Communicating Sarcasm

As social media becomes more and more prominent in our society, we have become more accustomed to digital media to express ourselves and communicate with others. However, this raises questions about our ability to signal different sentiments on these media. In particular, an issue that often arises is the question of communicating sarcasm over these media. Sarcasm is a disruptive factor that polarizes the meaning of a message and flips it completely. Before the digital era, studies often analyzed the verbal cues used to signal sarcasm. Signalling sarcasm is not a perfect science, however, studies have given us insight into how verbal cues, such as intonation, are used to signal sarcasm. Yet, in a digital medium the dimension of tonality is
removed and it seems as if there may be far more opportunity for misinterpretation of sarcastic messages in digital media than there is in verbal communication. This leads me to ask, can we signal sarcasm on digital media? Specifically, I will look at sarcasm signalling on Twitter because of its unique feature, the “hashtag”, which is a searchable annotation. Hashtags have been used to study the way sarcasm is signalled on Twitter because it is the only way to know the writer’s intention without assuming sarcastic intent. This allows researchers to analyze the lexical patterns in tweets to identify if there is a stereotypic semantic structure that signals sarcasm. Furthermore, studies have also looked into the types of lexical cues used in the signalling of sarcasm, such as, interjections and hyperbole. I will analyze the tools used to signal sarcasm verbally, lexically, and through hashtags on Twitter to determine whether sarcasm can effectively be signalled on Twitter.

To know if sarcasm can be signalled on Twitter, we must first understand how sarcasm is signalled verbally. Much of the work done around sarcasm detection focuses on early development and the age at which children begin to recognize sarcasm and how they do this. This is significant because it allows to identify the fundamental elements of how sarcasm is transmitted. The two most commonly discussed elements of sarcasm are the context in which the utterance is delivered and the intonation with which it is delivered. There is no academic consensus as to which one is more important to understanding sarcasm, but there are many studies that indicate sarcastic intonation is more likely to signal sarcasm to young children than context in early development.

One such study by Virginie Laval and Alain Bert-Erboul examined the way French children aged 3-7 responded to sarcastic requests given either context or sarcastic intonation. The children were told a story with either a sarcastic or non-sarcastic context, and which ended with a request, either given in a sarcastic or neutral intonation. They were then given three options: one that complied with the request (literal interpretation), one with the opposite action to the request (sarcastic interpretation), and a final option which was situational, meaning that the subject did not understand the request, either sarcastic or not. Intonation is “the combination of tonal features into larger structural units, associated with the acoustic parameter of voice fundamental frequency and its distinctive variations in the speech process” ((Laval and Bert-Erboul 2005). The results of this study found that children aged 5 and over were able to recognize sarcasm but only under the condition of sarcastic intonation. By the time children reached the age of 7, their understanding of sarcasm was good regardless of the situation. This supports the hypothesis that intonation comes first in the comprehension of sarcasm and thus is more fundamental to the way we signal sarcasm than context (Laval and Bert-Erboul 2005).

Another study done by Carol A.
Capelli, Noreen Nakagawa and Cary M. Madden also supports this hypothesis. The authors conducted experiments with 32 third grade students, 32 sixth grade students and 16 volunteer undergraduate students who volunteered to participate as a comparison group. The study performed a similar experiment as the previous authors, they told the children a story under different conditions of sarcastic intonation and context and then asked for the children’s interpretation of the story. According to the authors a sarcastic intonation is “a mocking tone, in which the speaker greatly exaggerated the modulation of pitch and drew out the emphasized syllables longer than was the case for the neutral intonation. The experiment used four questions to test the comprehension of the children. They used an open-ended question about the intention of the speaker, a question about how the children had interpreted the relevant context information in terms of its consistency with the ending remark, an either/or question about whether the speaker was trying to convey the literal meaning of their utterance, and the fourth question was a follow up to the third question and asked the children to explain why they gave the answer they did. The results of the study indicate that early ability to interpret sarcasm may depend more heavily on intonation than context. For all three groups (third grade, sixth grade, and adult subjects) the number of sarcastic interpretations were
highest when given both context and intonation. However, between just intonation and just context, the number of sarcastic interpretations were higher for intonation only amongst the children but lower in the adult group. The results support the hypothesis that intonation is more fundamental to signalling sarcasm (Capelli et al 1990).

One might argue that all of these studies are limited in their approach to signalling sarcasm by assuming that context and intonation are two separate routes by which we signal sarcasm.
The evidence supports the idea that both intonation and context contribute to the comprehension of sarcasm, which is indicative of the holistic nature of sarcasm. However, based on their findings, it is fair to say that intonation is an important element in signalling
sarcasm.

If we accept that intonation is fundamental to signalling sarcasm, what happens when sarcasm is removed from the oral medium and put into writing? The field of linguistics may have the answer. There is research that indicates lexical cues such as hyperbole, interjections, extreme adjectives or adverbs, and affective evaluations are all types of signals used to indicate sarcasm. Studying these cues in a verbal context can be difficult because it is impossible to ascertain the intent of the speaker. Twitter, on the other hand, provides an excellent alternative medium to study these claims because hashtags are often used to make explicit sarcastic intent. An interesting connection to the previous studies mentioned in this paper, tweets are often independent statements that do not have a context associated with them. This lends support to the idea that context is of less importance than the signalling mechanisms (e.g. intonation, lexical cues, and hashtags) used to express sarcasm. In order to evaluate the role of lexical cues in tweets marked with the hashtag “sarcasm”, we must first examine how the hashtag is used in relation to sarcastic tweets.

