Professional Documents
Culture Documents
net/publication/355043055
CITATION READS
1 259
2 authors:
Some of the authors of this publication are also working on these related projects:
Research Unit "Emerging Grammars in Language Contact Situations: A Comparative Approach", Project P6: "Noncanonical constituent linearisation in German across
heritage speakers" View project
Namdeutsch: The dynamics of German in the multilingual context of Namibia View project
All content following this page was uploaded by Heike Wiese on 04 October 2021.
Humboldt-Universität zu Berlin
Abstract
comparably novel devices in the rapidly evolving domain of digital messenging, emoji
provide an interesting example to observe change in progress. We present a corpus study and
an experimental study. Main results are (i) an overall salience of subjective and
intersubjective discourse meanings for emoji, with (ii) a general advantage for the former,
especially for emoji that iconically include more active elements, while (iii) dominance
relations can be modulated by left- vs. right-peripheral positions in favor of subjective vs.
studies contribute to our understanding of their pragmatic contribution and provide novel
Keywords
LP/RP differences
Phenomena of language variation can highlight ongoing tendencies of change and this is also
the dynamics of pragmatic markers in progress might shed a light on functional preferences
for different loci. A promising empirical domain for this is digital messenging and the
comparably new and rapidly developing domain of WhatsApp® messages has a special
potential to reveal patterns of variation and change, given its prevalence as a means of
restrictions.
WhatsApp® was developed in 2009 and, according to the company's website, it has
currently over two billion users worldwide.1 Instant messengers can be used for fast and
spontaneous communication among friends and family members in everyday life but there
are also additional usages, e.g. among work colleagues. They are not restricted to two
communication partners but also support group conversations, e.g. for school classes or
extended families, and temporary WhatsApp® groups are often used for organizing joint
activities.
Informal written language as used in WhatsApp® is interesting for at least two reasons.
First, informal language in general is particularly dynamic and open to innovations, since it is
less subject to the normative restrictions of standard language ideologies. Informal language
messages and similar computer- or digital-mediated communication, speakers do not only use
verbal markers but have also developed typographic and orthographic register-specific means
1
See https://www.whatsapp.com/about/ (last accessed 20.04.2020).
(e.g. Thurlow and Poff 2013, Androutsopoulos 2020a, 2020b). Among these are non-standard
spellings, punctuations and capitalizations, abbreviations and the use of emoji and emoticons
(e.g. Bieswanger 2013, Danesi 2017, Lamontagne and McCulloch 2017, Gawne and
McCulloch 2019; for recent studies on different varieties of German, see contributions in
Androutsopoulos and Busch 2020 and Marx et al. 2020; for recently published corpora, see
In this study, we are particularly interested in emoji as comparatively new devices that
perspective, we include emoji under the notion of discourse markers, which we understand as
a subset of pragmatic markers (cf. Fraser 2009). Following main approaches from the
literature, we define discourse markers (DM for short) as elements that are not fully
syntactically integrated and do not directly contribute to the propositional meaning and truth
value of an utterance but rather operate on the level of discourse (cf. Blakemore 2004, Fraser
2006, Blühdorn et al. 2017). This is a narrow definition in that it excludes modal particles and
connectives, as both are syntactically integrated. At the same time, as we will show in more
detail below, the definition is broad enough to include both verbal and graphic elements.
Furthermore, since it does not make any assumptions about the grammatical status of the
elements, it covers single as well as multi-word items (e.g. you know) and is also open, for
Another important point for our investigation is that the definition does not restrict DMs to a
particular position, e.g. the left periphery (as suggested for discourse markers by, for
instance, Imo 2012, who uses the term “discourse particles” instead for the more general
sense that we use for DMs here). Capturing DMs in both left and right peripheries will allow
Müller 2005), subjective and intersubjective (e.g. Traugott 2010, 2012). The textual function
of DMs is primarily oriented towards discourse contents and text organization, whereas the
subjective and intersubjective functions of DMs are primarily oriented towards discourse
participants. Subjective DMs are oriented towards the speaker and encode the speaker’s
stance and attitudes through evaluation of discourse contents, and their side in turn-taking,
especially through turn-keeping. Intersubjective DMs are oriented towards the addressee;
they encode the speaker’s attention to the social identity of the addressee and the addressee’s
“face” and can signal politeness (including hedging), elicitation of responses (including
management.
associations with emerging subjective vs. intersubjective functions of DMs, respectively (cf.
also Degand and Fagard 2011, Beeching and Detges 2014, Detges and Waltereit 2014). The
hypothesis states that elements in the left periphery (LP for short) are more likely to fulfil
subjective discourse functions, whereas elements in the right periphery (RP) are more likely
to be used for intersubjective functions. In this context, left and right peripheries are
understood with respect to sentences, identifying positions to the left or right, respectively, of
a strong tendency rather than a strict rule (e.g. Traugott 2012, Beeching and Detges 2014).
