to be overconfident in their ability to communicate
seriousness, sarcasm, anger, sadness, and humor over plain text email, as well as in their
ability to understand what was intended. Because of this overconfidence, interlocutors may
not realize that they have misunderstood, and communication may suffer as a result.
However, this problem is not limited to emoji; it also occurs in spoken communication
(Gumperz and Tannen 1979).

The Role of Discourse Context

Context plays an important role in discourse understanding. Our findings underscore the
importance of the local discourse context in determining emoji meanings. This context
matters more than the rendering of the emoji, as illustrated by the heart emoji items in the
survey (which render the same but occur in different contexts and have different
interpretations). Context also sometimes counts for more than emoji semantics, as illustrated
in examples 1-3, where the semantics of the emoji in isolation are marginally or not at all
relevant to the intended meaning of the emoji in the messages. This is not to imply that emoji
semantics play no role in the interpretation of pragmatic functions. In the case of tone
marking, for example, the actual tone conveyed (e.g. positive, loving, teasing, playful,
disgruntled) is usually cued by the sentiment of the emoji as well as the context. However,
the semantic meaning of an emoji alone is often insufficient to allow a recipient to interpret
the intended meaning (the illocutionary force) of an emoji-containing message.

Earlier, we suggested that the context of a tweet in previous studies may have been
insufficient to determine emoji meaning. The context we provided was richer. Although the
Facebook messages themselves were sometimes brief, we included prior context, as well as
user IDs that preserved gender and ethnicity information. Even so, our emoji items had
varying levels of contextual information, and the amount of available context appears to
affect the interpretability of the emoji. For example, the heart and grimace emoji items

28
 INTERPRETING EMOJI PRAGMATICS

tended to have limited context and correspondingly lower rates of agreement on their
functional interpretation.

The Status of Emoji as a Language

Meaning in language resides not just in the semantics of lexical items but also in the
pragmatics of their use. This study reveals emergent patterning for emoji at the level of
pragmatics, a level not previously considered in the debates about emoji as language. Our
findings strongly suggest that rather than simply expressing emotion, tone modification is
the basic function of emoji.

Emoji that function as tone modifiers, along with action and reaction emoji, could
conceivably be categorized as paralanguage, features that accompany verbal language that
“contribute to communication but are not generally considered to be part of the language
system.”15 Paralanguage includes facial expressions and gestures. Many emoji are faces, and
some emoticons and emoji reportedly function like gestures (Liebman & Gergle 2016;
McCulloch & Gawne 2018; Na’aman et al. 2017).

However, some emoji functions in the taxonomy used in this study do not clearly fit the
characterization of paralanguage (e.g. mention, decoration, some narrative sequences),
suggesting that while emoji can fulfill paralinguistic functions, their pragmatic range is more
expansive. Moreover, emoji are technically text (Pohl et al. 2017). Unlike paralanguage, they
are written (typed); there are a finite number of them; and they can substitute for words and
punctuation (Albert 2015). Like punctuation, they are illocutionary force markers (cf.
Dresner and Herring 2010). As such, emoji must be considered to be part of online language
at the pragmatic level.

This conclusion does not mean that emoji constitute a stand-alone language system.
Evidence from the literature indicates that emoji meanings and structural patternings, at least
in English-language contexts, are not (yet) conventionalized (e.g. Tatman 2016). Currently
their usage is flexible; their intended meanings can be open ended and imprecise, as
suggested by the variation in our survey responses. An exception is tone marking, which
appears to have become the conventional (default) interpretation of emoji use.

Conclusions

The Understanding Emoji Survey asked survey respondents to apply a taxonomy of

pragmatic functions to examples of emoji use in their local discourse contexts. Lay users
were able to assign pragmatic functions to emoji, despite not having seen the categories of
the taxonomy before, legitimizing the taxonomy but also revealing the privileged status of
tone modification as the default interpretation of emoji-in-use. It follows, therefore, that
researchers interested in how social media users understand emoji should not restrict their
study to emoji semantics but should also consider the pragmatic functions that motivate their
use. Also important is the finding that emoji are not functionally interchangeable: Different

15
Source: http://www.dictionary.com/browse/paralanguage

29
 INTERPRETING EMOJI PRAGMATICS

emoji types specialize to some extent for specific functions. This exploratory study has
provided preliminary insights into the functional specialization of 13 popular emoji.

