JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2002 Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

UNDERSTANDING IRONY How Do Stereotypes Cue Speaker Intent?

PENNY M. PEXMAN KARA M. OLINECK
University of Calgary

Katz and Pexman reported that certain occupations (e.g., comedian) were associated with ironic speech and that participants rated metaphors as more sarcastic when speakers were members of such occupations. In the present research, the authors investigated whether speaker occupation was a cue to ironic intent when the statements were not metaphors (e.g., literal statements such as you are a wonderful friend, potentially an ironic insult, and you are a terrible friend, potentially an ironic compliment). Results of Experiments 1 and 2 demonstrated that speaker occupation stereotypes were routinely integrated in the comprehension process but only cued ironic intent when other contextual cues were minimal. Results of Experiment 3 demonstrated that speaker occupation stereotypes involve particular types of information in the context of potentially ironic speech: a speakers perceived tendencies to be humorous, to criticize, to be sincere, and also a speakers perceived education level.

When a speaker says you are a wonderful friend, the statement could be interpreted as a literal compliment (e.g., you really are a great friend) or as an ironic1 insult (e.g., you are a rotten friend). The interpretation that the listener takes will depend on a number of factors, including, perhaps, their beliefs about the speaker. If the speaker is someone whom the listener believes to be highly sarcastic, then the statement may be more likely to be taken as ironic. Conversely, if the listener believes that the speaker hardly ever speaks sarcastically, the statement may be less likely to be interpreted ironically. It seems highly plausible that a listeners beliefs about a speakers conversational tendencies (e.g., a tendency to use sarcasm, to be funny, or to speak negatively) would influence a listeners interpretation of potentially ironic remarks. There is, however, only a small amount of
AUTHORS NOTE: This research was supported by a grant from the Natural Science and Engineering Research Council of Canada to PMP. We thank Susan Graham for helpful discussions about the analysis of Experiment 3 and Stacey Ivanko and Jodi Edwards for assistance with testing participants. In addition, two anonymous reviewers provided very helpful comments about an earlier version of this article.
JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY, Vol. 21 No. 3, September 2008 245-274 2008 Sage Publications

245

246

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

experimental evidence to support this possibility, and not all theories of irony would make such a prediction. The goal of the present article was to investigate the extent to which ironic interpretation may be cued by such beliefs about a speaker. Our focus is the interpretation of potentially ironic remarks (i.e., verbal, and not situational, irony). Our approach is motivated by interest in the interaction of the cognitive and social processes involved in communication. As argued by Fussell and Kreuz (1998; see also Gernsbacher, 1994), any act of message interpretation is based on both psycholinguistic processes (e.g., lexical retrieval, syntactic processing) and social-interactional factors such as beliefs about what a speaker is trying to achieve by his or her message (p. 3). In the present research, we investigated when and how these cognitive and social processes may interact in the context of understanding irony. Inferring speaker intent for a potentially ironic remark involves basic language comprehension processes; it also involves social factors such as relationship variables (Slugoski & Turnbull, 1988) and, perhaps, beliefs about the speaker. The potential influence of these and other social factors has not been addressed in most studies of verbal irony.

SPEAKER CHARACTERISTICS AS CUES TO IRONY The possibility that speaker characteristics might cue ironic intent has been addressed indirectly in some of the previous literature. For instance, Leggitt and Gibbs (2000) found that speakers who made sarcastic statements were perceived as feeling more angry, disgusted, and scornful than speakers who made nonsarcastic statements. Toplak and Katz (2000) reported that speakers who used a sarcastic criticism were perceived as being more verbally aggressive and offensive than speakers who used a literal criticism (see also Colston, 1997; see Dews & Winner, 1995, for a contrasting view, and Pexman & Olineck, 2002, for some resolution to this debate). Kreuz, Long, and Church (1991) reported that ironic statements fulfilled different communication goals than literal statements; in particular, ironic statements were used to be funny and to mock someone. Similarly, Roberts and Kreuz (1994) found that irony was used to express negative emotion. It is not clear from these findings, however, whether by manipulating these characteristics for the speaker (e.g., giving listeners knowledge that the speaker is angry or not angry), one could directly affect the likelihood that sarcasm would be perceived. This type of manipulation would provide a strong test of whether certain speaker characteristics are salient cues to ironic intent.

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

247

Slugoski and Turnbull (1988) manipulated aspects of the relationship between the speaker and listener and examined the effects on interpretation of potentially ironic remarks. They manipulated the closeness of the relationship between the speaker and listener and also the extent to which the speaker and listener liked each other. Liking was found to be a strong cue. If the speaker and listener liked each other, then literal insults were more likely to be interpreted as ironic compliments. If the speaker and listener disliked each other, then literal compliments were more likely to be interpreted as ironic insults. Speaker characteristics (conveyed by speaker occupation information) were directly manipulated by Katz and Pexman (1997). Their stimuli were metaphors such as children are precious gems that can be interpreted as metaphors (e.g., intended meaning is that children are valuable) but can also be interpreted ironically (e.g., intended meaning is that children are not valuable). They presented these statements following short written contexts with some of the speakers identified by occupation. Half of the speakers identified by occupation were members of high-irony occupations (e.g., comedian, cab driver, etc., according to ratings in a pilot study), and half were members of highmetaphor occupations (e.g., clergyman, scientist, etc.). Results showed that statements made by speakers from high-irony occupations received higher sarcasm ratings than statements made by speakers from high-metaphor occupations, particularly if other cues to ironic intent (e.g., incongruity between the positive or negative tone of the context and the negative or positive tone of the statement) were absent. Statements made by speakers from high-irony occupations were also more likely to be remembered in a subsequent recall task. This recall advantage for statements made by high-irony speakers was attributed to participants coding occupation in the comprehension process and carrying out particularly effortful processing for statements made by high-irony speakers to consider the possibility that a sarcastic message was intended. In support of this explanation, there is evidence that more effortful processing leads to a stronger memory trace (e.g., Gibbs, 1987; OBrien & Myers, 1985). Thus, speaker occupation stereotypes influenced interpretation of, and memory for, potentially ironic remarks. In an online reading task with the same stimuli, Pexman, Ferretti, and Katz (2000) found that when speaker occupation was mentioned, reading times were longer at the last word in the target metaphor. These longer reading times at statements end (and the absence of longer reading times for words after the statement) were taken to mean that speaker occupation information was integrated with other types of information (e.g., lexical, syntactic) as soon as the last word in the statement had been read. Based on the reading-time data, the authors concluded that this integration occurred relatively early in end-of-statement processing.

248

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

EFFECTS OF SPEAKER STEREOTYPES AND THEORIES OF IRONY No theory of irony makes explicit predictions about how speaker stereotypes should influence interpretation of irony. Nonetheless, for most theories, it is possible to derive predictions based on the tenets of the theories. In the present research, we investigated the extent and nature of speaker occupation stereotype effects to establish whether (and in what ways) theories of irony would need to be modified to account for such effects. According to the echoic mention theory (Sperber & Wilson, 1981, 1986), irony is made possible by an utterance that implicitly or explicitly echoes a previous event, norm, or expectation. Similarly, echoic reminder theory (e.g., Kreuz & Glucksberg, 1989) holds that the echoic nature of an ironic utterance reminds the listener of a violated norm or failed expectation, and as such expresses the ironic speakers negative attitude about a given situation. The Echoic Reminder view explains the asymmetry that is typically observed in irony, whereby positive statements are more readily interpreted as irony (as ironic insults) than are negative statements (as ironic compliments). This asymmetry arises because, even if there is no explicit antecedent, positive statements always echo the implicit norm whereby speakers are expected to make positive comments (e.g., if you cant say something nice dont say anything at all). Negative statements, in contrast, do not echo this implicit norm and thus seem to require an explicit antecedent (e.g., Gibbs, 1986; Hancock, Dunham, & Purdy, 2000; Kreuz & Glucksberg, 1989). Thus, according to these echoic theories, detection of irony depends on the presence of an implicit or explicit antecedent. There is no obvious reason why speaker occupation stereotypes would interact with the echoic mention or reminder, and there seems no reason why a model of this type would predict that speaker occupation should influence interpretation of potentially ironic utterances. Pretense models (e.g., Clark & Gerrig, 1984) posit that an ironic speaker pretends to address an imaginary listener who would likely take their utterance literally. The ironic speaker pretends to be a very optimistic person, perceiving events as if through rose-colored glasses. The result is that the ironic speaker expresses a negative attitude toward the situation or idea described, the imaginary listener, and the pretended optimist. By this account, occupation stereotypes could affect comprehension of irony if an occupation were strongly consistent, or strongly inconsistent, with this type of behavior. For instance, if the speakers occupation stereotype was consistent with jocularity, or perhaps with insincerity, then that might make a listener more likely to detect the speakers pretense. Conversely, if the speakers occupation stereotype was consistent with sincerity, and seriousness, then the listener might be less likely to detect the speakers pretense.