From a diachronic perspective, the hypothesis suggests that “expressions recruited to LP are
German WhatsApp® messages. While previous research has focused on verbal DMs, we will
focus on these graphic DMs. This adds an interesting additional empirical area to the
discussion, since emoji are comparably novel and particularly dynamic devices belonging to
In what follows, we first provide the background on emoji as graphic DMs (Section 2)
and then present two empirical studies: a corpus study that explores positions and discourse
study based on the findings from the corpus study, which tests subjective vs. intersubjective
interpretations of emoji in LP vs. RP positions in such messages (Section 4). The final section
Emoji are pictures based on Unicode. Coming from Japanese, the word emoji combines the
meanings ‘picture’ (e), ‘write’ (mo) and ‘character’ (ji). These graphic symbols were first
introduced in 1999 by Shigetaka Kurita (Bai et al. 2019). Unlike emoticons, which have
already been in use for a longer time, emoji are not limited to facial expressions but can also
depict objects (e.g. animals or food), actions (e.g. clapping hands) and symbols (e.g. flags).
Meanings of emoji found in the literature (e.g. Al Rashdi 2015, Cramer et al. 2016, Danesi
2017, Dürscheid and Siever 2017, Pappert 2017, McCulloch 2019, Beißwenger and Pappert
2019, Dürscheid 2020a, 2020b, Dainas and Herring forthc.) are manifold and not labeled
consistently. Understanding them as DMs in the sense defined above allows us to systematize
their diverse pragmatic contributions by subsuming them under subjective, intersubjective,
Emoji fulfil all the criteria we listed for DMs. As an illustration, consider (1) and (2),
taken from the RUEG corpus with WhatsApp® messages where speakers described a traffic
accident they just saw (Wiese et al. 2019; see below for details of the corpus).
(RUEG-DE_0.3.0, DEmo44FD)
(RUEG-DE_0.3.0, DEmo43MD)
In these examples, the emoji are not syntactically integrated but are rather placed outside the
the propositional meaning and truth value of the utterances but rather operate on the level of
discourse: they do not change the lexical semantics of the sentences but add a discourse
meaning to the message. In (1), the kiss emoji adds an intersubjective element of intimacy or
friendliness towards the addressee; in (2), the laughing-with-tears emoji conveys a subjective
narrative.
Note that in both cases, the emoji are not standing for the literal meaning depicted
iconically: in (1), the speaker does not convey that she is actually kissing at the moment of
writing and, in (2), the speaker does not express that he is in fact laughing with tears. As
pointed out by Glikson et al. (2017: 614), who found differences between the iconic meaning
of emoticons and their perceptions: “a smiley is not a smile” (cf. also Sergeant 2019, Albert
2020).
This points to semantic bleaching of the original iconic meaning, suggesting a process
observed for verbal DMs. Verbal DMs are often derived from referential elements with
lexical meanings through a process of semantic bleaching, where they lose this original
(e.g. Traugott 2020). In this process, the original element typically does not get lost but forms
a lexical counterpart to the DM. For instance, in English, man or you know are often used as
DMs (e.g. you know, this is great, man!). However, these expressions are still also used
Similarly, emoji that are used as DMs can also have a referential counterpart with a
lexical meaning. This is evident for emoji depicting objects or actions. Consider the two uses
(RUEG-DE_0.3.0, DEbi84MR)
(RUEG-DE_0.3.0, DEmo42FD)
In (3), the car emoji does not add to the propositional content of the message but rather is
used as a DM, acting on the discourse level: it visually emphasizes the message contents and
contributes intersubjectively by making the message more engaging. In contrast, the two
occurrences of the same emoji in (4) are not used as DMs but as icons that refer to concrete
objects, namely two cars. As such, they are even syntactically integrated, replacing lexical
items, i.e. nouns that could be used in those slots (cf. Danesi 2017, Dürscheid 2020a, 2020b,
Dainas and Herring forthc.). When understanding emoji as DMs, then, we refer to emoji in
their usage for discourse functions rather than to their possible counterparts with a referential
meaning – just as one refers to verbal elements (e.g. man) as DMs only in their discourse
usage.
marking tone have been characterized as “pragmatic functions” (Dainas and Herring forthc.).
Investigating emoji under the perspective of graphic DMs covers such functions, while
adding a novel empirical domain to the discussion of discourse markers that can shed new
light on associations of different positions (LP/RP) and different discourse functions. This
way, the view of emoji as DMs can integrate and further systematize different emoji
taxonomies and help us understand their dynamic meaning in terms of general patterns of
literature and subsume them under textual, subjective and intersubjective DM functions. Note
that, in each case, the categorization is not exclusive, since different meanings overlap and
emoji can – and often do – serve more than one function (as confirmed by our experimental
conversational openings or closings (e.g. Al Rashdi 2015, McCulloch 2019); organise text,
replacing punctuation marks (Sergeant 2019, Albert 2020, Androutsopoulos 2020b; cf.