The study also contributes to emoji research methodologically. Miller and her colleagues
used surveys effectively to collect evidence of how people understand emoji semantics. Our
study showed that surveys can also be used to assess lay user understandings of the pragmatic
functions of emoji-in-use. Further refinement of the survey method could lead to more
systematic study of which emoji are typically used for which specific functions, as well as
teasing out the contributions to overall meaning of the semantics versus the pragmatic
functions of emoji.

A limitation of this study is that the context provided for the Facebook messages was local
discourse context only. A thorough study of emoji pragmatics requires consideration not just
of the local discourse context,16 but also situational, interpersonal, and cultural contexts.
Even so, as Cramer et al. (2016) note, it may not be possible to interpret some emoji usage
correctly without understanding the sender’s intention due to idiosyncrasy (e.g. in-jokes,
private language). A possible way around this is to supplement survey and experimental
research with focus groups and interviews, ideally with the individuals who used the emoji.

An important variable is respondent age. A number of respondents commented at the end of

our survey that there are generational differences in emoji usage and understanding. Indeed,
Herring and Dainas (2018) found that the responses of the ‘other’ gender, which comprised
the youngest group of respondents, differed from those of the male and female respondents,
who were somewhat older on average. More recently, Herring and Dainas (under review)
analyzed the survey results based on respondents’ self-reported age and found systematic
generational differences in some emoji interpretations. This is an area in need of further
investigation.

In other future research, the pragmatic function taxonomy – expanded to include softening
– could be used to classify emoji meanings on other social media platforms, in order to
support and generalize from the present findings. Manual analysis could be supplemented
by automated analysis based on the taxonomy to allow larger amounts of data to be analyzed.
Researchers should also investigate the private usage of emoji in texting and chat, as public
comment threads may not be representative of other kinds of message exchanges.

Finally, the taxonomy could be applied to study how social media users understand other
types of graphicons-in-use. Emoji are currently the most popular graphicon type, but on
some platforms stickers, GIFs, and image macros are prominent features of computer-
mediated communication (Herring, 2018). Their interpretations by lay social media users
have yet to be explored.

16
Including the context of the thread. In this study, because of the prompt-focused nature of the Facebook
threads, the other comments were typically judged irrelevant to the interpretation of a given emoji-containing
comment.

30
 INTERPRETING EMOJI PRAGMATICS

References

Albert, Georg. 2015. Semiotik und Syntax von Emoticons. Zeitschrift für Angewandte
Linguistik 62 (1): 3–22. DOI: https://doi.org/10.1515/zfal-2015-0001

Annamalai, Subashini, and Sobihatun Nur Abdul Salam. 2017. "Undergraduates'

Interpretation on WhatsApp Smiley Emoji." Jurnal Komunikasi, Malaysian
Journal of Communication 33 (4): 89–103.

Barbieri, Francesco, German Kruszewski, Francesco Ronzano, and Horacio Saggion. 2016.
"How Cosmopolitan are Emojis?: Exploring Emojis Usage and Meaning over
Different Languages with Distributional Semantics." In Proceedings of the 2016
ACM on Multimedia Conference, 531–535. New York, NY: Association for
Computing Machinery. DOI: 10.1145/2964284.2967278

Bauer, Martin. 2000. “Classical Content Analysis: A Review.” In Qualitative Researching

with Text, Image and Sound, ed. by Martin Bauer, and George D. Gaskell, 131–
151. Thousand Oaks, CA: Sage.
DOI: http://dx.doi.org.proxyiub.uits.iu.edu/10.4135/9781849209731.n8

Baron, Naomi S., and Rich Ling. 2011. "Necessary Smileys & Useless Periods." Visible
Language 45 (1/2): 45-67.

Choi, Yuri, Kyung Hoon Hyun, and Ji-Hyun Lee. 2017. "Image-Based Tactile Emojis:
Improved Interpretation of Message Intention and Subtle Nuance for Visually
Impaired Individuals." Human–Computer Interaction. DOI:
10.1080/07370024.2017.1324305

Cohn, Neil. 2015, October 13. “Will Emoji Become a New Language?” BBC - Future.
http://www.bbc.com/future/story/20151012-will-emoji-become-a-new-language

Cramer, Henriette, Paloma de Juan, and Joel Tetreault. 2016. "Sender-Intended Functions
of Emojis in US Messaging." In Proceedings of the 18th International Conference
on Human-Computer Interaction with Mobile Devices and Services, 504-509. New
York, NY: Association for Computing Machinery. doi>10.1145/2935334.2935370

