FOOLPROOF CLUES TO SARCASM
Acoustic cues of sarcastic speech and the perceiving of sarcasm.LEONIE VAN DER LAND S2254751
Master thesis European Linguistics Faculty of Arts
University of Groningen
Supervisors: dr. D. G. Gilbers & dr. W. Kehrein Academic year: 2018-2019
Index
Abstract 3
1 Background 3-10
1.1 What is sarcasm? 3-7 1.2 Perception of sarcasm 7-8 1.3 Phonetic changes in sarcastic speech 9-10 1.4 Research questions and hypotheses 10
2 Method 11-13
2.1 Production 11-12
2.1.1 Participants 11 2.1.2 Design (procedure and materials) 11-12
2.2 Perception 12-13
2.2.1 Participants 12 2.2.2 Design (procedure and materials) 12-13
3 Results 13-18 3.1 Production 13-15 3.1.1 Descriptive results 13 3.1.2 Results production 14-15 3.2 Perception 15-18 3.2.1 Descriptive results 15-17 3.2.2 Results perception 17 3.2.3 Perception of sarcasm in sexes 17-18
4 Discussion 19-22 4.1 Production 19-20 4.1.1 Pitch 19 4.1.2 Pitch variation 19-20 4.1.3 Tempo 20 4.2 Perception 20-21 4.3 Perception in sexes 21-22 5 Conclusion 22-23 5.1 Overall conclusion 22 5.2 Limitations and recommendations 23
Appendix 24-25
Abstract
Sarcasm can be hard to detect, especially when one’s conversational partner is unfamiliar. If sarcastic speech has specific acoustic cues that are different from sincere speech, this could help the listener know if their collocutor is being sincere or sarcastic. These acoustic cues are, for example, pitch, variation in pitch, and tempo. In the present study, nine actors (two males, seven females) recorded sincere and sarcastic sentences. Analyses revealed that pitch and pitch variation do not differ significantly in sincere and sarcastic speech, but tempo (in syllables per minute) is slower in sarcastic speech. For males, there are no significant differences, but for females pitch and tempo are significantly lower in sarcastic speech. By using the recorded sentences, perception of sarcasm was also researched. In 83,2% of the cases, participants perceived the sentences correctly (87,2% sincere sentences and 79,2% sarcastic sentences) when they listened to either a sincere or a sarcastic sentence and had to score these sentences as such. When participants listened to both a sincere and sarcastic sentence and had to score which one was sarcastic, participants were correct in 92,6% of the cases.
Keywords: sarcasm, sarcastic speech, sincere speech, phonetics, features, cues, perception
1. Background
1.1 What is sarcasm?
When talking to someone you are not familiar with, you may find yourself wondering at one point if that person is being serious or ironic. To help listeners know when sarcasm is intended, specific acoustic cues are used, collectively known as the ‘ironic tone of voice’ (Bryant & Fox Tree, 2005; Clark and Gerrig, 1984; Sperber and Wilson, 1981). Bryant and Fox Tree (2005) define the ironic tone of voice (also referred to as sarcastic intonation) as a “particular consistent prosodic […] pattern with a distinct perceptual correlate that is systematically associated with verbal irony” (pp. 257). This prosodic pattern consists of pitch, volume, tempo, et cetera.
Gibbs (1994) defines two different types of irony: verbal and situational. Verbal irony is, not surprisingly, irony expressed in speech. An example of verbal irony would be “The paper you wrote on common grammatic errors is full of grammatic errors”. Situational irony derives from a situation, for instance when the CEO of an elevator company gets killed in an accident involving an elevator.
Sarcasm is one of the five different subtypes of verbal irony, along with jocularity, understatements, rhetorical questions, and hyperbole (Gibbs, 2000). Many theories about irony
exist and they all focus on different use of ironic language, such as linguistic, cognitive and social aspects. Wilson and Sperber (1992) and Sperber and Wilson (1995), for instance, theorize that irony is speakers’ repetition (echoing) of a previous statement, that is conveyed as ironic through its context. Other researchers (Clark 1996; Clark & Gerrig, 1984) view irony as pretense, not as echoic mention. In this view ironic speakers impersonate and, in some cases, pretend to talk to someone that is not the actual listener. Kreuz and Glucksberg (1989) state that irony is mostly a reminder of the speaker of a former event, but these reminders are not literal repetitions of previous utterances. A theory that combines both the echo and pretense theory, is the allusional pretense theory (Kumon-Nakamura, Glucksberg, & Brown, 1995). This theory proposes that irony is used in situations of failed expectations, where speakers violate the maxim that requires speakers to always be sincere. An alternative theory proposes that ironic utterances are special in that they conceal the negativity that causes the ironic comment (Dews, Kaplan, & Winner, 1995). Colston (1997), disagreeing, suggests that irony oftentimes can be more critical than literal statements.
As previously stated, sarcasm is only one of the five different subtypes of verbal irony (Clark and Gerrig, 1984), but most studies on ironic discourse focus on sarcasm, as they view this subtype to be the most typical one. The remaining four subtypes are, however, also very common (Gibbs, 2000). Jocularity, explains Gibbs (2000), is teasing one’s child, parent, friend, or romantic partner (i.e. a familiar person), and offers a gentler way of pointing out someone’s anomaly that deviates from the social norm. Understatements, like other subtypes of verbal irony, create a contrast with an event where, in the case of understatements, a speaker’s statement minimizes the event. For instance, when it is pouring with rain, an understatement would be “It is raining a little”. Understatements are mostly used to refer to a past event (“It was raining a little”), and to mock a person, event, or object rather than being critical (Colston & O'Brien, 2000). Outside the context of verbal irony, a rhetorical question is a figure of speech, where an answer to the question is not expected, for example “Why is he so stupid?” (Burton, 2007). Within the context of verbal irony, rhetorical questions are not usually an echo, but more often a speaker adopting pretense. In his study, Gibbs (2000) presents an example: a girl has guests over whom she clearly does not like and later makes some sarcastic comments about to her friend (“I would just love to have them here more often”). Right after, she asks “Isn’t it so nice to have guests here?” but is clearly not expecting an answer. Finally, hyperbole is an exaggeration of an actual situation. The girl who had guests over in Gibbs’ (2000) study later stated that “[she] was like the happiest person on earth” when the guests were present.
The present study will focus on sarcasm. By using sarcasm, speakers can express negative and critical attitudes towards events or persons by using positive expressions (Kreuz & Glucksberg, 1989). For example, if a colleague is always late for work and arrives late once more, colleagues waiting might have a look at the time and say “Good to see you arriving on time as always”. The colleagues express their critique by giving a positive phrased remark, yet the context and their tone of speech reveal their true intention.