Provine et al. 2007 for emoticons); and, interacting with intersubjective functions, emphasize
a topic through visual repetition or illustration of what has already been expressed verbally
(e.g. Al Rashdi 2015, Dürscheid and Siever 2017, Siebenhaar 2018, Dainas and Herring
forthc.) and indicate the type of speech act (e.g. Gawne and McCulloch 2019, Albert 2020;
Subjective functions cover usages where emoji express the speaker’s emotions or their
evaluation of the content (cf. Al Rashdi 2015, Cramer et al. 2016, Na’aman et al. 2017,
Pappert 2017) and generally convey the speaker’s state of mind (cf. Danesi 2017 on
While such functions can be regarded as central for emoji – in particular for those that
depict facial expressions and thus indicate emotions – several studies also point to functions
that can be analyzed as intersubjective. Most notably, emoji have been characterized as
adjusting the tone, e.g. in sarcasm or politeness, thus: serving as cues to guide the addressee’s
2020); making it more engaging to the recipient through “spicing up a text” (Cramer et al.
2016, 4); seeking a reaction from the addressee and keeping the communication channel open
(Danesi 2017 characterizes this as the “phatic” functions of emoji in the sense of Jakobson
1960).
As mentioned above, emoji can serve multiple, often interrelated functions at the same
time. For example, an emoji that indicates the type of speech act at the textual level is at the
same time intersubjectively guiding the addressee’s understanding, for instance, by indicating
the tone of the message. Such multifunctional uses can also bring together intersubjective and
subjective domains. Especially smile emoji can serve such related meanings: their subjective
function to express positive emotions can support an intersubjective pragmatic meaning, such
as indicating positive intent. By doing so, they can downgrade a strong complaint and signal
a friendly attitude towards the addressee (e.g. Dresner and Herring 2013, Gawne and
McCulloch 2019) or mark the speaker as friendly to present a positive social persona. Against
this background, we do not draw a hard line between subjective and intersubjective functions
for individual DMs but assume a continuum, parallel to what Traugott (2010) suggested for
verbal DMs. In particular, our analyses will allow for both functions to be present in one DM,
Approaching emoji from the perspective of co-speech gestures, Gawne and McCulloch
(2019) show that, like gestures and to a much larger extent than verbal language, the meaning
of emoji is shaped by context-specific use. Furthermore, the meaning of emoji can be culture-
dependent (Siever 2015) and can even differ within one cultural context, depending on setting
Taken together, this demonstrates that the analysis of emoji meanings and their
solely be based on interpretations of isolated elements alone but crucially needs to involve
their contextualization (cf. also Dainas and Herring forthc.), which constitutes an additional
In the following sections, we present two studies on emoji that take into account these
findings and approach them from the perspective of graphic DMs to investigate possible
The RUEG corpus (Wiese et al. 2019) has been compiled by the Research Unit “Emerging
Grammars in Language Contact Situations”.2 It consists of naturalistic data elicited with the
“Language Situations” set-up (Wiese 2020). In this set-up, participants are familiarized with a
fictional event, e.g. a traffic accident, through a non-verbal stimulus. They picture themselves
as a witness and describe the incident acting out different communicative situations that can
cover spoken and written modes and different settings. In the specific set-up used in RUEG,
the stimulus was a video of a (minor) car accident. Participants were asked to imagine
themselves as a witness of this accident and to tell different interlocutors about it in different
communicative situations, yielding 2x2, formal vs. informal and spoken vs. written settings.
In the situation relevant for the present study, participants were asked to send a text message
Participants came from two age groups: (i) 15- to 18-year-old adolescents who were
still attending school and (ii) young adults from 22 to 35 years of age. These age groups are
particularly suitable for our investigation, since both started using digital media when
creating and maintaining social relationships had already become a key function of the
internet (McCulloch 2019). This makes digital messenging in these cohorts particularly open
2
FOR 2537; see hu-berlin.de/rueg. Stimuli and other information, including a training video for elicitors, are
freely accessible on: https://osf.io/qhupg/. The study presented in this chapter was part of a larger investigation
into discourse patterns in different communicative situations. For corpora focusing on WhatsApp data alone, cf.
the MoCoDa2 corpus with digital messages from Germany (Beißwenger et al. 2020), and the corpus “What’s
Up Switzerland?” with WhatsApp messages from Switzerland (Stark et al. 2014–2020)
All data is integrated in a multimodal corpus that includes sound files and transcriptions
for spoken data and text for written data that preserves such graphic elements as emoji. The
ANNIS (Krause et al. 2016).3 It is openly available at Zenodo,4 with new versions regularly
In our study, we concentrate on the German data from Germany5 in the corpus version
RUEG-DE_0.3.0, which contains 121,998 tokens from 193 speakers in four communicative
(punctuation marks and filled pauses were excluded from the token count).
For our investigation of emoji as DMs, we excluded all emoji that depicted concrete objects
and replaced lexical items (e.g. instead of ‘car’, as in 4), and included all emoji that fitted
our defining criteria for DM, that is, those that had no propositional content but rather
operated at the level of discourse and were not syntactically integrated. This way, we
identified a total number of 142 emoji tokens, covering 43 different emoji types. A bit more
than a third of the speakers used emoji, namely 67 out of a total of 192 speakers (34.89%). In
our analysis, we therefore focused on those 67 speakers. In their productions, we found 2.2
3
https://korpling.org/annis3/
4
https://zenodo.org/record/3236069#.XXo2x3tCRPY
5
In addition, the corpus contains data from heritage speakers of German in the US. For the purpose of the
present chapter, this data was excluded from the analysis, though.