Derks, Daantje, Arjan E. R. Bos, and Jasper Von Grumbkow. 2007. "Emoticons and Social
Interaction on the Internet: The Importance of Social Context." Computers in
Human Behavior 23 (1): 842-849. https://doi.org/10.1016/j.chb.2004.11.013

Dresner, Eli, and Susan C. Herring. 2010. “Functions of the Non-Verbal in CMC:
Emoticons and Illocutionary Force.” Communication Theory 20: 249–268.
https://doi-org.proxyiub.uits.iu.edu/10.1111/j.1468-2885.2010.01362.x

Eisenchlas, Susana A. 2011. "On-Line Interactions as a Resource to Raise Pragmatic

Awareness." Journal of Pragmatics 43 (1): 51–61.
https://doi.org/10.1016/j.pragma.2010.08.013

31
 INTERPRETING EMOJI PRAGMATICS

Gullberg, Kajsa. 2016. Laughing Face with Tears of Joy: A Study of the Production and
Interpretation of Emojis Among Swedish University Students. Bachelor’s Thesis,
Lund University Department of Linguistics.

Gumperz, John J., and Deborah Tannen. 1979. “Individual and Social Differences in
Language Use.” In Individual Differences in Language Ability and Language
Behavior, ed. by Charles Fillmore, Daniel Kempler, and William S-Y. Wang, 305–
325. New York, NY: Academic. https://doi.org/10.1016/B978-0-12-255950-
1.50024-X

Herring, Susan C. 2013. “Discourse in Web 2.0: Familiar, Reconfigured, and Emergent.”
In Georgetown University Round Table on Languages and Linguistics 2011:
Discourse 2.0: Language and New Media, ed. by Deborah Tannen and Anna M.
Tester, 1–25. Washington, DC: Georgetown University Press.

Herring, Susan C. 2018. “Emergent Forms of Computer-Mediated Communication and

their Global Implications.” Linguapax Review 2017.
http://www.linguapax.org/english/publications/linguapax-review-en/linguapax-
review-2017

Herring, Susan C., and Ashley R. Dainas. 2017. "‘Nice Picture Comment!’ Graphicons in
Facebook Comment Threads.” In Proceedings of the Fiftieth Hawai'i International
Conference on System Sciences (HICSS-50). Los Alamitos, CA: IEEE Press.
DOI: 10.24251/HICSS.2017.264

Herring, Susan C., and Ashley R. Dainas. 2018. “Receiver Interpretations of Emoji
Functions: A Gender Perspective.” In Proceedings of the Emoji2018 Workshop,
ICWSM 2018, June 25, Stanford, CA. Published at http://ceur-ws.org

Herring, Susan C., and Ashley R. Dainas. Under review. “Gender and Age Influences on
Interpretation of Emoji Functions.”

Jaeger, Sara R., and Gastón Ares. 2017. "Dominant Meanings of Facial Emoji: Insights
from Chinese Consumers and Comparison with Meanings from Internet
Resources." Food Quality and Preference 62: 275–283.
https://doi.org/10.1016/j.foodqual.2017.04.009

Jaeger, Sara R., Yixun Xia, Pui-Yee Lee, Denise C. Hunter, Michelle K. Beresford, and
Gastón Ares. 2017. "Emoji Questionnaires Can Be Used With a Range of
Population Segments: Findings Relating to Age, Gender and Frequency of
Emoji/Emoticon Use." Food Quality and Preference 68: 397–410.
https://doi.org/10.1016/j.foodqual.2017.12.011

Kelly, Ryan, and Leon Watts. 2015. "Characterising the Inventive Appropriation of Emoji
as Relationally Meaningful in Mediated Close Personal Relationships." Paper

32
 INTERPRETING EMOJI PRAGMATICS

presented at Experiences of Technology Appropriation: Unanticipated Users,

Usage, Circumstances, and Design, September 20, Oslo, Norway.