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

249

The allusional pretense theory (Kumon-Nakamura, Glucksberg, & Brown, 1995) involves two main arguments: (a) Ironic utterances allude to a failed expectation (i.e., by implicit or explicit echo) and (b) ironic utterances involve pragmatic insincerity (i.e., by violating felicity conditions for speech acts, e.g., degree of politeness should be appropriate to the situation). As with pretense theory, it seems reasonable to assume that speaker stereotype information could enhance (or attenuate) the perception of pragmatic insincerity, and so allusional pretense theory could predict that speaker occupation influences the perception of irony. Colston (2000) provided a revision to the allusional pretense theory. That is, Colston argued that the necessary conditions for verbal irony are (a) portraying a contrast between expectations and reality and (b) flouting of Gricean Conversational Maxims (e.g., Maxims of Quality, Quantity, Relevance, or Manner). It seems possible that speaker stereotype information could provide insight about speaker expectations and thus accentuate the contrast necessary for verbal irony. For instance, if speaker occupation stereotypes were consistent with negative expectations, then detection of ironic intent might be enhanced. In addition, stereotypes that suggest insincerity and humor might facilitate detection of the flouting condition. Recently, Utsumi (2000) argued that many of the previous theories have failings: None of the previous irony theories can distinguish ironic utterances from nonironic ones completely (p. 1778). Utsumi argued that echoic mention (or reminding) and pretense are certainly important factors in perceiving irony but that previous theories have failed to find common properties present in all instances of irony and absent in all instances of nonirony. For instance, echoic theories fail to explain instances of irony that can be understood as such but are not at all echoic and fail to account for the effects of various contextual cues to ironic intent (both verbal and nonverbal). Furthermore, Utsumi questioned the adequacy of pretense theories, citing examples of nonirony that involve pretense. Pretense does not seem to be a sufficient condition for irony. Instead, Utsumi proposed the implicit display theory of verbal irony. The theory involves three main claims: First, an ironic utterance assumes an ironic environment, defined as a situational setting which motivates verbal irony (p. 1778). Second, an ironic utterance implicitly displays the ironic environment. Third, irony is a prototype-based category, so that utterances with more characteristics of implicit displays will be perceived as more ironic. Utsumi (2000) specified that the ironic environment involves the speakers having an expectation, the expectation not being met, and, most important for present purposes, the speaker having a negative emotional attitude toward the incongruity between expectation and outcome. Each of these components of the ironic environment must be expressed indirectlydirect expression would produce nonirony. Utsumi

250

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

noted that negative speaker attitude could be implicitly displayed by a number of cues, including verbal cues such as intonation, exaggeration, and the use of adjectives (e.g., really) and nonverbal cues such as gesture and facial expression. Utsumi did not address the possibility that these contextual cues might include beliefs about speakers. Nonetheless, it is seems likely that, by this model, speaker occupation would be predicted to have an effect on perceived irony. In particular, the implicit display theory emphasizes that ironic utterances indirectly express the speakers negative attitude toward the situation. The utterance conveys the extent to which the situation failed to meet the speakers expectations. If a speakers occupation helps convey their potentially negative attitude, then it should facilitate irony detection. This implicit display account holds that irony is understood more readily when the utterance is similar to the prototypical verbal irony (typicality effects). A speaker occupation stereotype that is consistent with ironic speech may facilitate irony detection by making an utterance seem more similar to that prototype. The theory seems to predict that speaker occupation stereotypes have an influence on perceived irony because they convey a negative attitude. As such, the key dimensions of those stereotypes should be traits such as cynicism and a critical outlook. Based on the implicit display theory, we also derived the specific prediction that occupation stereotypes are most likely to be relevant to potentially ironic utterances when the stereotype signals a person who tends to be critical but who tends to express that criticism indirectly. The speaker may choose to express the criticism indirectly because they tend to be humorous. The speaker may also be motivated to express their criticisms indirectly because of a concern about offending the addressee. That is, they may be in a position in which they have concerns about face saving. Irony certainly serves a face-saving function (Jorgensen, 1996). As Glucksberg (1995) suggested, When people either insult or compliment others, face-saving strategies are often used to offset the direct impact of such acts. Irony can be one such strategy (p. 51). Thus, current theories of irony can be extended to predictions about the likelihood of speaker occupation acting as a cue to speaker intent. In the present research we tested those predictions.

THE PRESENT RESEARCH Although it appeared that speaker occupation was a salient cue to ironic intent in the Katz and Pexman (1997) and Pexman et al. (2000) studies, there were several unresolved issues. In the present research, we addressed three of those issues to investigate the extent, and the

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

251

nature, of speaker occupation effects. The first issue was whether effects of occupation would be observed for stimuli that were not metaphors. In the Katz and Pexman (1997) and Pexman et al. (2000) studies, all of the target statements were metaphors. Metaphors used in an ironic sense are not the most typical form of verbal irony. For the metaphoric statements, as somewhat unconventional examples of irony, context cues may have been especially salient. In the present research, we examined the influence of occupation when statements were of the more conventional literal form, as in you are a wonderful friend. The second unresolved issue was the impact of speaker occupation on interpretation of ironic compliments (e.g., you are a terrible friend said when your friend has just done something gracious). In the Katz and Pexman (1997) and Pexman et al. (2000) studies, the target stimuli could be interpreted as ironic insults or literal compliments. Ironic compliments are less common than ironic insults in everyday speech (Gibbs, 2000). Furthermore, Kreuz and Glucksberg (1989) reported that ironic compliments were more difficult to interpret than ironic insults. Similarly, Pexman and Olineck (2002) found that participants were less certain of their interpretations for ironic compliments than they were for ironic insults. Because ironic compliments can be more difficult to interpret, speaker occupation might be a particularly salient cue to ironic intent in such cases. In the present research, we investigated this possibility. The third unresolved issue was why speaker occupation cued ironic intent. Which aspect or aspects of occupation stereotypes influence judgments of speaker intent? There has been substantial research on gender typing and occupations but very little research on other aspects of occupation stereotypes. One exception was a study by Slusher and Anderson (1987), in which the authors examined stereotypical traits of lawyers, artists, and clergymen. They found that these occupations were associated with particular traits: wealthy and aggressive lawyers, temperamental and creative artists, and kind and friendly clergymen. They also found that, when occupation was mentioned for a target person in a sentence context, participants imagined characteristics for the target person as a function of the interaction of those traits and the context. For instance, for the sentence George, a lawyer, is shopping for a car, their participants were more likely to imagine that the car was a fancy car (as compared to a situation in which the occupation was artist) because, they argued, the trait wealthy was activated and interacted with the activity (car shopping) described. Thus, Slusher and Anderson found that occupation stereotypes influenced the imaginations people generated for target individuals, as a function of the context in which the target individual was described. To our knowledge, occupation stereotypes have not been examined in the context of communication. It is clear, however, from the ratings

252

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

obtained by Katz and Pexman (1997) that people perceive members of certain occupations as likely to use irony and members of other occupations as unlikely to use irony. What is unclear is which aspects of occupation stereotypes are related to these perceptions. As mentioned, research has shown that speakers who make sarcastic statements are perceived to be more scornful, and more verbally aggressive, than speakers who make nonsarcastic statements (Leggitt & Gibbs, 2000; Toplak & Katz, 2000). In the present study, we determined whether these or other aspects of occupation stereotypes are related to use of sarcasm. Based on the predictions we derived from pretense theories, the relevant aspects of the occupation stereotypes should be sincerity and, perhaps, tendency to use humor. Based on Colstons contrast construct, we also added tendency to have negative expectations (cynicism) to this list. Based on the predictions we derived from the implicit display theory, the relevant aspects of the occupation stereotypes should be tendency to be critical and tendency to express criticism indirectly, perhaps manifest as tendency to use humor, or low social status (which could prompt a need for face saving). To summarize, in this study, we first investigated whether speaker occupation cued ironic intent for statements that could be interpreted literally (Experiments 1 and 2) and then examined what it is about occupation stereotypes that contributes to a perception of irony (Experiment 3).