The six most common emoji types in our data are *Face with Tears of Joy* (16 hits;
0.24 per 100 tokens), *Facepalming* (woman or man) (11 hits; 0.17 per 100 tokens),
*Face with Rolling Eyes* (10 hits; 0.15 per 100 tokens), *Face Screaming in Fear* (10
hits; 0.15 per 100 tokens), *Flushed face* (10 hits; 0.15 per 100 tokens), and *Grinning
Emoji as graphic DMs in our data are exclusively positioned in the right periphery, in line
with findings from the literature that emoji occur most frequently at the end of messages and
rarely at the beginning (e.g. Al Rashdi 2015, Cramer et al. 2016, Dürscheid 2020a). If the
association of RP positions with intersubjective DM functions holds for graphic DMs as well,
this might then be related to intersubjectivity. However, as mentioned above, emoji are highly
context-sensitive and it is therefore important to take the context into account when analyzing
their meaning.
In the corpus, simple smiling or grinning emoji often occur after salutations – as in
(5) Hiii omg wisst ihr was mir grade passiert ist ?
(RUEG-DE_0.3.0, Demo57MD_iwD)
In such cases, we can assume that the emoji signals politeness and friendliness towards the
addressee, rather than simply providing the information that the speaker is happy, and hence
a smile emoji that is used to downgrade a strong complaint (cf. Cramer et al. 2016,
McCulloch 2019, cf. also Dresner and Herring 2013 for emoticons). In order to check for this
function, we classified the content preceding each emoji as: negative, as in (6) and (7);
eines Unfalls
of.an accident
(RUEG-DE.0.3.0, DEbi12FR)
(RUEG-DE.0.3.0, DEmo57FD)
(RUEG-DE.0.3.0, DEmo02MD)
(RUEG-DE_0.3.0, DEbi12FT)
Over two thirds of the emoji in our data appear after negative content (72.21%).6 Looking at
the six most frequent emoji in our corpus, for three of them, namely , , , such
negative messages were the only context in which they were used, while the other three
10 - -
11 - -
10 - -
9 1 -
9 6 1
6 3 1
Table 1: Classification of the preceding content for the most common emoji in RUEG-DE_0.3.0 (absolute
numbers)
Given the findings on intersubjective uses for softening negative contents, these emoji might
hence fulfil intersubjective functions. However, a closer look at our data did not provide such
a clear-cut picture since, in many cases, the interpretation could not be determined
6
Note, though, that negative content is probably somewhat over-represented in our data given the elicitation
stimulus (a video of a car accident).
(10) ein Auto hat schön nen Schaden in den Hintern.
bekommen
gotten
(RUEG-DE_0.3.0, DEbi24FR)
In this example, the emoji might indicate that the speaker is happy or amused or indeed feels
schadenfreude about what happened or it could signal the speaker’s intention to soften the
statement and make clear for the addressee that nothing really bad happened. While the first
mutually exclusive but can both be involved in the same occurrence. Given that emoji used as
DMs typically represent happy, unhappy etc. faces, we can regard subjective functions as
more basic, since they are closer to an iconic indication of the speaker’s emotion.
interactional meanings. The interesting question is then whether the dominant usage of emoji
in the RP is associated with additional intersubjective functions that interact with subjective
ones and build on them. Corpus data can give a first indication for this but it does not allow
us to unambiguously infer such intersubjective functions (cf. also Dürscheid 2020a on this
issue). To test this question, we therefore conducted an experimental study, presented in the
next section.
4 The discourse interpretation of emoji: An experimental study
We conducted a study to test whether emoji do not only serve subjective functions but also
acquire intersubjective functions that are salient in interpretation. To link up with our corpus
findings, we looked at speakers comparable to one of the age groups in the corpus, namely
4.1 Methods
4.1.1 Participants
A total of 44 adults between 21-34 years old were tested. Subjects were recruited by student
4.1.2 Material
Experimental stimuli were twelve short sentences constructed to look like parts of
WhatsApp® messages. Each stimulus sentence was presented with an emoji and each in two
versions: with an emoji at the end and with an emoji at the beginning. Stimulus versions were
mixed, yielding two sets of stimuli, each with an equal number of sentences with an emoji at
the beginning and an emoji at the end, and each stimulus sentence appearing only once in
each set. An example (translated from German) is given in (11), with the different versions of
two stimuli, once from set 1, and once from set 2 (all stimuli are listed in the Appendix, with
(11) a. The women strolled across the car park (from set 1)
emoji, we used sentences with a neutral content loosely related to the narratives in the RUEG
person(s), a motion verb in the perfect tense (the commonly used tense for informal
The emoji employed in this study are frequently used positive emoji of facial
expressions. Based on our corpus data and findings from previous studies (e.g. Gullberg
2016, Ljubešić and Fišer 2016, Annamalai and Salam 2017, Dürscheid 2020a, 2020b), we
followed by three different paraphrases that provided (i) a neutral interpretation, (ii) a
was counterbalanced across stimuli sentences. Paraphrases followed a fixed pattern. The first
part was identical in all paraphrases and picked up the verbal contents of the stimulus
sentence. In the paraphrases for subjective and intersubjective paraphrases (ii) and (iii), this
was followed by a second part expressing, for subjective interpretations, the speaker’s stance
towards the contents and, for intersubjective interpretations, the speaker’s desire to convey a
certain social persona. Hence, paraphrases were constructed according to the pattern in Table
2.
paraphrases (see the Appendix for the German original), where (12a) provides a subjective
a. Your friend wants to tell you that the women strolled across the car park and
likes it.
b. Your friend wants to tell you that the women strolled across the car park and
c. Your friend wants to tell you that the women strolled across the car park.