Sarcasm can be divided into four types: propositional, embedded, like-prefixed, and illocutionary sarcasm (Camp, 2012). First, propositional sarcasm occurs in situations where a statement seems to be a proposition, but an implicit sentiment is actually involved. An example of propositional sarcasm would be “That’s a great idea!”. Embedded sarcasm involves an embedded incongruity in the words and phrase itself. Zvolenszky (2012) exemplifies this type of sarcasm by the sentence “John has turned out to be such a diplomat that no one takes him seriously”. The word “diplomat” is incongruous with “not taken seriously”, as a diplomat is generally taken seriously. The third type of sarcasm, like-prefixed, is a so-called like-phrase that implies rejection of an argument that is made by someone’s conversation partner. A common sarcastic remark that is like-prefixed is “like you care!”. This type of sarcasm is, of course, specific for the English language. A possible Dutch equivalent of the like-prefix could be “alsof” in the sentence “alsof het jou wat kan schelen” (as if you care). The fourth and final type of sarcasm, illocutionary sarcasm, is not only verbal, but also includes non-verbal clues that indicate to a sarcastic intention rather than a sincere one. For instance, when saying “yeah right” at the same time as rolling your eyes, hints to a sarcastic attitude (Zvolenszky. 2012).
Inconsistency in language (discrepancy in what is said and what is meant) occurs not only in sarcasm, but in deception, metaphors, and humor as well. It seems that deception is closely related to sarcasm (Lee & Katz, 1998). When A says to B “I love your new hat”, A could be telling the truth (a literal proposition), be lying (deception), or being sarcastic (sarcasm). The difference between the truth and a lie is dependent on the intention of the speaker (Gibbs, 1994), whereas the difference between sarcasm and a lie relies on the shared knowledge between the speaker and the listener (Long & Graesser, 1988). If the new hat in the mentioned statement was, for example, full of stains, the speaker was most likely being sarcastic and if the listener also had this information, they would most likely be able to detect the sarcasm. When the speaker actually disliked the hat but perhaps wanted to be nice, the speaker was lying. Phillips et al. (2015) write that it has to be noted that speakers do not mean to be deceiving when sarcastic remarks are made. Although sarcasm and deception share the characteristic of a
cues to let the listener know they are being sarcastic, whereas in deception speakers try to not give clues of their true intention.
Incongruous language is also found in metaphors. Metaphors and sarcasm can be distinguished by the plausibility of statements. In the statement “John is a night owl” it is metaphorically meant that he stays up late instead of him being a literal owl, which, of course, is impossible. However, this statement, as in “you stay up late” is plausible for people who do stay up late, but sarcastic when made to a person who goes to bed early (Gibbs, 1994).
Sarcasm is very dependent on language- and culture specific traits, as well as sex. Rockwell and Theriot (2001) studied sarcasm in sex and culture. They found men to be more sarcastic than women, and men are more sarcastic to other men than to women. Women, in turn, were more sarcastic in conversations with men than in conversations with other women. Thus, both sexes are more sarcastic toward men than toward women. Rockwell and Theriot (2001) explain this by noting that “men view conversation as an opportunity to express their power and self-image” (pp. 49). Women, who are generally more interested in preserving relationships and do not want to hurt their collocutor, are less likely to be sarcastic. The reason why men are more sarcastic to other men than to women, might be explained by men feeling more comfortable to be themselves when they are around other men than when they are around women. Also, when a man is in the presence of another man or other men, they might feel a sense of competition and try to express their power through sarcasm. The researchers had trouble explaining why women are more sarcastic to other men than to other women. A possible explanation is that use of sarcasm is prompted when a woman is in the presence of a man. If a man uses sarcasm in a conversation with a woman, the woman might feel attacked and starts to defend herself by responding with sarcasm. When there is no man involved in a conversation, women may be more concerned about their conversational partner and therefore do not ‘attack’ the other woman. However, the researchers note that there are numerous possibilities to explain this result. Colston and Lee (2004) looked into perceiving and judging ironic language in genders. They found that ironic speakers were judged to be more likely male than female. It was also discovered that men reported they were more likely to use verbal irony relative to women. Like Rockwell and Theriot (2001), they explain this by males being riskier in discourse.
Rockwell and Theriot (2001) also looked into cultural features of sarcasm. They discovered that an individualist significantly uses more sarcasm than a collectivist. As individualists are more expressive in their emotions and giving their opinions than collectivists,
encouraged to have a positive attitude toward others rather than express negativity. Generally, people who live in individualist-based cultures may use more sarcasm than people in collectivist-based cultures.
Summarizing, men and individualists use more sarcasm than women and collectivists. Considering, Ivanko, Pexman, and Olineck (2004) and Kreuz, Dess, and Link (2006) suggest that the ability to perceive sarcasm is influenced by the frequency of sarcasm usage.
1.2 Perception of sarcasm
Lots of researchers agree that recognizing sarcasm can be hard and is dependent on multiple factors, including context, non-verbal cues, and tone of voice (Rockwell, 2000; Voyer & Techentin, 2010).
Adults are capable of noticing ambiguous messages through (the tone of) speech, gestures and facial expressions that are characteristic in sarcasm, where the exact opposite of an opinion is stated. However, adults also have trouble recognizing sarcasm and in particular older adults (Phillips, Allen, Bull, Hering, Kliegel, & Channon, 2015). To correctly interpret someone’s true intention, the use of paralinguistic and contextual cues is necessary. The literal meaning of the statement has to be suppressed and pragmatic cues have to be interpreted. One also has to be able to put themselves in another’s shoes, also known as Theory of Mind (Leverage, 2011). Without context, or gestures and facial expressions, recognizing sarcasm evidently becomes more complicated.
However, in Woodland and Voyer’s (2011) study, participants listened to short stories with either a congruent or incongruent context and tone of voice. A congruent context and tone of voice would be a positive context with a sincere tone, or a negative context with a sarcastic tone. An incongruent context and tone of voice would either be a positive context with a sarcastic tone, or a negative context with a sincere tone. Participants scored the stories as sincere or sarcastic on a Likert scale, where 1 was very sincere and 7 very sarcastic. Negative contexts elicited high ratings and were, in other words, perceived as sarcastic (M = 5.2). Ratings were lower in positive contexts, i.e. were perceived as sincere (M = 3.4). Participants were also able to distinguish between sincere and sarcastic tones of voice. When tone of voice and context were incongruent, participants rated the statement as neutral. Statements that were paired with a congruent tone of voice were scored corresponding to that specific tone of voice and context. The researchers predicted that context would be more important than tone of voice, but a listener considers both context and tone of voice when judging the intention of an utterance.