EXPERIMENT 1
METHOD

Participants Participants in Experiment 1 were 48 undergraduate students (41 female) who received bonus credit in a psychology course in exchange for their participation. Twenty-four versions of the stimuli were used, and participants were assigned to version by their order of appearance at the lab, such that Participant 1 was assigned to Version 1, Participant 2 to Version 2, and so on. Materials In this experiment, there were three paper-and-pencil tasks: a ratings task, a distraction task, and an unexpected free recall memory task. There were 24 items on the ratings task, and each item involved a three-sentence context followed by a target statement (for examples, see Table 1). These items were selected from a pilot study.

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

253

Table 1 Sample Item (Experiment 1) Context Strongly positive Laura and her friend, a scientist, were having coffee at a local caf. They were talking about a romantic date that Laura had been on the night before. Laura and her date had gone for dinner and dancing at a new club downtown. The scientist said: Laura and her friend, a scientist, were having coffee at a local caf. They were talking about a pleasant date that Laura had had the night before. The scientist said: Laura and her friend, a scientist, were having coffee at a local caf. They were talking about a boring date that Laura had had the night before. The scientist said: Laura and her friend, a scientist, were having coffee at a local caf. They were talking about a boring date that Laura had had the night before. They had watched old reruns of cartoons all evening. The scientist said: Positive Statement That sounds pretty exciting

Weakly positive

Weakly negative

Negative

That sounds pretty dull

Strongly negative

Note. The contexts illustrate a nonsarcastic occupation version of the contexts. In a sarcastic occupation version, the scientist would be replaced with a sarcastic occupation (e.g., a comedian).

Pilot study. In the pilot study, 40 participants were asked to rate a set of 45 potential occupations, 38 were asked to rate a set of potential statements, and 48 were asked to rate a set of potential contexts. All of these participants were undergraduate students who did not participate in the subsequent experiments. For the occupation ratings, participants were told to imagine, for each occupation in a list, that a member of that occupation got a flat tire on the way to work. They then rated the likelihood (on a 7-point scale) that the member of that occupation would make a sarcastic remark about the situation. For the experiment proper, we selected the eight occupations with the highest likelihood ratings (comedian, talk show host, movie critic, journalist, cab driver, truck driver, actress, salesman) as our sarcastic occupations (M = 5.41, SD = 0.64), and the eight occupations with the lowest likelihood ratings (army sergeant, accountant, doctor, clergyman, scientist,

254

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

librarian, mortician, dentist) as our nonsarcastic occupations (M = 2.92, SD = 0.27, t(14) = 10.16, p < .001, SE = 0.25). Mean sarcasm ratings for all 45 occupations are listed in the appendix. The potential statements we devised each involved a positive and a negative version (e.g., that sounds pretty exciting, that sounds pretty dull). These were presented on paper, in random order, and participants were asked to rate each statement on a 7-point scale for (a) how likely it was that the statement would be made in real conversation, (b) how familiar the statement was, and (c) how positive the statement was. For the experiment proper, we selected 24 pairs of statements, for which the positive versions of the statements were rated to be significantly more positive (M = 5.53, SD = 0.61) than the negative versions of the statements (M = 2.04, SD = 0.53, t(23) = 17.98, p < .001, SE = 0.21). As much as possible, within the pairs, we also tried to match likelihood (M = 5.40, SD = 0.80, for positive versions; M = 5.04, SD = 0.80, for negative versions) and familiarity ratings (M = 4.80, SD = 0.85, for positive versions; M = 4.15, SD = 0.84, for negative versions). The potential contexts we devised were in sets of four, including a strongly positive version, a weakly positive version, a weakly negative version, and a strongly negative version. These were presented on paper, in random order, and participants were asked to rate each context scenario on a 7-point scale for (a) how likely it was that the scenario would really happen and (b) how positive the scenario was. We selected 24 sets of contexts, for which mean positivity ratings for the strongly positive contexts (M = 6.32, SD = 0.31) were significantly higher than for the weakly positive contexts (M = 5.27, SD = 0.66, t(23) = 8.86, p < .001, SE = 0.12), mean positivity ratings for the weakly positive contexts were significantly higher than for the weakly negative contexts (M = 3.43, SD = 0.86, t(23) = 9.25, p < .001, SE = 0.20), and mean positivity ratings were significantly higher for the weakly negative contexts than for the strongly negative contexts (M = 2.28, SD = 0.69, t(23) = 7.78, p < .001, SE = 0.15). As much as possible, we chose contexts in such a way as to equate likelihood ratings across context conditions (for strongly positive: M = 5.53, SD = 0.53; for weakly positive: M = 5.55, SD = 0.31; for weakly negative: M = 5.32, SD = 0.51; for strongly negative: M = 4.99, SD = 0.59). The materials for the experiment were 24 pairs of statements and 24 sets of contexts involving a 2 (context strength: strong or weak) by 2 (context positivity: positive or negative) by 2 (statement positivity: positive or negative) by 3 (occupation: sarcastic occupation, nonsarcastic occupation, or no occupation) combination of conditions. Each statement appeared in each of the four contexts, with each of the three types of occupation, across the 24 versions of the experiment. Thus, one third of the items were presented in the no occupation condition. This condition was devised to create filler items. That is, we did not want to mention occupation in every one of the contexts because we thought that

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

255

might lead participants to adopt a response strategy in which they anticipated occupation information in each context. The stimuli were presented in booklets. At the top of each page in the booklet, we printed a context and target statement (the target statement was presented in bold) and then four rating scales. The four rating scales were as follows: (a) Is the speaker being sarcastic? (1 = not at all sarcastic, 7 = extremely sarcastic), (b) Is the speaker saying something polite? (1 = not at all polite, 7 = extremely polite), (c) Is the speaker mocking someone? (1 = not at all mocking, 7 = extremely mocking), and (d) How certain are you that you correctly interpreted the speakers intent? (1 = not at all certain, 7 = extremely certain). The distraction task involved six arithmetic problems. On the free recall task, participants were asked to try to remember the target statements, which had been presented in boldface. Procedure Testing for this experiment took place in a group setting with five to seven participants tested at a time. Participants first completed the ratings task and then spent approximately 10 minutes on the distraction task. Finally, the unexpected free recall task was completed.
RESULTS

There were five dependent measures in this experiment: four ratings dimensions and the free recall data. In Table 2, we presented the correlations among the dependent measures separately for the positive statements (above the diagonal) and the negative statements (below the diagonal). The patterns of correlations for these two types of statements are somewhat different because the positive statements are potentially ironic insults and the negative statements are potentially ironic compliments, and these two forms of verbal irony have slightly different social functions (e.g., Dews & Winner, 1995; Pexman & Olineck, 2002). These results, however, are tangential to the central issue of the present work. Means and standard deviations for all of the dependent variables are presented in Table 3. To address our central question of the influence of speaker occupation as a cue to sarcastic-ironic intent, we conducted analyses of all of the conditions in which occupation was mentioned with 2 (context strength: strong or weak) by 2 (context positivity: positive or negative) by 2 (statement positivity: positive or negative) by 2 (occupation: sarcastic occupation or nonsarcastic occupation) ANOVAs. In all analyses, the data were analyzed by subjects and, separately, by items (F1 and F2, respectively).

256

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

Table 2 Intercorrelations for Ratings Dimensions and Free Recall Data in Experiment 1 Measure 1. Sarcasm ratings 2. Mocking ratings 3. Politeness ratings 4. Certainty ratings 5. Free recall data 1 .10* .20** .06 .05 2 .75** .31** .01 .01 3 .66** .63** .01 .04 4 .08 .04 .05 .04 5 .09 .07 .14** .03

Note. Intercorrelations for negative statements are presented below the diagonal, and intercorrelations for positive statements are presented above the diagonal. df= 384. *p < .05. **p < .01