This set-up allows us to systematically tap into subjective vs. intersubjective discourse
meanings through a number of features that distinguish it from previous studies on emoji
meaning where participants were asked to describe the intended meaning of emoji (e.g.
Cramer et al. 2016) or to choose from a list of descriptions (e.g. Dainas and Herring forthc.).
In contrast, our set-up does not isolate the emoji interpretation but integrates it into that
of the context sentence. This does not only systematically take into account the context-
on the propositional semantics of the message in its scope. Since this semantics is part of all
paraphrases, it also allowed us to include a neutral paraphrase as part of the choices, which
we considered relevant in view of findings for emoticons suggesting that these sometimes do
not have an effect on the interpretation (Walther and D’Addario 2001). Another key feature
of our set-up is that it targets two general types of DM meaning, namely subjective and
intersubjective functions, for emoji, rather than more specific meaning aspects. The main
associations of pragmatic markers in general and, vice versa, benefit from it for our
understanding of emoji. This led to a smaller and more streamlined set of options, which had
the further advantage of making it possible to ask participants to rank all options, rather than
to choose only one. This allowed us to take into account the multifunctionality of DMs,
which has also been observed for emoji (see section 2).
4.1.3 Procedure
Participants were invited via email to take part in an online study. The study was conducted
using SurveyMonkey. On that site, we provided information on the study and the terms and
conditions for informed consent. After giving their consent, participants got the following
“In what follows, you will see short sentences, one at a time. Imagine that a friend
has written this sentence to you via WhatsApp®. There are three explanations under each
Participants were divided into two groups, with each group seeing one of the two stimulus
lists (see Section 4.1.2). Stimulus sentence, together with their respective paraphrases, were
presented one by one. After ranking paraphrases for all stimulus sentences, participants filled
out a short questionnaire asking their age, gender, languages spoken in the family in
4.2 Results
One item was excluded from the analysis because participants commented afterwards that
one of the paraphrases did not fit the contents. A post hoc poll with ten additional speakers
confirmed this. Another item was excluded because the emoji was accidently positioned at
the beginning of the sentence in both lists. Hence, we included in the analysis two sets of ten
stimuli (five stimuli with emoji at the beginning, five stimuli with emoji at the end) ranked by
subjective 256 94 81
and type of paraphrase (X2 (4, N = 440) = 285.4, p < 2.2e-16). The patterns indicate three
main results:
(i) The neutral paraphrases, without a discourse-pragmatic content, were evaluated as least
fitting, receiving rank 3 most often both across paraphrases and across rankings for
neutral paraphrases. This indicates that the emoji in our stimuli contribute a discourse
(ii) Paraphrases that involved an additional subjective discourse meaning received rank 1
most often, both across paraphrases and across rankings for subjective paraphrases.
This confirms our suggestion that subjective meanings are a central part of emoji used
as DMs.
(iii) Paraphrases that involved additional intersubjective discourse meanings received rank 2
most often, followed by rank 1, in both cases across paraphrases and across rankings
Hence, even though emoji depicting facial expressions point to the emotions that these
expressions reflect and thus to subjective meanings, these findings confirm that they are also
associated with intersubjective functions. Our results further confirm that emoji meanings do
not exclude each other but they can be multi-faceted, indicating that the two functions
identified here build on each other, with intersubjective meanings based on subjective
primary RP association that was evident from our corpus study and is also confirmed in the
literature (e.g. Cramer et al. 2016, Dainas and Herring forthc.; see 3.3). To further investigate
this, we looked at the results for the two stimuli variants separately, comparing rankings for
stimulus sentences that were preceded by an emoji (LP) with those for the same sentences
followed by an emoji (RP). In each case, we calculated the relative rankings for subjective vs.
intersubjective ratings. We counted how often participants rated the subjective paraphrase
higher than the intersubjective one (i.e. we counted each time that the subjective paraphrase
was ranked 1 and the intersubjective paraphrase was ranked 2 or 3 or that the subjective
paraphrase was ranked 2 and the intersubjective paraphrase was ranked 3). And vice versa,
we counted how often participants rated the intersubjective paraphrase higher than the
subjective one. Figure 1 shows the respective figures for all stimulus sentences (across the
Figure 1: Distribution of relative rankings over the two stimuli variants (emoji in the left vs. right periphery)
160
146
139
140
120
100
80
80 72
60
40
20
0
emoji left emoji right
As the figures show, there was an advantage for the subjective reading in both conditions, but
this became smaller when the emoji was on the right, with the subjective reading less often in
the lead and the intersubjective reading more often in the lead than when the emoji was on
the left. However, this difference is not large and it did not reach significance in a Chi-square
test, which might also be due to the comparably small number of stimulus items and of
salience for right-peripheral graphic DMs but this will need further corroborations through
larger studies.