In their study, Joshi, Bhattacharyya and Carman (2017) described automatic detection of sarcasm in texts through prediction by using algorithms. A sarcastic sentence can carry a positive surface sentiment, like “I love going to the dentist”, or a negative surface sentiment, for example “Their national football team performed terribly all year anyway”, when actually they have just won the world cup. A sarcastic sentence can also carry no surface sentiment at all, like the example given by Joshi et al. (2017) “and I am the Queen of England”, which is an idiomatic expression that can be used sarcastically. The positive surface sentiment sentence can be interpreted as sarcasm, based on the linguistic notion of incongruity (Joshi, Bhattacharyya, & Carman, 2018). “Love”, a positive word, is not congruent to the - usually - negatively associated word “dentist”. These two words put together in one statement that are incongruous, could indicate sarcasm.
Liebrecht, Kunneman, and Van den Bosch (2013) researched automatic detection of sarcasm in mircotexts through prediction and used Twitter to collect a training corpus of 78 thousand Dutch tweets. Sarcastic messages on Twitter are often marked with the hashtag ‘#sarcasm’ (in Dutch ‘#sarcasme’) to avoid the message from being misunderstood. The researchers created a classifier (an algorithm) using their collected examples and next applied it to a test set of one day’s stream 3.3 million Dutch Tweets. Their classifier was able to detect 75% tweets correctly that originally were marked with the hashtag, with the hashtag removed. When they tested the classifier on the top 250 tweets most likely to be sarcastic, it could only detect 30% correctly. Therefore, the researchers concluded that it is rather hard to discriminate sarcastic tweets from literal tweets in an open setting (their second test).
Voyer (2003) studied how sarcastic speech is represented in the brain. They stimulated the right and left ear simultaneously using dichotic listening. The results showed an advantage for sincere tones in the left hemisphere of the brain, and an advantage for sarcastic tones in the right hemisphere. This suggests that sincere and sarcastic discourse are distinctively represented in the brain. Bryden and MacRae’s (1989) study indicated that tone of voice is a paralinguistic feature, which is processed in the right hemisphere. Additionally, in the study of Shamay-Tsoory, Tomer, and Aharon-Peretz (2005) sarcasm is linked to the right hemisphere, as well as the prosody that goes along with sarcasm. Rockwell (2000) claims that the capability to understand sarcasm depends on the listener’s perception of prosody.
1.3 Phonetic changes in sarcastic speech
Recognizing sarcastic speech is hard, and therefore speakers must differ their sarcastic speech’s features from their sincere speech to inform the listener about their intention. Some researchers have looked into acoustic differences in sincere and sarcastic speech, one being pitch. These studies have found contradicting results. Rockwell (2000) recorded speakers reading three different types of sentences: non-sarcastic, spontaneous sarcastic, and posed sarcastic sentences. Next, listeners were asked to rate the sentences on amount of sarcasm. They could distinguish posed sarcasm from non-sarcasm, but were unable to discriminate spontaneous sarcasm from non-sarcasm. Rockwell also looked into the vocal features of the recorded utterances and found that a lower pitch in was favored in sarcastic speech than in non-sarcastic speech. Other studies, however, discovered that sarcastic speakers’ pitch was raised (Anolli, Ciceri, & Infantino, 2000; Rockwell, 2007).
Differences in pitch range have also been looked into. Several studies found a more flattened (monotone) pitch range in sarcasm (Milosky & Wroblesky, 1994; Shapley, 1987), whereas other researchers noted that sarcastic speakers maintained a wide pitch range and even had exaggerated changes in pitch (Haiman, 1998; Milosky & Ford, 1997).
Finally, diversity in tempo was researched and tempo was found to be slower in sarcastic speech than in non-sarcastic speech (Kreuz & Roberts, 1995; Rockwell, 2000). Haiman (1998) and Schaffer (1982) noted that in sarcastic speech longer pauses were present.
As happens often in research, there is no conclusive answer to which factors determine sarcastic speech. Other researchers than mentioned before, claim that, like any other prosody, sarcasm can be identified by a combination of various acoustic cues (Bryant & Fox Tree, 2005; Voyer & Techentin, 2010). Because of all the conflicting data mentioned before, this would seem a logical conclusion (Woodland & Voyer, 2011). Alternatively, the apparent impossibility of defining the vocal features of sarcastic speech could also indicate a total lack of a systematic acoustic difference.
As mentioned in section 1.1, five different subtypes of irony are described by Gibbs (2000), namely sarcasm, jocularity, understatements, rhetorical questions, and hyperbole. Furthermore, sarcasm can be divided into the four subtypes propositional, embedded, like-prefixed, and illocutionary sarcasm. Different results in pitch, pitch variation and tempo could be a result of different types of verbal irony and/or types of sarcasm used in these studies, or because different methods were used.
Few of the studies mentioned before where performed in the Dutch language, and pitch patterns vary in languages. Therefore, pitch patterns in sarcastic speech in Dutch could differ from the languages used in previous studies, which is mostly English.
1.4 Research questions and hypotheses
Due to multiple studies finding different results regarding vocal features in sarcasm, the present study aims to clarify these contradictions. The following research questions and hypotheses are presented:
Q1: What are the phonetical differences in sincere utterances and sarcastic utterances?
Q1a: Does average pitch in sincere utterances and sarcastic utterances differ?
H1a: The average pitch in sincere and sarcastic utterances differ, and pitch in sarcastic
speech is either higher or lower than pitch in sincere speech.
Q1b: Does average pitch variation in sincere utterances and sarcastic utterances
differ?
H1b: It is expected that pitch variation in sarcastic utterances differs from the pitch
variation in sincere utterances, and pitch variation in sarcastic speech is either higher or lower than pitch variation in sincere speech.”
Q1c: Does average tempo (in duration per syllable) in sincere utterances and sarcastic
utterances differ?
H1c: It is expected that the average duration per syllable is longer in sarcastic
utterances than in sincere utterances.
Q2: Are people able to distinguish sincere utterances from sarcastic utterances without context, gestures, and facial expressions?
H2: It is expected that, without context, gestures, and facial expressions, people are not able to differentiate sincere utterances from sarcastic utterances.
Q3: Are males better at recognizing sarcastic speech compared to females?”
2. Method
2.1 Production
2.1.1 Participants
Nine part-time actors with at least one year of acting experience took part in this study (two males and seven females). Their age ranged from 20-23 years old with an average age of 21,5. All of the participants’ native language was Dutch and they had no other native language.