Sarcasm ratings. For ratings of the extent to which a speaker was being sarcastic, there was a significant interaction of Context Strength Context Positivity Statement Positivity, F1(1, 47) = 33.70, p < .001, MSE = 3.23, and F2(1, 23) = 15.36, p < .001, MSE = 3.42, and a significant interaction of context positivity and statement positivity, F1(1, 47) = 412.73, p < .001, MSE = 4.18, and F2(1, 23) = 271.59, p < .001, MSE = 3.20, because sarcasm ratings were highest when contexts were strongly negative or positive and were incongruent with statement positivity. This effect demonstrates that incongruity between context and statement can produce a perception of sarcastic intent and replicates this finding in the past literature (e.g., Katz & Lee, 1993; Katz & Pexman, 1997; Pexman & Olineck, 2002). There was also a main effect of context positivity, F1(1, 47) = 31.48, p < .001, MSE = 3.31, and F2(1, 23) = 18.60, p < .001, MSE = 1.26, because sarcasm ratings were higher for statements presented in negative contexts (potentially ironic insults). This difference is likely due to the fact that ironic insults are the more common form of sarcasm (Gibbs, 2000). The difference in conventionality for ironic insults and ironic compliments likely also explains the main effect of statement positivity, F1(1, 47) = 37.16, p < .001, MSE = 3.98, and F2(1, 23) = 16.44, p < .001, MSE = 4.37: Sarcasm ratings were higher for positive statements (potentially ironic insults). Notably, there were no significant effects involving speaker occupation, so this variable did not affect ratings of sarcasm for the statements presented in this experiment. Mocking ratings. For ratings of the extent to which the speaker was mocking someone, there was an interaction of Context Strength Context Positivity Statement Positivity, F1(1, 47) = 8.01, p < .001, MSE = 3.14, and F2(1, 23) = 7.89, p < .01, MSE = 1.64, and a significant interaction of context positivity and statement positivity, F1(1, 47) = 92.14, p < .001, MSE = 2.93, and F2(1, 23) = 46.50, p < .001, MSE = 2.55, because mocking ratings were highest when contexts were strongly negative or positive and incongruent with statement positivity. There was also a

Table 3 Ratings and Recall Percentages for Conditions in Experiment 1 Strongly Positive Context Positive Negative Statement, Statement, Literal Ironic Compliment Compliment Measure Sarcasm ratings Occupation Condition Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation M 2.17 2.19 2.29 2.08 5.38 5.52 5.60 5.62 43.8 33.3 SD 1.55 1.81 1.70 1.80 1.45 1.58 1.33 1.39 50.1 47.6 M 5.31 5.04 3.56 3.44 3.69 3.69 5.23 5.31 35.4 43.8 SD 1.84 2.19 1.89 1.89 1.85 1.74 1.32 1.27 48.3 50.1 Weakly Positive Context Positive Statement, Literal Compliment M 3.06 2.94 3.02 2.50 4.85 5.15 5.40 5.48 45.8 55.9 SD 1.91 1.99 1.93 1.88 1.66 1.57 1.25 1.43 50.4 42.5 Negative Statement, Ironic Compliment M 4.25 4.23 3.75 3.04 3.08 3.56 5.04 4.73 25.0 45.8 SD 2.18 2.21 2.15 1.86 1.65 1.61 1.44 1.59 43.8 50.4 Strongly Negative Context Positive Statement, Ironic Insult M 6.31 6.44 4.94 4.88 3.35 3.31 5.69 5.92 43.8 43.8 SD 1.45 1.13 2.12 2.01 1.88 1.87 1.29 1.38 50.1 50.1 Negative Statement, Literal Insult M 2.02 1.73 3.06 3.15 3.06 3.08 5.81 5.69 25.0 39.6 SD 1.73 1.25 2.13 2.06 1.62 1.71 1.10 0.97 43.8 49.4 Weakly Negative Context Positive Statement, Ironic Insult M 5.87 5.65 4.79 4.54 3.29 3.10 5.83 5.60 35.4 41.7 SD 1.67 1.73 1.90 2.21 1.77 1.61 1.08 1.30 48.3 49.8 Negative Statement, Literal Insult M 2.42 2.60 3.94 3.37 2.71 2.98 5.40 5.33 35.4 29.2 SD 1.82 1.81 2.08 1.94 1.38 1.45 1.25 1.51 48.3 45.9

Mocking ratings

Politeness ratings

Certainty ratings

Recall 257

258

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

main effect of context positivity, F1(1, 47) = 51.15, p < .001, MSE = 4.69, and F2(1, 23) = 35.31, p < .001, MSE = 3.27, because mocking ratings were higher for statements presented in negative contexts. Notably, there was an interaction of context strength and speaker occupation, F1(1, 47) = 3.97, p < .05, MSE = 2.89, and F2(1, 23) = 4.05, p < .05, MSE = 1.89. The nature of this interaction was that speaker occupation had a greater impact on mocking ratings for statements presented in the weak context conditions than for statements presented in the strong context conditions. There was also a main effect of speaker occupation, F1(1, 47) = 6.43, p < .05, MSE = 2.50, and F2(1, 23) = 4.16, p < .05, MSE = 1.93, such that mocking ratings were higher for statements made by sarcastic speakers than for statements made by nonsarcastic speakers. The fact that there was not a significant interaction between speaker occupation and context positivity suggested that occupation was an equally salient cue to speaker intent for statements that were potentially ironic insults and for statements that were potentially ironic compliments. Thus, the hypothesis that occupation might be a particularly salient cue for statements that were potentially ironic compliments was not supported. Politeness ratings. For ratings of the extent to which the speaker was saying something polite, there was a significant interaction of context positivity and statement positivity, F1(1, 47) = 32.12, p < .001, MSE = 2.98, and F2(1, 23) = 19.77, p < .001, MSE = 2.42, such that politeness ratings were highest for positive statements presented in positive contexts and lowest for negative statements presented in negative contexts. There were also main effects of context strength, F1(1, 47) = 6.06, p < .05, MSE = 2.74, and F2(1, 23) = 7.06, p < .05, MSE = 1.18; context positivity, F1(1, 47) = 111.65, p < .001, MSE = 2.70, and F2(1, 23) = 46.73, p < .001, MSE = 3.22; and statement positivity, F1(1, 47) = 41.20, p < .001, MSE = 4.78, and F2(1, 23) = 27.80, p < .001, MSE = 3.54: Politeness ratings were highest for statements presented in strong contexts, statements presented in positive contexts, and positive statements. Notably, there were no significant effects involving speaker occupation, so this variable did not affect politeness ratings for the statements presented in this experiment. Certainty ratings. Participants ratings of how certain they were of their interpretations were generally quite high, and analysis of those ratings produced only three significant main effects. There was a main effect of context strength, F1(1, 47) = 8.82, p < .005, MSE = 1.42, and F2(1, 23) = 4.67, p < .05, MSE = 1.37, such that participants were more certain of their interpretations when statements were presented in the strongly biasing contexts. There was also a main effect of context positivity, F1(1, 47) = 13.27, p < .001, MSE = 1.87, and F2(1, 23) = 9.75, p < .005, MSE = 1.25, such that participants were more certain of their

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

259

interpretations for statements presented in negative contexts. Finally, there was a main effect of statement type, F1(1, 47) = 13.10, p < .001, MSE = 1.53, and F2(1, 23) = 12.16, p < .005, MSE = 0.84: Participants were more certain of their interpretations for positive statements. Recall. In scoring the free recall data, participants were given one point for each of the 24 items that they recalled correctly. To be scored as correctly recalled, the statements had to be remembered almost verbatim. That is, remembered statements had to preserve the literal meaning of the original statement, and a maximum of two errors in wording were permitted. For recall data, the four-way interaction of Context Strength Context Positivity Statement Positivity Speaker Occupation was significant, F1(1, 47) = 6.90, p < .05, MSE = 1275.49, and F2(1, 23) = 4.38, p < .05, MSE = 1004.30. The nature of this interaction was revealed with simple main effects tests. For statements presented in positive contexts (which were potentially ironic compliments), there was a significant three-way interaction of Context Strength Statement Positivity Speaker Occupation in the subjects analysis, F1(1, 47) = 4.06, p < .05, MSE = 1444.48, and F2(1, 23) = 3.14, p = .09, MSE = 597.83, and a significant two-way interaction of Statement Positivity by Speaker Occupation, F1(1, 47) = 6.62, p < .05, MSE = 2867.35, and F2(1, 23) = 15.27, p < .001, MSE = 767.66, because recall percentages were highest for positive statements made by sarcastic speakers, particularly in the weakly positive context. For nonsarcastic speakers, memory was best for negative statements in the strongly positive context and for positive statements in the weakly positive context. Neither of these interactions was significant for statements presented in negative contexts (which were potentially ironic insults). The fact that speaker occupation interacted with these other factors only when the statements were potentially ironic compliments (and not ironic insults) lends some support to our hypothesis that speaker occupation might be a particularly salient cue for ironic compliments. Analysis of the recall data also showed a significant three-way interaction of Context Positivity Statement Positivity Speaker Occupation, F1(1, 47) = 5.42, p < .05, MSE = 1883.87, and F2(1, 23) = 6.30, p < .05, MSE = 810.69. The nature of this interaction was that for statements in positive context (potentially ironic compliments), speaker occupation seemed to influence recall percentages only for negative statements, with lower recall for statements made by sarcastic speakers. For statements in negative context, however, speaker occupation seemed to have the same small effect on recall percentages for both positive and negative statements, with slightly lower recall percentages for statements made by sarcastic speakers. There was also an interaction of statement positivity and speaker occupation that approached significance by subjects and was significant by items, F1(1, 47) = 3.26, p = .08, MSE = 2871.23, and F2(1, 23) = 13.38, p < .001, MSE