Kruger, Justin, Nicholas Epley, Jason Parker, and Zhi-Wen Ng. 2005. "Egocentrism Over
E-Mail: Can We Communicate as Well as We Think?" Journal of Personality and
Social Psychology 89 (6): 925- 936. 10.1145/2818048.2819945

Liebman, Noah, and Darren Gergle. 2016. "It's (Not) Simply a Matter of Time: The
Relationship between CMC Cues and Interpersonal Affinity." In Proceedings of the
19th ACM Conference on Computer-Supported Cooperative Work & Social
Computing, 570-581. New York, NY: Association for Computing Machinery. DOI:
http://dx.doi.org/10.1145/2818048.2819945

Lo, Shao-Kang. 2008. “The Nonverbal Communication Functions of Emoticons in

Computer-Mediated Communication.” CyberPsychology & Behavior 11 (5): 595-
597. DOI: 10.1089/cpb.2007.0132

Miller, Hannah, Daniel Kluver, Jacob Thebault-Spieker, Loren Terveen, and Brent Hecht.
2017. "Understanding Emoji Ambiguity in Context: The Role of Text in Emoji-
Related Miscommunication." In Proceedings of the 11th International Conference
on Web and Social Media (ICWSM 2017). Palo Alto, CA: Association for the
Advancement of Artificial Intelligence.

Miller, Hannah, Jacob Thebault-Spieker, Shuo Chang, Isaac Johnson, Loren Terveen, and
Brent Hecht. 2016. “‘Blissfully Happy’ or ‘Ready to Fight’: Varying
Interpretations of Emoji.” In Proceedings of the 10th International Conference on
Web and Social Media (ICWSM 2016). Palo Alto, CA: Association for the
Advancement of Artificial Intelligence.

Na'aman, Noa, Hannah Provenza, and Orion Montoya. 2017. "Varying Linguistic Purposes
of Emoji in (Twitter) Context." In Proceedings of ACL 2017, Student Research
Workshop, 136-141. Stroudsburg, PA: Association for Computational Linguistics.
https://doi.org/10.18653/v1/P17-3022

Novak, Petra Kralj, Jasmina Smailović, Borut Sluban, and Igor Mozetič. 2015. "Sentiment
of Emojis." PloS One 10 (12) https://doi.org/10.1371/journal.pone.0144296

Oliveira, Joana. 2017, August 16. “Emoji, the New Global Language?” OpenMind.
https://www.bbvaopenmind.com/en/emoji-the-new-global-language/

Pavalanathan, Umashanthi, and Jacob Eisenstein. 2016. “More Emojis, Less :) The
Competition for Paralinguistic Function in Microblog Writing.” First Monday 21
(11). DOI: http://dx.doi.org/10.5210/fm.v21i11.6879

Pohl, Henning, Christian Domin, and Michael Rohs. 2017. “Beyond Just Text: Semantic
Emoji Similarity Modeling to Support Expressive Communication.” ACM

33
 INTERPRETING EMOJI PRAGMATICS

Transactions on Computer-Human Interaction 24 (1): article 6.

doi>10.1145/3039685

Provine, Richard R., Robert J., Spencer, and Darcy Mandell. 2007. “Emotional
Expression Online: Emoticons Punctuate Website Text Messages.” Journal of
Language and Social Psychology 26 (3): 299–307. https://doi-
org.proxyiub.uits.iu.edu/10.1177/0261927X06303481

Rodrigues, David, Marília Prada, Rui Gaspar, Margarida V. Garrido, and Diniz Lopes.
2017. "Lisbon Emoji and Emoticon Database (LEED): Norms for Emoji and
Emoticons in Seven Evaluative Dimensions." Behavior Research Methods 50 (1):
392–405. https://doi-org.proxyiub.uits.iu.edu/10.3758/s13428-017-0878-6

Tatman, Rachael. 2016, December 7. “Do Emojis Have Their Own Syntax?” [Blog post].
Making Noise & Hearing Things.
https://makingnoiseandhearingthings.com/2016/12/07/do-emojis-have-their-own-
syntax/

Thompson, Clive. 2016, April 19. “The Emoji is the Birth of a New Type of Language (No
Joke).” Wired. https://www.wired.com/2016/04/the-science-of-emoji/

Tigwell, Garreth W., and David R. Flatla. 2016. “Oh That’s What You Meant!: Reducing
Emoji Misunderstanding.” In Proceedings of the 18th International Conference on
Human-Computer Interaction with Mobile Devices and Services Adjunct
(MobileHCI ’16), 859–866. New York, NY: Association for Computing
Machinery. DOI:10.1145/2957265.2961844

Walther, Joseph B., and Kyle P. D’Addario. 2001. "The Impacts of Emoticons on Message
Interpretation in Computer-Mediated Communication." Social Science Computer
Review 19 (3): 324-347. https://doi-
org.proxyiub.uits.iu.edu/10.1177/089443930101900307

Yus, Francisco. 2014. “Not All Emoticons Are Created Equal.” Linguagem em (dis)curso
14 (3): 511-529. http://dx.doi.org/10.1590/1982-4017-140304-0414

34