2.1.2 Design (procedure and materials)
Every participant was first asked to fill in a form of consent. In this part of the study, the part-time actors read ten Dutch sentences aloud and their speech was recorded with a voice recorder. Prior to the actual test, the participants read an instruction and two practice sentences. The structure of each sentence was the same; an introduction sentence followed by a target sentence (see Appendix A). The target sentences were either pronounced sincerely or sarcastically, depending on the introduction sentence. Two introduction sentences were created for each target sentence; one introduction sentence that would elicit a sincere pronunciation of the target sentence, and one introduction sentence that would elicit a sarcastic pronunciation of the target sentence. The introduction sentences were made to create context and to indicate if the target sentence should be pronounced either sincerely or sarcastically. Examples of introduction and target sentences are shown in Table 1.
Table 1
Example of target sentences and their counter introduction sentences.
Tone Introduction sentence Target sentence Original Dutch sentences:
Sincere
Sarcastic
De kok in het driesterrenrestaurant had zeven uur lang in de keuken gestaan om een kerstmaal te bereiden. De kok raffelde in twintig minuten een kerstmaal af en serveerde oude diepvriespizza’s.
Het eten smaakte fantastisch.
English translation: Sincere
Sarcastic
The famous restaurant’s cook was in the kitchen for seven hours to cook a Christmas dinner.
The restaurant’s cook served spoiled frozen pizzas for Christmas dinner.
The food was delicious.
The introduction sentence was always read previous to the corresponding target sentence. To prevent the order of the target sentences from influencing each other, two versions of the test were created. The order and variety of the target sentences differed in both versions. Each version contained, however, five sincere target sentences and five sarcastic target sentences. Every participant only read a specific target sentence once, either the sincere version or the sarcastic version, but never both.
The recorded target sentences, both sincere and sarcastic, were used for analyses. The target sentences were isolated in Adobe Audition and were then imported in PRAAT for analyses. In PRAAT the average pitch, pitch variation, and speech rate (duration per syllable) of the sincere and sarcastic utterances were measured. Next, the measurements of the sincere and sarcastic utterances were compared to each other.
2.2 Perception
2.2.1 Participants
52 students between the age of 18-26 participated in this study (31 females and 21 males). The average age was 22. At the time of testing, participants were all students in the Arts faculty of the University of Groningen. All participants were native speakers of Dutch and had no other native language.
2.2.2 Design (procedure and materials)
In this part of the study, the material gained in the production part was used (see section 2.1.2). Means of pitch, pitch variation and duration per syllable were measured for sincere and sarcastic utterances in both sexes, and the utterances with the most extreme features were used. For example, mean pitch in sincere utterances in females was 215 Hz, and 199 Hz in sarcastic utterances and therefore sincere utterances with the highest measured pitch were selected to be used in part 2 and the lowest measured pitch for sarcastic utterances. Means of pitch variation and duration per syllable were measured and selected the same way. An utterance that met at least two out of three criteria, were considered to be suitable to use in part 2.
There were two different sections in the perception part of the study:
1. The participants listened to the isolated target sentences - without the introduction sentences - and scored these sentences as either sincere or sarcastic.
2. The participants listened to both versions (sincere and sarcastic) of the same isolated target sentence (but recorded by different participants) and scored which sentence was
sincere and which sentence was sarcastic. The sentences in this part were never the same sentences as the ones used in part one and the first section of part two.
The participants were allowed to listen to every audio fragment as many times as they preferred. Subsequent to each question, participants filled in a 5-point Likert-scale where they scored how sure they were of their answer, ranging from ‘very unsure’ to ‘very sure’. Answers were next imported into SPSS for analyses.
3. Results
3.1 Production
3.1.1 Descriptive results
All the sincere and sarcastic target sentences were isolated using Adobe Audition and were analyzed in PRAAT. The average pitch and pitch variation (minimum pitch subtracted from maximum pitch) of each isolated sentence, and average duration per syllable were measured, for examples see Table 2. The average duration per syllable was measured by dividing the sentence duration in seconds by the number of syllables. This delivered a list of ninety measurements per parameter, since nine participants recorded ten sentences. All means were measured in semitones.
Table 2
Calculation of pitch variation and duration per syllable (real measurements).
Pitch variation Duration per syllable
Highest pitch measured Lowest pitch measured Pitch variation Total duration sentence Syllables in sentence Duration per syllable 319 Hz 162 Hz 157 Hz 1,3 secs 8 0,16 secs 196 Hz 83 Hz 113 Hz 2,15 secs 10 0,22 secs 195 Hz 89 Hz 106 Hz 1,33 secs 6 0,22 secs
Next, the data were imported in SPSS for statistical analyses. There were two variables per hypothesis, namely a sincere-variable and a sarcastic-variable. Levels of significance were measured by performing a Mann Whitney U test. The mean for all parameters were measured as well, to determine the direction of a possible significant outcome.
3.1.2 Results production
As mentioned before, the average pitch and pitch variation of the sincere and sarcastic utterances were measured, as well as the duration per syllable. The Shapiro-Wilk test was performed to test normality of the data, see Table 3. Normality of the data is tested to check if the data is normally distributed, which determines what statistical test has to be performed.
Table 3
Results of the Shapiro-Wilk test.
Average pitch Pitch variation Duration per syllable Sincere p = 0,002* Sincere p = 0,248 Sincere p = 0,052
Sarcastic p = 0,807 Sarcastic p = 0,007* Sarcastic p = 0,001*
* = significant result
The Shapiro-Wilk test indicated that not all the data were normally distributed. Therefore, a Mann Whitney U test was performed instead of an independent t-test to account for the lack of normality.
Due to pitch differences between males and females and also in between individuals, all means of semitones were measured. An octave consists of twelve semitones and this system controls for the pitch-system not being linear. When a frequency is doubled in Hz, for example 65 Hz to 130 Hz (a difference of 65 Hz), equals twelve semitones (one octave). However, 130 Hz to 260 Hz equals twelve semitones as well, yet this is a difference of 130 Hz (Moon, 1953). If means are compared using the non-linear system in the present study, this would lead to erroneous assumptions.
An average pitch of 90,5 Hz was found in sincere utterances, slightly higher than the average pitch in sarcastic utterances (89,5 Hz), see Table 4. The Mann Whitney U test indicated that the average pitch in sincere utterances was not significantly higher than the average pitch in sarcastic utterances (p = 0,07).
The average pitch variation in sincere utterances (15,7 Hz) was slightly lower than the average pitch variation in sarcastic utterances (17,2 Hz). According to the Mann Whitney U test that was performed, these differences were not significant (p = 0,40).
The differences of average duration per syllable between sincere and sarcastic utterances were also measured. The Mann Whitney U test revealed that the average duration of syllables in sincere sentences (0,197 seconds) differed significantly (p < 0,001) from the average duration of syllables in sarcastic sentences (0,228 seconds).
Table 4
Overview of vocal features in sincere and sarcastic speech.