260

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

= 498.19. For statements made by sarcastic speakers, recall percentages were higher for positive statements than for negative statements, but for statements made by nonsarcastic speakers, there was a much smaller difference in recall for the two types of statements. Occupation effects in the recall data suggested that participants processed incongruencies between speaker occupation stereotypes and the tone of statements and contexts. Also, speaker occupation did influence overt ratings of speaker intent, at least in terms of ratings of the extent to which the speaker was mocking someone. Speaker occupation did not, however, have a significant effect on ratings of perceived sarcasm. Thus, it appears that, for the types of statements presented here, occupation stereotypes influence perceptions of speaker intent but do not add significantly to perceived level of sarcasm. Participants seemed to assess stereotype information in deciding how mocking a speaker intended to be but did not modify their sarcasm ratings as a function of the occupation stereotype. Perhaps mocking ratings involve a more direct assessment of speaker characteristics, whereas sarcasm ratings are also influenced by contextual factors like incongruity. Because Katz and Pexman (1997) observed a significant effect of speaker occupation on sarcasm ratings for their metaphor stimuli, it seems that occupation effects for sarcasm ratings may be limited to situations involving less conventional target statements than were presented here. Indeed, Katz and Pexman found effects of speaker occupation on sarcasm ratings for only their less familiar (and not for more familiar) metaphors. It is also possible, however, that it is not just statement ambiguity that leads to occupation effects; ambiguous context (even with a quite conventional statement) might cause occupation information to become a more influential cue to ironic intent. That is, if context information was minimal, speaker occupation might be a more salient cue to speaker intent and might influence sarcasm ratings. This notion seems to be supported by the fact that in the present Experiment 1, occupation effects for mocking ratings were stronger in the weak (as compared to strong) context conditions. This possibility was tested in Experiment 2.

EXPERIMENT 2
METHOD

Participants Participants in Experiment 2 were 48 undergraduate students (35 female) who received bonus credit in a psychology course in exchange for their participation. None of these students had participated in

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

261

Experiment 1. Six versions of the stimuli were used, and participants were assigned to version by their order of appearance at the lab, such that Participant 1 was assigned to Version 1, Participant 2 to Version 2, and so on. Materials Materials in Experiment 2 were a subset of the materials used in Experiment 1, with slight modification. The contexts were modified to remove information about the tone, or positive/negative outcome, of the events described. As such, there was only one type of context presented, which could be labeled neutral. The materials for the experiment were 12 pairs of statements involving a 2 (statement positivity: positive or negative) by 3 (occupation: sarcastic occupation, nonsarcastic occupation, or no occupation) combination of conditions. Each statement appeared with each of the three types of occupation across the six versions of the experiment. Procedure The procedure for this experiment was the same as that used in Experiment 1, except that only 12 items (rather than 24) were presented.
RESULTS

As in Experiment 1, there were five dependent measures in this experiment: four ratings dimensions and the free recall data. The correlations between these dependent measures are presented in Table 4. Means and standard deviations for these variables are presented in Table 5. The goal of the experiment was to examine the influence of speaker occupation as a cue to sarcastic-ironic intent. To address this issue, we conducted analyses of all of the conditions in which occupation was mentioned with 2 (statement positivity: positive or negative) by 2 (occupation: sarcastic occupation or nonsarcastic occupation) ANOVAs. Sarcasm ratings. For ratings of the extent to which a speaker was being sarcastic, there was, importantly, a significant main effect of speaker occupation, F1(1, 47) = 5.02, p < .05, MSE = 1.06, and F2(1, 11) = 3.88, p = .07, MSE = 0.34, such that sarcasm ratings were higher for statements made by speakers from sarcastic occupations. The only other significant effect was a main effect of statement positivity, F1(1, 47) = 21.62, p < .001, MSE = 1.78, and F2(1, 11) = 13.15, p < .005, MSE = 0.73, because sarcasm ratings were higher for positive statements.

262

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

Table 4 Intercorrelations for Ratings Dimensions and Free Recall Data in Experiment 2 Measure 1. Sarcasm ratings 2. Mocking ratings 3. Politeness ratings 4. Certainty ratings 5. Free recall data 1 .37** .15* .20** .06 2 .77** .24** .05 .06 3 .80** .72** .02 .01 4 .13* .12 .22** .08 5 .04 .08 .11 .09

Note. Intercorrelations for negative statements are presented below the diagonal, and intercorrelations for positive statements are presented above the diagonal. df= 288. *p < .05. **p < .01.

Table 5 Ratings and Recall Percentages for Conditions in Experiment 2 Neutral Context Positive Statement, Potentially Ironic Insult Measure Sarcasm ratings Mocking ratings Politeness ratings Certainty ratings Recall Occupation Condition Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation Sarcastic occupation Nonsarcastic occupation M 3.81 3.45 3.61 3.24 4.47 4.77 5.02 4.82 40.6 40.6 SD 2.21 1.98 2.09 2.02 1.88 1.72 1.56 1.74 49.4 49.4 Negative Statement, Potentially Ironic Compliment M 2.89 2.58 3.30 3.90 3.07 2.79 5.04 4.91 42.7 39.6 SD 1.87 1.64 1.88 1.80 1.51 1.44 1.44 1.41 49.7 49.2

Mocking ratings. For ratings of the extent to which a speaker was mocking someone, the only significant effect was an interaction of speaker occupation and statement positivity, F1(1, 47) = 8.45, p < .01, MSE = 1.33, and F2(1, 11) = 8.60, p < .05, MSE = 0.33. As illustrated by the pattern of means in Table 5, the nature of this interaction was that for positive statements, mocking ratings were somewhat higher when the speaker was from a sarcastic occupation, whereas for negative statements, mocking ratings were somewhat lower when the speaker was from a sarcastic occupation. Speakers from sarcastic occupations are more likely to be using both types of statements ironically than speakers from nonsarcastic occupations, resulting in positive statements that are interpreted as ironic insults and negative statements that are interpreted as ironic compliments.

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

263

Politeness ratings. For ratings of the extent to which a speaker was saying something polite, there was a significant interaction of speaker occupation and statement positivity by subjects only, F1(1, 47) = 4.41, p < .05, MSE = 0.93, and F2(1, 11) = 2.48, p = .14, MSE = 0.41, and a significant main effect of statement positivity, F1(1, 47) = 78.65, p < .001, MSE = 1.74, and F2(1, 11) = 24.52, p < .001, MSE = 1.39. The nature of the main effect was that positive statements were considered to be more polite than negative statements. Certainty ratings. Despite the minimal context provided in this experiment, participants ratings of how certain they were of their interpretations were reasonably high. In the analysis of these ratings, there was only one significant effect: a main effect of speaker occupation by items only, F1(1, 47) = 1.66, p = .20, MSE = 0.80, and F2(1, 11) = 5.44, p < .05, MSE = 0.06, such that certainty ratings were somewhat higher for statements made by members of sarcastic occupations. Because this effect was not significant by subjects, it was probably the case that whereas some participants showed this pattern in their ratings, others did not. Recall. The recall data were scored as in Experiment 1, and analysis of those data showed no significant effects. The results of Experiment 2 showed, most important, significant effects of speaker occupation for sarcasm ratings. Thus, participants used information about speaker occupation as a cue to sarcastic intent, rating target statements as more sarcastic when the speaker was a member of an occupation associated with sarcastic speech and less sarcastic when the speaker was a member of an occupation that was not associated with sarcastic speech. Speaker occupation also influenced perceived intent to mock someone and perceived politeness. These results suggest that speaker occupation stereotypes do cue ironic intent when there is little other contextual information provided. In Experiment 3, we investigated what it is about occupation stereotypes that cues such intent.