interesting specific patterns for different subgroups of emoji, suggesting a continuum along
different dominance relations between subjective or intersubjective meanings, which interact
with LP and RP positions. In our data, this continuum is characterized by three patterns:
The first pattern is attested for , , . For stimuli with these emoji, the overwhelming
majority of participants ranked the subjective paraphrase higher than the intersubjective one
for both LP and RP occurrences. The second pattern is displayed by four emoji in our data:
, , , . For this group, the subjective paraphrase received a higher rank when the
emoji was in the LP but not when it was in the RP. The third pattern is represented by one
emoji in our study, . For this stimulus, the intersubjective paraphrase received a higher
rank than the subjective one when the emoji was in the RP, while relative ranks were about
the same when it was in the LP. Table 4 gives an overview of the frequencies within the
different patterns.
Pattern Emoji position Emoji Subjective >intersubjective Intersubjective >subjective
(i) LP 21 1
17 5
22 0
RP 22 0
21 1
21 0
(ii) LP 15 7
15 7
13 9
12 10
RP 10 11
11 10
10 12
10 12
(iii) LP 10 11
RP 8 14
Table 4: Frequency of higher relative ranks for subjective vs. intersubjective paraphrases (out of 22 responses)
While the numbers for each group are too small to allow a definite conclusion, these findings
suggest that the distribution of emoji over different patterns is related to their form. Emoji
that are more complex and depict not just smiles but include more active elements, pointing
to activities such as laughing, tend to display pattern (i). Since facial emoji are understood to
represent the speaker, such active elements anchor the interpretation more strongly on the
speaker’s side, which then favor subjective discourse interpretations. In contrast, for emoji
that depict smiling faces with no such additional active elements, as in group (ii), the
intersubjective aspect is strong enough to compete with the subjective one for a higher rank.
The association of linear (LP/RP) and functional aspects (subjective/intersubjective) can then
intersubjectification for these emoji related to their position in the RP, in line with what
functions to a different degree, depending not only on the context they are presented in but
also on the type of pictorial elements they combine. Further research is needed to investigate
5 Conclusions
In this chapter, we presented an investigation into emoji as graphic discourse markers (DMs).
We showed that emoji are typically used in a way that meets key criteria for DMs: they do
not contribute to the propositional meaning and truth value of a message but rather operate on
the level of discourse organization and they are not a syntactically integrated part of
sentences but occupy external positions at the peripheries. We showed that specific emoji
meanings identified in the literature can be subsumed under subjective, intersubjective and
textual discourse functions and argued that emoji share functional and positional properties
with verbal DMs. Further parallels to verbal DMs include: the existence of referential
counterparts for some emoji, especially (but not only) those depicting concrete objects; a
strong reliance on context for their meaning contribution; and, related to this, the possibility
As part of a comparably novel and vibrant area, namely that of digital social
further light on the pragmatic contribution of emoji on the one hand and on associations
between intersubjective and subjective discourse meanings of DMs with left periphery (LP)
vs. right periphery (RP) on the other hand, which have been an ongoing topic in the
We presented two studies investigating this for the example of emoji depicting facial
study testing subjective and intersubjective interpretations for sentences with emoji in their
left vs. right periphery. The corpus study was based on the open-access RUEG corpus that
provides naturalistic language productions by speakers from two age groups, adolescents and
registers.
Given that a lot of emoji used as DMs represent happy, unhappy etc. faces that indicate
emotions and thus relate to the speaker, we regarded subjective functions as more basic.
However, since intersubjective and subjective functions do not exclude each other but rather
can be realized by the same element, we hypothesized that the emoji in our corpus might
involve additional intersubjective functions that interact with these subjective ones and build
on them. Such intersubjective functions were plausible for the emoji in our data and they
were further supported by their contexts: emoji depicting facial expressions appeared
exclusively in the RP and they appeared mostly after negative contents, that is, in a position
emoji do not only serve subjective functions but also acquire intersubjective functions that
are salient in interpretation. Participants saw sentences with emoji in the RP or LP and had to
indicate, through ranking, how suitable they found paraphrases expressing only the
sentence’s content (neutral), or additional subjective discourse meanings (speaker expressing
Our study yielded three main results: (i) neutral paraphrases were evaluated as least
fitting, indicating that the emoji in our stimuli contribute a discourse meaning, which further
confirms their status as (graphic) DMs; (ii) subjective paraphrases received rank 1 most often,
confirming our assumption that subjective meanings are a central part of emoji used as DMs;
(iii) intersubjective paraphrases received rank 2 most often, followed by rank 1, indicating the
While there was an overall advantage for subjective paraphrases, this advantage was
smaller for stimuli with RP emoji compared to those with LP emoji, which points to an effect
individual emoji revealed three patterns that could be located on a continuum along different
dominance relations between subjective or intersubjective meanings, which interact with LP-
and RP-positions. Our findings suggest that subjective meanings are particularly strong for
emoji that include more active elements, e.g. laughing. As an explanation for this, we
suggested that such active elements anchor the interpretation more strongly on the speaker’s
side, thus making subjective discourse meanings more salient. In contrast, for emoji that
depict smiling faces with no additional active elements, the subjective aspect was comparably
less strong and thus their interpretation was open to intersubjective aspects as well. Our
findings suggest that this can be modulated by their position in the RP (vs. LP), indicating
Taken together, such results further contribute to the complex picture of associations
between linear and functional aspects of DMs. A key finding in our study was the coexistence
and interaction of different discourse meanings within the same DM, and their interaction
with left vs. right positions. By understanding emoji as a special, graphic case of DMs, our
results, on the one hand, provide novel evidence supporting findings on left vs. right
associations and intersubjective vs. subjective discourse meanings. On the other hand, they
contribute to the discussion of emoji as vibrant and quickly developing elements of written
intersubjective and textual discourse meanings. From an empirical and methodological point
of view, we believe our study has shown that informal-written language as captured by
linguistic dynamics makes it valuable not only for research into computer- or digitally-
mediated communication but also for lexical, grammatical, and discourse-pragmatic analyses.