Parameter Sincere Sarcastic
Mean Sd. Mean Sd. Sig. N Pitch 90,5 4,3 89,5 3,7 0,07 45 Pitch variation 15,7 7,8 17,2 8,0 0,40 45 Duration Syllable (sec) 0,197 0,033 0,228 0,047 <0,001* 45 * = significant result
Differences in features in sincere and sarcastic speech were also measured within sexes, see Table 5. Female’s speech turns out to change, whereas male’s speech does not change at all. Female’s pitch in sarcastic speech lowers significantly (p = 0,005), as well as female’s tempo (p < 0,001).
Table 5
Vocal features of males’ and females’ sincere and sarcastic speech.
Male Female
Sincere Sarcasm Sig. Sincere Sarcasm Sig. Pitch 83,3 83,9 0,281 92,5 91,1 0,005* Pitch variation 10,7 11 0,529 17,1 19,0 0,344 Duration/ syllable 0,193 0,194 0,971 0,198 0,238 <0.001* * = significant result 3.2 Perception 3.2.1 Descriptive results
The isolated target sentences were imported in a survey using rug.qualtrics.com. The survey consisted of two parts; ten questions with an isolated target sentence which participants had to score as either sincere or sarcastic (part 1), and five questions with both an isolated sincere and sarcastic utterance where participants had to score which utterance was sarcastic (part 2). After each question participants answered how sure they were of their answer.
The answers of the participants were assigned points. If participants perceived an utterance as sincere, one point was assigned. Sarcastic answers were assigned two points
(regardless of the correct answer). In part 2, perceiving sentence 1 as sarcastic was assigned one point and sentence 2 two points. This produced a mean between one and two for both parts, which is shown in Table 6. A mean between 1-1,5 suggests that the majority of the participants perceived the utterance as sincere, and a mean between 1,5-2 indicates that it was perceived sarcastic. In part 2, a mean between 1-1,5 implies that the majority of the participants perceived sentence 1 as sarcastic, and sentence 2 was perceived as sarcastic if the mean is between 1,5-2. As shown in Table 6, the majority of the participants answered the questions correctly.
Table 6 also shows the level of certainty with which the participants answered the questions regarding the utterances. A 5-point Likert scale was used, and participants could choose between the following answers: 1) very unsure, 2) unsure, 3) neutral, 4) sure and 5) very sure. Again, this produced a mean between one and five. The higher the mean, the more certain participants were of their given answer. A Likert scale was filled in after each of the fifteen questions.
Table 6
Overview of the results of the perception section.
Question
Correct
answer Given answer
% correct Mean Certainty (mean) Sincere Sarcastic 1 Sincere 44 8 85% 1,15 3,42 2 Sarcastic 4 48 92% 1,92 4,5 3 Sincere 48 4 92% 1,08 3,88 4 Sincere 35 17 67% 1,33 3,33 5 Sarcastic 12 40 77% 1,77 3,48 6 Sarcastic 11 41 79% 1,79 3,63 7 Sarcastic 11 41 79% 1,79 3,2 8 Sincere 51 1 98% 1,02 4,08 9 Sarcastic 16 36 69% 1,69 3,6 10 Sincere 49 3 94% 1,06 3,58 Question Correct
answer Given answer
%
correct Mean Certainty Sentence 1 Sentence 2 11 Sentence 1 51 1 98% 1,02 4,12 12 Sentence 1 45 7 87% 1,13 3,33 13 Sentence 2 6 44 88% 1,88 3,4 14 Sentence 2 4 47 92% 1,92 4,25 15 Sentence 1 50 1 98% 1,02 3,94
As shown in Table 6, question 4 (67%) and 9 (69%) were not as much correctly answered as the other questions. On the other hand, participants performed extremely well on question 8 (98%) and 10 (94%). When looking into the vocal features of these utterances, nothing in particular stands out. The necessary criteria were met to use these utterances in the perception part of the study.
3.2.2 Results perception
The percentage of correct given answers in part 1 and part 2 were compared by performing a Mann Whitney U test. In part 1, an average of 83,2% of the participants answered correctly (but of these, sincere sentences were correctly perceived in 87,2% of the cases and sarcastic sentences only 79,2%), whereas in part 2 participants answered correctly almost 10% more often (92,6%). This difference turned out not to be significant (p = 0,108).
Difference in levels of certainty in part 1 and 2 were also compared, see Table 7. In part 1, participants scored how sure they were of their answer being the correct answer with 3,67 (1 is very unsure, 5 is very sure). In part 2, participants were slightly more certain their answer was correct (3,81). Again, a Mann Whitney U test was performed to compare averages. With
p = 0,581 this difference was not significant.
Table 7
Mean scores of accuracy and certainty of participants answers.
% correct Certainty
Part 1 83,2 3,67
Part 2 92,6 3,81
Sig. 0,108 0,581
3.2.3 Perception of sarcasm in sexes
The difference in percentage of correct given answers by males and females were also compared, see Table 8. Their answers were acquired by using a filter in the questionnaire program Qualtrics.
Table 8
Mean scores of accuracy and certainty in sexes.
% correct Certainty
Male Female Sig. Male Female Sig. 82,9 85,0 0,93 3,8 3,7 0,39
Males answered correctly on an average of 82,9%, females performed slightly better with 85%. A Mann Whitney U test showed that this difference was not significant (p = 0,93). Certainty of given answers by men and women were compared as well. On a scale of 1 to 5, men scored the chance of their answers being correct with an average of 3,8. Women were almost as certain of their answers as men and scored their answers on average 3,7. Again, a Mann Whitney U test was performed and this difference of 0,1 turned out not to be significant (p = 0,39).
Table 9 shows results on the relation between the sex of the speaker and which sex scored best per question. In questions 1, 2, 6, 7, and 10, the sex of the speaker and sex of the best performer differ. Considering participants listened to two different speakers in part two, these questions were left out of this analysis.
Table 9
Relation between speaker’s sex and best scoring sex.
Question Speaker % correct Best result Male Female 1 Female 85,7 83,8 Male* 2 Male 85,7 96,8 Female* 3 Female 90,4 93,5 Female 4 Male 71,4 64,5 Male 5 Female 66,6 83,8 Female 6 Female 80,9 77,4 Male* 7 Female 85,7 74,2 Male* 8 Male 100 96,8 Male 9 Female 47,6 83,8 Female 10 Female 100 90,3 Male* * Sex of speaker and best performing sex do not match
4. Discussion
4.1 Production
The present study aimed to determine what the vocal features of sarcastic speech are, and how these features in sarcastic speech differ from sincere speech. The research question in respect to this topic was: “What are the phonetical differences in sincere utterances and sarcastic utterances?” To be able to answer such a broad question, three sub-questions and respecting hypotheses were composed, regarding pitch, pitch variation and tempo (in duration per syllable).