EXPERIMENT 3 The purpose of Experiment 3 was to investigate which aspects of occupation stereotypes may be related to perceived use of sarcasm. As such, we collected ratings for 45 occupations on dimensions that might be involved in the occupation stereotypes (and that would be relevant to perceived use of sarcasm). Based on previous research, we selected the following dimensions: (a) tendency to be humorous (e.g., Kreuz et al., 1991), (b) tendency to criticize (e.g., Colston, 1997; Kreuz et al., 1991), (c) tendency to be sincere (e.g., Clark & Gerrig, 1984; Haverkate,

264

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

1990; Kumon-Nakamura et al., 1995), (d) tendency to have close relationships (e.g., Slugoski & Turnbull, 1988), (e) tendency to be aggressive (e.g., Toplak & Katz, 2000), (f) education level, and (g) social status (Katz & Pexman, 1997).
METHOD

Participants Participants in Experiment 3 were 50 undergraduate students (41 female) who received bonus credit in a psychology course in exchange for their participation. None of these students had participated in Experiments 1 or 2. Materials In this experiment, the materials consisted of a booklet on which 45 occupations were listed. Under each occupation, seven rating scales were presented, one for each of the dimensions of interest: (a) likelihood that a person with this occupation would use humor during daily interactions (1 = not at all likely, 7 = very likely), (b) likelihood that a person with this occupation would express criticisms during daily interactions (1 = not at all likely, 7 = very likely), (c) level of sincerity that you would attribute to a person with this occupation (1 = low, 7 = high), (d) how typical it would be for a person with this occupation to have close relationships (1 = not at all typical, 7 = very typical), (e) how aggressive you would expect a person with this occupation to be (1 = not at all aggressive, 7 = very aggressive), (f) level of education that you would attribute to a person with this occupation (1 = low, 7 = high), and (g) level of perceived social status of a person with this occupation (1 = low, 7 = high). Sarcasm ratings for the same 45 occupations had been collected in the pilot study preceding Experiment 1. Procedure Testing for this experiment took place in a group setting with 5 to 7 participants tested at a time.
RESULTS

Mean ratings for the 45 occupations are listed in the appendix. Correlations among the variables in this experiment are displayed in Table 6. As illustrated in Table 6, several of the predictor variables were significantly correlated with sarcasm ratings. The only variables not significantly correlated with tendency to use sarcasm were tendency

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

265

Table 6 Correlations for Sarcasm Ratings and Predictor Variables Across 45 Occupations in Experiment 3 Variable Sarcasm ratings Predictor variable 1. Humor 2. Criticism 3. Sincerity 4. Close relationships 5. Aggressive 6. Education 7. Social status *p < .05. **p < .01. 1 .61** 2 .34* .04 3 .55** .06 .28 4 .19 .41** .33* .60** 5 .32* .11 .64** .53** .54** 6 .40** .23 .35* .38* .06 .16 7 .23 .09 .49** .25 .00 .37* .86**

to have close relationships and perceived social status. In addition, the intercorrelations between several of the predictor variables were quite high. For the regression analysis, to avoid multicollinearity, we removed three predictors that were significantly correlated with other predictors and had the lowest correlations with sarcasm ratings: social status (which had a strong positive correlation with education level), tendency to have close relationships (which had a strong positive correlation with sincerity and a strong negative correlation with tendency to be aggressive), and tendency to be aggressive (which had a strong positive correlation with tendency to criticize). Hence, the predictor variables were tendency to be humorous, tendency to criticize, tendency to be sincere, and education, all from ratings obtained in Experiment 3. The dependent variable was tendency to use sarcasm, from ratings obtained in the pilot study for Experiment 1. To determine which of these four predictor variables have significant, unique relationships with sarcasm ratings, all four predictors were entered in a standard multiple regression analysis. A summary of that regression analysis is displayed in Table 7. With all four predictors, the regression equation was significant, R2 = .74, F(4, 40) = 28.71, p < .001. Importantly, all four predictor variables had significant, unique relationships with tendency to use sarcasm. It is certainly possible that occupation stereotypes involve aspects other than those examined here. Nonetheless, the fact that 74% of the variance in sarcasm ratings is explained by the combination of humor, criticism, sincerity, and education level indicates that we have captured many of the critical components of these stereotypes. Thus, when thinking about different occupations, there is evidence that people can activate information about each of these characteristics. Whether they do activate such information will depend on the context. When the context is a situation in which a listener (or reader) is trying to understand

266

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

Table 7 Summary of Regression Analysis Predicting Sarcasm Ratings With Occupation Characteristics (n = 50) in Experiment 3 Variable Tendency to be humorous Tendency to criticize Tendency to be sincere Education level **p < .01. ***p < .001. B 0.57 0.45 0.32 0.17 SEB 0.09 0.12 0.11 0.06 .54 .39 .29 .30 t 6.53*** 3.94*** 2.92** 2.91** sr
2

.28 .10 .05 .05

the message conveyed by a potentially sarcastic remark (and when speaker occupation is provided), these characteristics will be activated. In other contexts, different characteristics would likely be relevant.

GENERAL DISCUSSION The purpose of this research was to investigate whether speaker occupation cued sarcastic-ironic intent for literal (rather than metaphoric) statements and, if so, what it was about occupation stereotypes that contributed to a perception of irony. Our results showed, in Experiment 1, that speaker occupation information did not significantly affect ratings of sarcasm. That is, sarcasm ratings were not significantly higher when a statement was made by a speaker from an occupation associated with sarcastic speech (e.g., comedian) than they were for the same statement made by a speaker from an occupation associated with nonsarcastic speech (e.g., clergyman). In that experiment, level of perceived sarcasm was not affected by occupation of the speaker. Speaker occupation did, however, influence mocking ratings, such that statements were considered to be more mocking when made by speakers from occupations associated with sarcastic speech. Speaker occupation also influenced memory for the target statements. This suggested that speaker occupation was coded in the comprehension process, although it did not influence sarcasm ratings. In Experiment 2, we investigated the possibility that speaker occupation might only cue sarcastic intent for literal statements when other contextual cues (e.g., positive/negative outcome of preceding events) were not available. This possibility was supported by the results of Experiment 2 because both positive and negative statements presented in neutral contexts were perceived to be more sarcastic when the speaker was a member of an occupation associated with sarcastic speech. Taken together, the results of Experiments 1 and 2 suggested that participants seemed to take multiple sources of information, including

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

267

speaker stereotypes, into consideration as they tried to determine what meaning speakers intended. There was evidence that speaker occupation was routinely integrated during comprehension of potentially ironic statements but only cued sarcastic intent when other cues were absent. If there was clear congruity or incongruity of context and statement (in Experiment 1), speaker occupation did not add significantly to perceived sarcasm. If context-statement congruity was ambiguous (in Experiment 2), then speaker occupation did add significantly to perceived sarcasm. An additional issue of interest for the present study was whether speaker occupation was a cue for both ironic insults and ironic compliments. One hypothesis was that because ironic compliments are less common and can be more difficult to understand, speaker occupation would be a particularly salient cue for ironic compliments. There was only weak support for this hypothesis: In Experiment 1, the recall data showed that speaker occupation had a stronger effect on memory for potentially ironic compliments than it did on memory for potentially ironic insults. The results of Experiment 3 provided an indication of the types of information that are activated by speaker occupation stereotypes in discourse context: perceived tendencies to be humorous, to criticize, to be sincere, and perceived education level. Members of occupations that are associated with sarcastic speech are perceived to be funnier, more critical, less sincere, and to have lower education levels. This information, when integrated with the statement and other aspects of context, shapes perceptions of speaker intent. We take our findings as evidence that, in the process of understanding speech acts, social variables are integrated with linguistic information. One might argue that this integration happens only when one is required to make an explicit decision about speaker intent, as in the present experiments. Recall, however, that in the Pexman et al. (2000) study, in which participants were not required to make any explicit decision about speaker intent (instead, they just read contexts and statements in a self-paced reading task), there was evidence that participants integrated speaker occupation information as soon as they had finished reading the last word in the target statement. The same was very likely the case in the present study. Our results included effects of speaker occupation stereotypes on interpretation of ironic intent. We should note, however, that there are certainly aspects of occupation stereotypes (e.g., stereotype strength, consistency, etc.) that we have not addressed and that likely influence these effects. It is also the case that occupation stereotypes are only one example of social knowledge that probably influences interpretation of ironic utterances. In the introduction, we derived predictions regarding speaker occupation effects from the various theories of verbal irony. We noted that