Acknowledgments
The work presented here was conducted within the Research Unit FOR 2537, projects P6
German Research Foundation). For their support in collecting data for the emoji study, we
thank Guendalina Reul, Luisa Koch, Myrto Rompaki, Yeşim Bayram, Tjona Sommer, Lea
Coy, Marvin Brink and Marius Keller. We thank Oliver Bunk for valuable input on the two
studies. For helpful suggestions, we thank the editors of the present volume, Jolanta
Šinkūnienė and Daniël Van Olmen, and two anonymous reviewers. For the discussion of
different aspects of the work presented here, we thank the participants of the workshop
“Pragmatic Markers and Clause Peripheries”, organized by Jolanta Šinkūnienė and Daniël
Van Olmen as part of the 52nd Annual Meeting of the Societas Linguistica Europaea,
Leipzig, August 21st to 23rd, 2019, and of the workshop “Empirical Studies of Word Sense
Dominik Schlechtweg as part of the Annual Meeting of the German Society for Linguistics
References
Al Rashdi, Fathiya. 2015. Forms and Functions of Emojis in WhatsApp Interaction among
Albert, Georg. 2020. “Emojis und soziale Registrierung (enregisterment).” In Register des
Gruyter.
Register des Graphischen, ed. by Jannis Androutsopoulos, and Florian Busch, 133–158.
Berlin: De Gruyter.
vielfältig: IDS-Jahrbuch 2019, ed. by Konstanze Marx, Henning Lobin, and Axel
Androutsopoulos, Jannis, and Florian Busch (eds). 2020. Register des Graphischen:
Beeching, Kate, and Ulrich Detges. 2014. “Introduction.” In Discourse Functions at the Left
Change, ed. by Kate Beeching, and Ulrich Detges, 1–23. Leiden: Brill.
Beißwenger, Michael, Marcel Fladrich, Wolfgang Imo, and Evelyn Ziegler 2020. “Die
Beißwenger, Michael, and Steffen Pappert. 2019. Handeln mit Emojis: Grundriss einer
Universitätsverlag Rhein-Ruhr.
Blakemore, Diane. 2004. Relevance and Linguistic Meaning: The Semantics and Pragmatics
Gesprächsforschung.
Cramer, Henriette, Paloma De Juan, and Joel R. Tetreault. 2016. “Sender-intended Functions
Human Computer Interaction with Mobile Devices and Services, MobileHCI: 504–509.
Dainas, Ashley R., and Susan C. Herring. forthc. “Interpreting Emoji Pragmatics.” In
Internet Pragmatics: Theory and Practice, ed. by Chaoqun Xie, Francisco Yus, and
Danesi, Marcel. 2017. The Semiotics of Emoji: The Rise of Visual Language in the Age of the
Degand, Liesbeth, and Benjamin Fagard. 2011. “Alors between Discourse and Grammar.”
Detges, Ulrich, and Richard Waltereit. 2014. "Moi je ne Sais pas vs. Je ne Sais pas moi:
French Disjoint Pronouns in the Left vs. Right Periphery." In Discourse Functions at the
Left and Right Periphery: Crosslinguistic Investigations of Language Use and Language
Change, ed. by Kate Beeching, and Ulrich Detges, 24–46. Leiden: Brill.
Dresner, Eli, and Susan C. Herring. 2013. “Emoticons and Illocutionary Force.” In
Gruyter.
Dürscheid, Christa, and Christina M. Siever. 2017. “Jenseits des Alphabets: Kommunikation
Frick, Karina. 2017. Elliptische Strukturen in SMS: eine korpusbasierte Untersuchung des
Gawne, Lauren, and Gretchen McCulloch. 2019. “Emoji as Digital Gestures.” Language@
Internet 17 (2).
Glikson, Ella, Arik Cheshin, and Gerben A. van Kleef. 2017. “The Dark Side of a Smiley:
Gullberg, Kajsa. 2016. Laughing Face with Tears of Joy: A Study of the Production and
Jakobson, Roman. 1960. “Linguistics and Poetics.” In Style and Language, ed. by Thomas A.