4.1.1 Pitch
In her study, Rockwell (2000) found that pitch in sarcastic speech is lower in comparison to pitch non-sarcastic speech. Other researchers (Anolli, Ciceri, & Infantino, 2000; Rockwell, 2007), however, have found the exact opposite. In their studies, pitch was actually higher in sarcastic speech rather than lower. Therefore, the research question is “Does average pitch in sincere utterances and sarcastic utterances differ?” Because of the varying results, the hypothesis in regard to the research question is as follows: “The average pitch in sincere and sarcastic utterances differ, and pitch in sarcastic speech is either higher or lower than pitch in sincere speech.”
The results of the present study do not confirm the findings of Rockwell (2000), nor the findings of Anolli, Ciceri and Infantino (2000) and Rockwell (2007), as no significant difference in pitch in sincere and sarcastic speech was found in the present study. There is, however, a significant difference in pitch in female participants. In sarcastic speech, their pitch is lower with respect to their pitch in sincere speech.
4.1.2 Pitch variation
Various studies have looked into pitch range in sarcastic speech. For example, Milosky and Wroblesky (1994), and Shapley (1987) found that the pitch range in sarcastic speech is more monotone (flattened) than in non-sarcastic speech. Other researchers’ studies (Haiman, 1998; Milosky & Ford, 1997), on the contrary, showed that sarcastic speakers’ pitch range maintained wide and the changes in pitch were even more extreme. Again, the results of different studies are contradictive and thus the research question in respect to pitch variation is “Does average pitch variation in sincere utterances and sarcastic utterances differ?” Respectively, the
variation in sincere utterances, and pitch variation in sarcastic speech is either higher or lower than pitch variation in sincere speech.”
In the present study, again no significant differences were found in sincere and sarcastic speech and consequently, the stated hypothesis was not confirmed.
4.1.3 Tempo
Kreuz and Roberts (1995), and Rockwell (2000) looked into tempo in sarcastic speech and discovered that speakers’ tempo is slower in sarcastic speech than in non-sarcastic speech. Haiman’s (1998) findings on tempo add that longer pauses are present in sarcastic speech. Due to these researchers’ findings, the research question regarding tempo is “Does average tempo (in duration per syllable) in sincere utterances and sarcastic utterance differ?” In respect to Kreuz and Roberts (1995), and Rockwell’s (2000) findings, the hypothesis is that the average duration per syllable is longer in sarcastic utterances than in sincere utterances.
The hypothesis in the present study regarding tempo was confirmed. Tempo is indeed slower in sarcastic speech than in sincere speech. On average, the duration of a syllable in sarcastic speech was almost 17% longer relative to the duration of a syllable in sincere speech. This is, however, not the case for male participants. General tempo is slower in sarcastic speech, but when looking into tempo in sexes, male’s tempo does not change significantly. As only two male participants took part in this study as opposed to seven females, this result could thus be explained.
Listening to participants’ utterances, pauses are also present in sarcastic utterances and even in words in sarcastic utterances. This often happened in content words where emphasis can be put on, like “amazing” and “fantastic”, in Dutch “geweldig” and “fantastisch”. These words were pronounced with pauses in-between the syllables “ge-wel-dig” and “fan-tas-tisch” (the dash marking a pause). This contributes to the syllables’ duration. Kreuz and Roberts (1995) and Rockwell (2000) noted that slower tempo may actually be a longer duration of vowels. The slower tempo in sarcastic speech could be the result of a combination of more pauses and vowel-lengthening.
4.2 Perception
Distinguishing sarcastic speech from non-sarcastic speech is no easy task, especially without context. Context, gestures, and facial expressions (like eye-rolling) help listeners recognizing sarcasm. Furthermore, to be able to interpret someone’s true intention, one needs to use
paralinguistic and contextual cues, suppress the literal meaning of the statement, and interpret pragmatic cues (Phillips, Allen, Bull, Hering, Kliegel, & Channon, 2015; Rockwell, 2000).
In respect to previous research on the perception of sarcasm, the following research question in the present study was presented: “When context, gestures, and facial expressions are not presented to participants, will they still be able to recognize sarcastic speech?” The corresponding hypothesis was constructed: “It is expected that, without context, gestures, and facial expressions, people are not able to differentiate sincere utterances from sarcastic utterances”.
In this part of the study, participants listened to isolated sincere and sarcastic utterances. In part 1 they had to answer if the heard utterance was sincere or sarcastic. Surprisingly, 83,2% of the questions were correctly answered, where sincere utterances were correctly perceived in 87.2% of the cases and sarcastic utterances only 79.2%. In part 2, participants listened to a sincere and a sarcastic utterance per question and had to answer which of the utterances was sarcastic. Now, 92,6% of the questions were correctly answered.
All together, these results were unexpected. Without a presented context, participants had to fully rely on acoustic cues (pitch and tempo). This could suggest that the so-called ‘ironic tone of voice’ (Clark and Gerrig, 1984; Sperber and Wilson, 1981) is more important than context to recognize sarcasm, as no context was presented to participants, but further research is required. Sperber and Wilson (1981) describe the ‘ironic tone of voice’ as people leaving their own voices behind in pretense for new voices. They exemplify this by an actor playing Othello who adjusts his voice that is appropriate to the character they are portraying. The same could be the case for sarcasm, where sarcastic speech and the corresponding attitude elicits a new voice. However, participants in this study had never met the participants whose speech was recorded, and therefore could not know if their voice changed in sarcastic speech. Unless this new voice has general acoustic features that are identical for every speaker, it is unlikely that the ironic tone of voice caused the results found in this part of the present study.
4.3 Perception in sexes
Rockwell and Theriot (2001) looked into sarcasm in gender and culture. According to their study, men are more sarcastic than women, and both sexes are more likely to be sarcastic toward men than toward women. Ivanko et al. (2004) and Kreuz et al. (2006) suggest that the ability to perceive sarcasm is linked to the frequency of the use of sarcasm. This would imply that, as women use less sarcasm, they are therefore less able to recognize sarcastic speech. Accordingly,
females?” was formulated. In respect to the studies of Ivanko et al. (2004) and Kreuz et al. (2006), it was hypothesized that males are able to distinguish sarcasm in more cases than females can.
The results showed that there is no significant difference in sex when it comes to recognizing sarcasm, as well as for certainty about given answers. The material used in this study were recordings of two males and seven females, and questions about these recordings were answered by 21 males and 31 females. In both tests the majority consisted of women, which could have influenced the results. However, when looking at the results in part 1 of the perception section, in 50% of the questions the sex opposite from the speaker performed better. It can therefore be concluded that sex of speakers and listeners did not influence the results of this study.