268

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

echoic theories (e.g., echoic mention and echoic reminder) seemed to predict null speaker occupation effects, whereas theories involving pretense (e.g., pretense theory and allusional pretense theory) would likely predict significant speaker occupation effects if the occupation cued level of sincerity and/or jocularity on the part of the speaker. From the recent implicit display theory, in contrast, we derived the prediction that speaker occupation can cue irony if the speaker occupation signals a tendency to be indirectly critical. Our results showed that speaker occupation stereotypes cue ironic intent and that those stereotypes involve information about a speakers tendencies to be humorous, critical, insincere, and to have lower levels of education. These findings are consistent with pretense theories because according to those theories, sarcastic speakers should be perceived to be joking and insincere, and speaker occupation stereotypes that are consistent with that perception should facilitate detection of ironic intent. Our findings are also consistent with Colstons (2000) revision of allusional pretense theory. Colston argued that detection of verbal irony requires a contrast between expectation and reality. Speaker occupation stereotypes may facilitate detection by providing information about speaker expectations. Colston provided evidence that for ironic compliments in particular, speaker expectations are perceived to be negative (and thus contrast with the positive context). That is, the perception of irony for an ironic compliment is created because the literal interpretation of the statement alludes to the speakers negative expectations, which were inferred by the listener. A stereotypic tendency to be critical may facilitate that inference. For an ironic insult, it seems likely that perception of ironic intent is heightened by the contrast between the speakers generally negative attitude (which, again, could be inferred from stereotypic traits) and the positive literal interpretation of their statement. This is consistent with our finding that a tendency to be critical was an important aspect of the sarcastic stereotype. Our findings are also consistent with predictions we derived from the implicit display theory (Utsumi, 2000). The occupation stereotype influences interpretation because it contributes to the ironic environment. It contributes to that environment by indicating that the speaker is likely to have a negative attitude (tendency to be critical) and that such an attitude is likely to be indirectly expressed (through humor and insincerity). This theory even explains the somewhat surprising relationship between education and tendency to use sarcasm. We found that the most sarcastic occupations tended to be those with lower perceived education levels and that education level was strongly associated with social status. The direction of this relationship is perhaps surprising but, we would argue, could potentially be explained by the implicit display theory. Speakers with a lower perceived education

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

269

level or social status (e.g., in this study, cab driver, truck driver, comedian) may use ironic insults to hedge the offense they might cause to higher status listeners. That is, because ironic insults serve a facesaving function, and lower status individuals may have a concern about directly insulting individuals with higher status, they may have a greater tendency to use ironic utterances. Higher status individuals are no less likely to be critical; indeed, there was actually a significant positive correlation (r = .35) between tendency to be critical and education in Experiment 3. That is, speakers with higher education levels were perceived to be relatively more critical. We argue, however, that higher status individuals are perceived to have fewer face-saving concerns and therefore are thought to be more likely than lower status individuals to express criticism directly. As a result, high status or high education level cues an expectation for direct criticism, whereas lower status or education level cues an expectation for indirect (ironic) criticism. This dissection of the relationship between tendency to use sarcasm and education level seems justified, given the significant and unique relationship observed in our regression analysis. Nonetheless, we do not suspect that education level by itself is a particularly useful cue to ironic intent. The source of the occupation effect is the combination of tendencies that are captured by occupation stereotypes: lower status, a tendency to express oneself with humor and insincerity, and a cynical outlook. Utsumi (2000) did not make explicit mention of how the implicit display theory could handle cases of ironic compliments. On one hand, this seems to be a problem for the implicit display theory because the theory holds that the ironic speaker has a failed expectation and also a negative emotional attitude (e.g., disappointment, anger, reproach, envy) toward the incongruity between what is expected and what is actually the case (p. 1783). Although, as mentioned, Colston (2000) found that speakers who made ironic compliments were perceived to have relatively negative expectations, the notion that those speakers have a negative emotional response toward the difference between their negative expectation and the positive outcome of events seems implausible. That is, it seems unlikely that speakers who choose to deliver praise with an ironic compliment are angry or disappointed that events had a more positive outcome than expected. On the other hand, Utsumi (2000) did mention in a footnote that there are negative attitudes of different degrees and types. Some are strong, and others are weak; some are serious but others are joking (p. 1783). This raises the possibility that when speakers deliver ironic compliments, they jokingly present a negative attitude (e.g., mock disappointment) about the incongruity between negative expectation and positive outcome. This seems a more plausible emotional response. In this way, it may be possible for the implicit display theory to account for this complimentary

270

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

form of verbal irony. This issue could be resolved in future research by investigating perceived emotions (e.g., mock disappointment vs. genuine disappointment) of speakers making ironic compliments.

CONCLUSION In the present study, we investigated one instance of the interaction between social knowledge and language comprehension processes. The results suggest that people have shared beliefs about the linguistic tendencies of different social groups and that those beliefs influence perceived communicative intent. To understand this interaction more generally, future research should address the impact of other social factors (e.g., age and gender stereotypes, authority, and expertise) on the interpretation of different forms of figurative language.

APPENDIX Mean Ratings and Standard Deviations for Occupations

Pilot Study (n = 40) Sarcasm Occupation Comedian Talk show hosta a Movie critic a Journalist a Cab driver a Truck driver a Actress a Salesman Army sergeantb b Accountant b Doctor b Clergyman b Scientist b Librarian Morticianb Dentistb Lawyer Teacher Police officer Politician Artist Factory worker Professional athlete
a

Experiment 3 (n = 50) Humor M 6.84 5.84 4.84 4.36 4.42 4.68 5.30 5.30 2.40 3.06 4.86 3.84 3.10 3.22 3.26 5.18 3.34 5.14 3.74 4.18 4.50 4.02 4.78 SD 0.55 0.91 1.08 1.06 1.36 1.38 1.04 0.95 1.20 1.24 0.93 1.39 1.27 1.36 2.05 1.10 1.15 0.93 1.43 1.32 1.25 1.46 1.15 M 5.62 5.60 6.42 5.28 4.18 4.50 5.26 3.72 6.34 4.54 4.68 3.40 5.12 4.14 3.24 4.82 6.04 5.00 5.52 5.84 4.30 4.46 4.54 Critic SD 1.28 1.12 0.67 1.43 1.66 1.54 1.16 1.47 0.80 1.05 1.13 1.63 1.19 1.14 1.52 1.40 0.83 1.01 1.05 1.08 1.59 1.34 1.45 Sincere M 3.26 4.34 4.44 5.02 4.10 4.40 3.32 2.70 3.94 4.50 5.90 6.06 5.10 4.88 4.76 5.38 4.02 5.76 4.76 2.64 5.46 4.20 4.32 SD 1.32 1.47 1.36 1.13 1.33 1.25 1.41 1.28 1.57 1.15 0.79 1.32 1.25 1.14 1.48 1.01 1.38 0.77 1.59 1.27 1.09 1.26 0.94 Close Rels M 4.62 4.84 4.66 4.92 4.78 3.90 4.26 4.62 3.82 4.90 5.08 5.62 4.32 4.86 3.62 5.22 4.34 6.04 4.82 4.52 5.26 4.78 4.68 SD 1.05 1.18 1.00 1.05 1.31 1.42 1.55 1.23 1.44 1.11 1.18 1.41 1.32 1.11 1.60 1.00 1.38 0.78 1.26 1.20 1.41 1.15 1.32 Aggress M 4.30 4.96 4.82 5.58 4.52 4.78 4.80 5.76 6.74 3.64 3.98 2.44 3.62 2.18 2.88 3.70 6.14 3.54 6.10 5.64 2.88 3.54 6.02 SD 1.52 1.64 1.37 1.57 1.25 1.33 1.56 1.56 0.53 1.52 1.70 1.73 1.59 1.22 1.32 1.45 1.29 1.27 0.79 1.48 1.29 1.15 1.02 Education Level M 3.18 3.86 4.42 5.76 1.84 2.08 3.34 3.24 4.60 5.72 6.96 4.02 6.62 3.84 4.18 6.72 6.86 5.68 4.70 5.36 4.00 1.92 3.56 SD 1.22 1.14 1.16 0.72 0.79 0.94 1.17 1.10 1.25 0.90 0.20 1.45 0.67 1.45 1.30 0.54 0.35 0.74 0.86 1.29 1.32 0.99 1.18 Social Status M 3.66 5.52 4.24 5.36 1.80 2.04 5.88 2.88 5.30 4.98 6.88 5.04 5.82 3.28 2.88 6.36 6.44 4.96 5.46 5.44 3.48 2.00 6.34 SD 1.17 1.11 1.06 0.96 0.64 0.90 1.04 1.10 1.09 1.17 0.39 1.41 1.10 1.20 1.47 0.75 0.70 0.86 1.01 1.26 1.30 0.95 0.80

M 6.37 6.10 5.95 4.88 4.88 5.15 5.28 4.70 3.00 2.75 2.95 2.48 3.03 2.68 3.20 3.20 4.60 3.47 3.65 4.18 3.60 4.63 4.08

SD 1.46 1.06 1.20 1.49 1.96 1.64 1.45 1.81 1.81 1.56 1.52 1.47 1.48 1.33 1.70 1.54 1.81 1.30 1.70 1.85 1.72 1.88 1.58

271

(continued)