Twitter is a popular micro-blogging platform, notably, messages (commonly referred to as “tweets”) can contain searchable markers called “hashtags”. These markers are commonly used to link messages to broader topics or to express certain sentiments, such as sarcasm. One study done by David Kovaz, Roger J. Kreuz, and Monica A. Riordan examined tweets marked with the hashtag “sarcasm” and examined the use of various lexical cues to determine whether it is possible to signal sarcasm solely using these lexical cues. The researchers collected tweets marked with the hashtag “sarcasm” and filtered the results based on certain criteria (e.g. no retweets) and then collected a control group with tweets from the same users but not marked with the hashtag. The sarcastic tweets were then compared to the control tweets to detect the presence of interjections, adjectives, and adverbs; and specific co-occurrences of parts of speech (i.e., “adverb + adjective” and “adjective + adjective”). The findings of the study support the hypothesis that there are certain lexical cues that help to signal sarcasm. In particular, interjections were found to be more frequent in sarcastic tweets (Kovaz et al 2013).

Another study done by Florian Kunneman, Christine Liebrecht, Margot van Mulken, and Antal van den Bosch tested a similar theory. The group collected a training sample of about 406 thousand Dutch tweets with the hashtag “sarcasm”. This sample was then used to train a classifier to apply to a daily stream of Dutch tweets, which is about 2.25 million tweets. On this particular day, there were 335 explicitly marked sarcastic tweets. The classifier was able to detect 309 with the hashtag removed, for an 87% accuracy. For tweets without the explicit hashtag, the group compiled the top 250 tweets ranked by the classifier to likely be sarcastic and annotated them. 35% of these top 250 tweets were indeed sarcastic. This shows that it is far more difficult to detect sarcasm in an open setting, where an explicit hashtag is not used (Kunneman et al 2015). This supports the idea that the hashtag fills a signalling gap that pure lexical cues cannot achieve.
Ultimately, the aim of both of these studies was to see if it would be possible to detect sarcasm in tweets without the hashtag “sarcasm”. There seems to be something counterintuitive about labelling sarcasm explicitly. By using the explicit hashtag there is no opportunity for inference by the reader. The message is not to be taken literally; it is sarcastic. There is a level of inferability associated with the use of sarcasm. Inferability refers to the degree of likelihood that someone will be able to understand the sarcastic nature of an utterance. If an author believes they need to make their sarcastic intentions explicit by using a hashtag, this may be indicative of low inferability of the message. Inferability is one of many factors referred to as pragmatic factors. These are the social, contextual, and interpersonal elements of sarcasm. According to various studies, people take into account the amount of shared common ground and perceived closeness to the listener when deciding whether to use sarcasm. By using the hashtag, the inferability of the message is assured but the value of the pragmatic factors is diminished. Again, it seems as if sarcasm has a holistic nature to it which can be lost partially when it is moved into a digital medium. Holism is a concept that is explained by Sven Ouzman in “Seeing is Deceiving”, holism is the multi-sensory aspects that are vital to an experience which is encompassed in a mindscape. The value of the mindscape is diminished when one aspect is taken away (Ouzman 2001).

Some authors have hypothesized that explicit markers, such as hashtags, are the digital extra-linguistic equivalent of non-verbal expressions that people use in verbal interactions when conveying sarcasm. This is reminiscent of the secondary orality discussed in Walter J. Ong’s “Orality and Literacy” because it calls on the idea of the revival of a signalling tool used in orality but applying it to a new medium. However, as Ong explains in his work, this secondary orality similar but also strikingly different from primary orality. The idea that the hashtag is nothing more than a replacement for the expressions used verbal communication of sarcasm vastly oversimplifies the complex and multi-sensory dimensions of sarcasm and the value that the hashtag adds to the message. Using the hashtag adds an explicit factor to the signalling mechanism that has never been present in verbal cues nor lexical cues alike. This is an example of the way digital media have adapted to signal sarcasm. Yet, there is controversy about the adverse affects of including an explicit hashtag to the value of sarcasm as a disruptive force. This, accompanied by the idea that sarcasm is a holistic experience that requires one to use multiple senses to comprehend, leads to a very complex understanding of sarcasm. Signalling sarcasm is no easy feat, from verbal, to non-verbal, to lexical cues, human communication patterns have adapted to try and convey the complexities of sarcasm. Similarly, today it seems like Twitter users have adapted the hashtag as a way to signal sarcasm on the digital medium. However, this approach is not yet fully developed, and while the goal is not to replicate orality, signalling sarcasm in digital media, such as Twitter, still needs refinement. There is no golden formula to signalling sarcasm. Sarcasm is more of an art than a science, and the Twitter-verse is still learning and developing its techniques.