Rockwell and Theriot (2001) noted that women are less likely to use sarcasm in conversations because they do not want to hurt their conversation partner, but in the present study participants did not have a conversation partner since they simply recorded utterances and could not hurt anyone directly. Consequently, female participants may have felt freer to use ‘true’ sarcasm. In turn, according to Rockwell and Theriot (2001) wrote that men use more sarcasm out of a feeling of superiority. But again, there was no conversation partner and they had no one to be superior to. The circumstance of not having a conversation partner could elicit more sarcastic features congruent to sarcasm in women and the opposite for men. This equalized the features of sarcasm in their sarcastic speech.
5. Conclusion
5.1 Overall conclusion
When comparing acoustic cues of sincere and sarcastic speech, tempo is slower in sarcastic speech with respect to sincere speech. However, when comparing the acoustic cues within sexes, there is no difference in tempo in male participants. Furthermore, pitch is lower in female’s sarcastic speech compared to pitch in female’s sincere speech. There are no differences in pitch variation. Additionally, recognizing sarcasm is easier when both a sincere and sarcastic sentence are presented, but isolated sentences are still perceived correctly most of the time.
5.2 Limitations and recommendations
One of the limitations in this study was the uneven distribution of male and female participants in the production part. Seven females participated versus only two males. Comparing samples of at least five male participants would have given a more trustworthy outcome. Also, in this study, the participants did not record a specific sentence twice (i.e. the sincere version and the sarcastic version of the same sentence). This was done to not let the different versions of the sentences influence the other sentence, for instance, if the sincere version was recorded subsequent to the sarcastic version, the sincere version might have had ‘traces’ of sarcastic cues. When using the method of the present study in future research, this could be prevented by taking a break before the other versions of the same sentences are recorded. The latter would allow to detect acoustic differences by directly comparing the same sentences spoken by the same speaker.
Interestingly, in this study, almost no acoustic differences were found in sincere and sarcastic speech, yet participants did perceive a difference. This might suggest that clues other than examined in previous studies and the present study are important for recognizing sarcasm, for example pauses. A recommendation for future research is to also include a questionnaire where participants are asked what clues they looked for when assessing a sentence. Were they focusing on pitch, tempo, or something else?
Finally, it is debatable how natural ‘acting sarcastic’ is. The participants in this study that recorded the sincere and sarcastic sentences were actors and most had theatre experience, and the acting in plays is often exaggerated. In future research, sincere and sarcastic sentences could be, for example, elicited in a natural conversation. As opposed to acting, you would be able to get the participants’ true acoustic cues of their sarcastic speech.
Appendix: Material used for production
Sincere: Toen de klas naar de Efteling ging, bleek al het eten daar vandaag gratis te zijn, was het de hele dag zonnig en had iedereen een goed humeur.
Sarcastic: Toen de klas bij de Efteling aankwam, bleek het park die dag gesloten te zijn en moesten ze weer naar huis. De verwarming van de bus deed het niet, waardoor het ijskoud was. Target sentence: Het was een geweldige dag.
Sincere: De kok in het driesterrenrestaurant had zeven uur lang in de keuken gestaan om een kerstmaal te bereiden.
Sarcastic: De kok raffelde in twintig minuten een kerstmaal af en serveerde oude diepvriespizza’s met bedorven mayonaise.
Target sentence: Het eten smaakte fantastisch.
Sincere: De schoolmeester hielp alle kinderen geduldig met hun werkstuk.
Sarcastic: De schoolmeester kraakte de werkstukken van alle kinderen af waar ze hard aan hadden gewerkt en gaf iedereen een onvoldoende.
Target sentence: Hij is ontzettend aardig.
Sincere: Toen we pech kregen onderweg kwam de monteur er binnen 5 minuten aanrijden en repareerde de auto in 10 minuten.
Sarcastic: Toen we pech kregen onderweg was de monteur pas na 5 uur aanwezig en kon de auto niet repareren.
Target sentence: Hij heeft ons fantastisch geholpen.
Sincere: De bruid had een jurk van 5000 euro aan en had haar make-up en haar door een professional laten doen.
Sarcastic: De bruid had twee handdoeken aan elkaar genaaid en droeg dat als bruidsjurk, waarna ze er rode wijn op knoeide.
Target sentence: Ze zag er prachtig uit.
Sincere: Vorige week zijn we op vakantie geweest naar Bali. Het weer was fantastisch en de vluchttijden waren ideaal.
Sarcastic: Vorige week zijn we op vakantie geweest naar Terschelling. Het regende alle dagen en de boot ging op de terugweg kapot waardoor we 3 uur lang vastzaten op zee.
Target sentence: Dat gaan we zeker nog een keer doen.
Sincere: Tijdens de presentatie gisteren luisterde iedereen geïnteresseerd, werden er veel leuke vragen gesteld en volgde er na afloop een luid applaus.
Sarcastic: Tijdens de presentatie werkte de PowerPoint niet, viel de stroom uit na 5 minuten en liep iedereen weg voordat de presentatie klaar was.
Target sentence: Beter kon het niet gaan.
Sincere: De winnaar van de New York Marathon zette een nieuw wereldrecord neer met zijn tijd van 1 uur, 59 minuten en 31 seconden.
Sarcastic: De bus die de jongen wilde halen stond op 100 meter afstand en vertrok over 5 minuten. Hij rende er naartoe, maar kwam alsnog te laat.
Target sentence: Wat kan hij hard rennen.
Sincere: De vrouw deed al 40 jaar lang elke dag vrijwilligerswerk bij het asiel en had meerdere malen een beestje geadopteerd.
Sarcastic: De vrouw had een hond uit het dierenasiel gehaald, maar dumpte het beestje gelijk nadat hij de volgende dag op haar tapijt plaste.
Target sentence: Haar levenskeuzes jaloersmakend.
Sincere: De weersvoorspellingen van de weerman op tv komen al jarenlang elke dag uit. Sarcastic: De weerman op tv heeft al jarenlang de weersvoorspellingen dag na dag fout. Target sentence: Hij is echt een vakman.
References
Anolli, L., Ciceri, R., & Infantino, M. G. (2000). Irony as a game of implicitness: Acoustic profiles of ironic communication. Journal of Psycholinguistic Research, 29, 275–311. Anolli, L., Ciceri, R., & Infantino, M. G. (2002). Behind dark glasses: Irony as a strategy for
indirect communication. Genetic, Social, and General Psychology Monographs, 128(1), 76-95.
Attardo, S., Wagner, M., (Educator), & Urios-Aparisi, E. (Eds.). (2013). Prosody and
humor (Benjamins current topics, v. 55). Amsterdam: John Benjamins Publishing
Company. (2013).