272 Pilot Study (n = 40) Sarcasm Occupation Mechanic Chef Waiter Nurse Janitor Counselor Prison guard Secretary Bank teller Fashion designer Gardener Plumber Aerobics instructor Musician Farmer Electrician Astronaut Engineer Hairdresser Insurance agent Butcher Veterinarian M 4.20 3.77 4.38 3.21 3.78 3.28 3.47 4.38 3.93 4.00 3.55 4.32 4.13 3.98 4.10 4.25 3.47 3.53 4.52 3.58 3.70 3.28 SD 1.81 1.31 1.46 1.36 1.37 1.54 1.93 1.56 1.42 1.54 1.38 1.46 1.38 1.64 1.57 1.46 1.77 1.95 1.43 1.62 1.32 1.45 Humor M 4.54 4.66 5.50 4.90 4.24 4.56 2.82 4.52 4.34 4.50 4.00 4.74 5.08 4.74 4.32 4.24 4.52 3.68 5.46 3.66 4.32 4.60 SD 1.28 0.92 1.15 0.93 1.59 1.23 1.57 1.13 1.14 1.07 1.43 1.07 1.19 1.14 1.45 1.00 1.09 1.48 0.95 1.10 1.35 1.09 M

APPENDIX Continued
Experiment 3 (n = 50) Critic SD 1.13 1.12 1.57 1.22 1.64 1.37 1.37 1.45 1.32 0.97 1.29 1.16 1.49 1.13 1.21 1.02 1.22 0.92 1.38 1.24 1.23 1.18 Sincere M 4.16 5.02 4.14 5.88 4.32 5.90 3.78 4.56 4.34 4.42 5.02 4.12 4.48 5.00 5.28 4.34 5.08 4.76 4.58 3.72 4.50 5.76 SD 1.13 1.04 1.26 0.87 1.46 0.93 1.36 0.93 1.04 1.18 1.22 1.06 1.07 1.03 1.13 0.94 1.08 1.00 1.37 1.20 1.11 0.94 Close Rels M 4.80 5.16 5.22 5.66 4.68 5.94 3.78 5.52 5.12 4.92 5.22 4.70 5.36 5.10 5.64 4.96 4.84 4.42 5.58 4.70 4.94 5.60 SD 1.05 1.08 1.13 0.92 1.25 0.82 1.30 0.97 1.14 1.21 1.22 1.15 0.92 1.25 1.21 1.01 1.18 1.28 0.97 0.95 1.10 0.73 Aggress M 4.12 3.52 3.66 3.34 2.80 3.30 6.20 2.94 3.18 4.54 2.36 3.52 4.12 3.62 3.62 3.48 4.54 4.24 2.86 4.66 4.38 3.28 SD 1.14 1.36 1.29 1.39 1.25 1.59 1.03 1.24 1.29 1.79 1.14 1.28 1.72 1.40 1.46 1.40 1.40 1.46 1.18 1.67 1.34 1.37 Education Level M 3.38 4.04 2.24 5.58 1.66 5.22 3.36 3.38 3.86 3.68 2.56 2.84 2.80 3.76 2.76 3.86 6.72 6.46 2.32 4.94 2.16 6.38 SD 1.14 1.11 0.74 0.81 0.94 1.04 1.08 1.09 1.23 1.19 1.23 1.06 1.21 1.38 1.48 1.11 0.61 0.71 1.02 1.08 0.96 0.83 Social Status M 2.96 4.18 2.36 4.86 1.40 4.84 3.56 3.04 3.64 5.00 2.54 2.30 2.96 4.30 2.96 3.44 6.52 6.08 2.54 4.56 2.14 5.72 SD 1.11 1.02 0.88 0.90 0.70 1.13 1.07 1.05 0.98 0.93 1.13 1.04 1.05 1.47 1.03 0.93 0.86 0.83 0.97 1.16 0.86 0.93

4.58 5.18 4.06 4.46 4.28 4.14 5.42 4.26 3.98 5.80 3.62 4.54 4.28 4.84 4.34 4.32 4.36 5.24 4.24 4.62 3.80 4.48

Note. Critic = tendency to criticize; Close Rels = tendency to have close relationships; Aggress = tendency to be aggressive. a. Sarcastic occupations from Experiment 1. b. Nonsarcastic occupations from Experiment 1.

Pexman, Olineck / SPEAKER STEREOTYPES AND IRONY

273

NOTE
1. In this article, we use the terms irony or ironic to refer to utterances in which the speakers intended meaning is opposite to the literal meaning. This is the form of verbal irony that is perceived as sarcasm.

REFERENCES
Clark, H. H., & Gerrig, R. J. (1984). On the pretense theory of irony. Journal of Experimental Psychology: General, 113, 121-126. 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. (2000). On the necessary conditions for verbal irony comprehension. Pragmatics and Cognition, 8, 277-324. Dews, S., & Winner, E. (1995). Muting the meaning: A social function of irony. Metaphor and Symbolic Activity, 10, 3-19. Fussell, S. R., & Kreuz, R. J. (1998). Social and cognitive approaches to interpersonal communication. Mahwah, NJ: Lawrence Erlbaum. Gernsbacher, M. A. (Ed.). (1994). Handbook of psycholinguistics. San Diego, CA: Academic Press. Gibbs, R. W. (1986). On the psycholinguistics of sarcasm. Journal of Experimental Psychology: General, 115, 3-15. Gibbs, R. W. (1987). Memory for requests in conversation revisited. American Journal of Psychology, 100, 179-191. Gibbs, R. W. (2000). Irony in talk among friends. Metaphor and Symbol, 15, 5-27. Glucksberg, S. (1995). Commentary on nonliteral language: Processing and use. Metaphor and Symbolic Activity, 10, 47-57. Hancock, J., Dunham, P. J., & Purdy, K. (2000). Childrens comprehension of critical and complimentary forms of verbal irony. Journal of Cognition and Development, 1, 227248. Haverkate, H. (1990). A speech act analysis of irony. Journal of Pragmatics, 14, 77-109. Jorgensen, J. (1996). The functions of sarcastic irony in speech. Journal of Pragmatics, 26, 613-634. Katz, A. N., & Lee, C. J. (1993). The role of authorial intent in determining verbal irony and metaphor. Metaphor and Symbolic Activity, 8, 257-279. Katz, A. N., & Pexman, P. M. (1997). Interpreting figurative statements: Speaker occupation can change metaphor to irony. Metaphor and Symbol, 12, 19-41. Kreuz, R. J., & 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., Long, D. J., & Church, M. B. (1991). On being ironic: Pragmatic and mnemonic implications. Metaphor and Symbolic Activity, 6, 149-162. Kumon-Nakamura, S., Glucksberg, S., & Brown, M. (1995). How about another piece of pie: The allusional pretense theory of discourse irony. Journal of Experimental Psychology: General, 124, 3-21. Leggitt, J. S., & Gibbs, R. W. (2000). Emotional reactions to verbal irony. Discourse Processes, 29, 1-24. OBrien, E. J., & Myers, J. L. (1985). When comprehension difficulty improves memory for text. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11, 1221. Pexman, P. M., Ferretti, T. R., & Katz, A. N. (2000). Discourse factors that influence online reading of metaphor and irony. Discourse Processes, 29, 201-222.

274

JOURNAL OF LANGUAGE AND SOCIAL PSYCHOLOGY / September 2008

Pexman, P. M., & Olineck, K. M. (2002). Does sarcasm always sting? Investigating the impact of ironic insults and ironic compliments. Discourse Processes, 33, 199-217. Roberts, R. M., & Kreuz, R. J. (1994). Why do people use figurative language? Psychological Science, 5, 159-163. Slugoski, B. R., & Turnbull, W. (1988). Cruel to be kind and kind to be cruel: Sarcasm, banter, and social relations. Journal of Language and Social Psychology, 7, 101-121. Slusher, M. P., & Anderson, C. A. (1987). When reality monitoring fails: The role of imagination in stereotype maintenance. Journal of Personality and Social Psychology, 52, 653-662. Sperber, D., & Wilson, D. (1981). Irony and the use-mention distinction. In P. Cole (Ed.), Radical pragmatics (pp. 296-318). New York: Academic Press. Sperber, D., & Wilson, D. (1986). Relevance: Communication and cognition. Cambridge, MA: Harvard University Press. Toplak, M. T., & Katz, A. N. (2000). On the uses of sarcastic irony. Journal of Pragmatics, 32, 1467-1488. Utsumi, A. (2000). Verbal irony as implicit display of ironic environment: Distinguishing ironic utterances from nonirony. Journal of Pragmatics, 32, 1777-1806.