Krause, Thomas, Ulf Leser, and Anke Lüdeling. 2016. “graphANNIS: A Fast Query Engine
for Deeply Annotated Linguistic Corpora.” Corpus Linguistic Software Tools 31 (1): 1–
25.
Lamontagne, Jeffrey, and Gretchen McCulloch. 2017. “Wayyy Longgg: Orthotactics and
Ljubešić, Nikola, and Darja Fišer. 2016. “A Global Analysis of Emoji Usage.” Proceedings
McCulloch, Gretchen. 2019. Because Internet: Understanding the New Rules of Language.
Müller, Simone. 2005. Discourse Markers in Native and Non-native English Discourse.
Amsterdam: Benjamins.
Workshop: 136–141.
Pappert, Steffen. 2017. “Zu kommunikativen Funktionen von Emojis in der WhatsApp-
Provine, Robert, Robert Spencer, and Darcy Mandell. 2007. “Emotional Expression Online:
Sergeant, Philip. 2019. The Emoji Revolution: How Technology is Shaping the Future of
Siebenhaar, Beat. 2018. “Funktionen von Emojis und Altersabhängigkeit ihres Gebrauchs in
Stark, Elisabeth, Simone Ueberwasser, and Anne Göhring. 2014–2020. Corpus “What’s up,
Mediated Communication, ed. by Susan C. Herring, Dieter Stein, and Tuija Virtanen,
Traugott, Elizabeth Closs, 1982. “From Propositional to Textual and Expressive Meanings:
Amsterdam: Benjamins.
Traugott, Elizabeth Closs. 2012. “Intersubjectification and Clause Periphery.” English Text
https://www.youtube.com/watch?v=FrkKMMWGfA0&feature=youtu.be .
Ueberwasser, Simone, and Elisabeth Stark. 2017. “What’s up, Switzerland? A Corpus-based
Walther, Joseph, and Kyle D’Addario. 2001. “The Impacts of Emoticons on Message
19 (3): 324–347.
Wiese, Heike. 2020. “Language Situations: A Method for Capturing Variation Within
Wiese, Heike, Artemis Alexiadou, Shanley Allen, Oliver Bunk, Natalia Gagarina, Kateryna
Iefremenko, Esther Jahns, Martin Klotz, Thomas Krause, Annika Labrenz, Anke
Lüdeling, Maria Martynova, Katrin Neuhaus, Tatiana Pashkova, Vicky Rizou, Christoph
Schroeder, Luka Szucsich, Rosemarie Tracy, Wintai Tsehaye, Sabine Zerbian, and Yulia
Zuban. 2019. RUEG Corpus (Version 0.3.0) [Data set]. Zenodo. DOI:
10.5281/zenodo.3236069
Appendix
Deine Freundin möchte dir mitteilen, dass die Frau über den Parkplatz geschlendert ist, und
Deine Freundin möchte dir mitteilen, dass die Frau über den Parkplatz geschlendert ist, und
Deine Freundin möchte dir mitteilen, dass die Frau über den Parkplatz geschlendert ist.
Deine Freundin möchte dir von dem Jungen mit dem Ball erzählen und dabei humorvoll
wirken.
Deine Freundin möchte dir von dem Jungen mit dem Ball erzählen.
Deine Freundin möchte dir von dem Jungen mit dem Ball erzählen und findet es sehr lustig,
Deine Freundin möchte dir berichten, in welche Richtung die Familie gegangen ist, und freut
sich darüber.
Deine Freundin möchte dir berichten, in welche Richtung die Familie gegangen ist, und
Deine Freundin erzählt dir von dem Mann, der an dem Hochhaus entlang ging.
Deine Freundin erzählt dir von dem Mann, der an dem Hochhaus entlang ging, und möchte
fröhlich wirken.
Deine Freundin erzählt dir von dem Mann, der an dem Hochhaus entlang ging, und findet das
Gebäude amüsant.
Deine Freundin möchte dir von dem Sprint des Postboten erzählen und findet sehr komisch,
Deine Freundin möchte dir von dem Sprint des Postboten erzählen.
Deine Freundin möchte dir von dem Sprint des Postboten erzählen und möchte witzig wirken.
Deine Freundin möchte dir von den Kindern berichten und findet sie süß.
Deine Freundin möchte dir von den Kindern berichten und möchte herzlich wirken.
Deine Freundin möchte dir von der Polizistin berichten und ist gut gelaunt.
Deine Freundin möchte dir erzählen, wie der Typ die Bordsteinkante überwunden hat.
Deine Freundin möchte dir erzählen, wie der Typ die Bordsteinkante überwunden hat, und ist
Deine Freundin möchte dir erzählen, wie der Typ die Bordsteinkante überwunden hat, und
Deine Freundin möchte dir von den langsamen Rentnern berichten und möchte humorvoll
wirken.
Deine Freundin möchte dir von den langsamen Rentnern berichten und findet die lustig.
Deine Freundin möchte dir von dem Verkäufer erzählen und amüsiert sich darüber.
Deine Freundin möchte dir von dem Verkäufer erzählen und möchte lässig wirken.