Bryant, G. A., & Fox Tree, J. E. (2005). Is there an ironic tone of voice? Language and Speech, 48: 257–277.
Bryden, M. P., & MacRae, L. (1989). Dichotic laterality effects obtained with emotional words.
Neuropsychiatry, Neuropsychology and Behavioural Neurology, 1, 171–176.
Burton, G. (2007). “Rhetorical questions”. Specialized language definitions. Brigham Young University.
Camp, E. (2012). Sarcasm, pretense, and the semantics/pragmatics distinction. Noûs 46, 4 (2012), 587–634
Cheang, H., & Pell, M. (2008). The sound of sarcasm. Speech Communication, 50(5), 366-381. doi:10.1016/j.specom.2007.11.003.
Clark, H. (1996). Using language. Cambridge: Cambridge University Press, 1996. Pp. xi+432.. Journal of Linguistics. 3s5. 167 - 222. 10.1017/S0022226798217361.
Colston, H. L. (1997). Salting a wound or sugaring a pill: The pragmatic functions of ironic criticisms. Discourse Processes, 23, 24-45.
Colston, H. L., & Sabrina Y. Lee (2004) Gender Differences in Verbal Irony Use, Metaphor and Symbol, 19:4, 289-306.
Colston, H. L., & O'Brien, J. (2000). Contrast and pragmatics in figurative language: Anything understatement can do, irony can do better. Journal of Pragmatics - J PRAGMATICS. 32. 1557-1583. 10.1016/S0378-2166(99)00110-1.
Dews, S., Kaplan, J., & Winner, E. (1995). Why not say it directly? The social functions of irony. Discourse Processes, 19, 347-367.
Ducharme, L. (1994). Sarcasm and interactional politics. Symbolic Interaction, 17(1), 51-62. doi:10.1525/si.1994.17.1.51.
Gibbs Jr., R. W. (1994). The Poetics of Mind: Figurative Thought, Language, and
Understanding. Cambridge University Press.
Gibbs Jr., R. W. (2000). Irony in talk among friends. Metaphor and Symbol 15 (1-2), 5-27 Haiman, J. (1998). Talk is cheap: Sarcasm, alienation, and the evolution of language. New
York: Oxford University Press. (2001). Retrieved December 18, 2018.
Ivanko, S. L., Pexman, P. M., & Olineck, K. M. (2004). How sarcastic are you? Individual differences and verbal irony. Journal of Language & Social Psychology, 23: 244–271. Joshi, A., Bhattacharyya, P., & Carman, M. (2017). Automatic sarcasm detection. Acm
Computing Surveys, 50(5), 1-22. doi:10.1145/3124420
Joshi, A., Bhattacharyya, P., & Carman, M. (2018). Investigations in computational
sarcasm (Cognitive systems monographs, 37). Singapore: Springer.
doi:10.1007/978-981-10-8396-9.
Kreuz, R. J, Dress, M. L. and Link, K. E.July 2006. “Regional differences in the spontaneous use of sarcasm”. In Paper presented at the annual meeting of the Society for Text and
Discourse.
Kreuz, R., & Glucksberg, S. (1989). How to be sarcastic: The echoic reminder theory of verbal irony. Journal of Experimental psychology: General, 118, 374-386.
Kreuz, R. J., & Roberts, R. M. (1995). Two cues for verbal irony: Hyperbole and the ironic tone of voice. Metaphor and Symbolic Activity, 10, 21–30.
Kumon-Nakamura, S., Glucksberg, S., & Brown, M. (1995). How about another piece of pie: The allusional pretense of theory of discourse irony. Journal of Experimental
Psychology: General, 124, 3-121.
Lee, C. J., & Katz, A. N. (1998). The differential role of ridicule in sarcasm and irony.
Metaph. Symb. 13, 1 (1998), 1–15.
Leverage, P. (2011). Theory of mind and literature. West Lafayette, Ind.: Purdue University Press.
Liebrecht, C. C., Kunneman, F. A., & Van den Bosch, A. P. J. (2013). The perfect solution for detecting sarcasm in tweets #not (2013). In Proceedings of the Workshop on
Computational Approaches to Subjectivity, Sentiment, and Social Media by the North American Chapter of the Association for Computational Linguistics.
Long, D. L., & Graesser, A. C. (1988). Wit and humor in discourse processing. Discourse
Processes 11, 1 (1988), 35–60.
Milosky, L. M., & Wrobleski, C. A. (1994). The prosody of irony. Paper presented at the International Society for Humor Studies Conference. Ithaca, NY.
Moon, P. (1953). A scale for specifying frequency levels in octaves and semitones. The Journal
of the Acoustical Society of America, 25(3), 506-515.
Phillips, L., Allen, R., Bull, R., Hering, A., Kliegel, M., & Channon, S. (2015). Older adults have difficulty in decoding sarcasm. Developmental Psychology, 51(12), 1840-1852. doi:10.1037/dev0000063.
Rockwell, P. (2000). Lower, slower, louder: Vocal cues of sarcasm. Journal of
Psycholinguistic Research, 29(5), 483–495.
Rockwell, P. (2007). Vocal features of conversational sarcasm: A comparison of methods. Journal of Psycholinguistic Research, 36(5), 361-369. doi:10.1007/s10936-006-9049-0.
Rockwell, P., & Theriot, E. M. (2001). Culture, gender, and gender mix in encoders of sarcasm: A self‐assessment analysis, Communication Research Reports, 18:1, 44-52.
Shamay-Tsoory, S. G., Tomer, R., & Aharon-Peretz, J. A. (2005). The neuroanatomical basis of understanding sarcasm and its relationship to social cognition. Neuropsychology, 19, 288 –300.
Shapley, M. (1987). Prosodic variation and audience response. IprA: Papers in Pragmatics,
1(2), 66–79.
Sperber, D., & Wilson, D. (1995). Relevance: Communication and cognition (2nd ed.).
Cambridge, MA: Blackwell.
Voyer, D. (2003). Reliability and magnitude of perceptual asymmetries in a dichotic word recognition task. Neuropsychology, 17, 393– 401.
Voyer, D. & Techentin, C. (2010). Subjective acoustic features of sarcasm: Lower, slower, and more. Metaphor and Symbol, 25: 1–16.
Wilson, D., & Sperber, D. (1992). On verbal irony. Lingua, 87, 53-76.
Woodland, J., & Voyer, D. (2011). Context and Intonation in the Perception of Sarcasm, Metaphor and Symbol, 26:3, 227-239
Zvolenszky, Z. (2012). A gricean rearrangement of epithets. Twenty Years of Theoretical
