The Cognitive Psychology of Humor

C H A P T E R

5

The Cognitive Psychology of Humor

The Psychology of Humor DOI: https://doi.org/10.1016/B978-0-12-812143-6.00005-9

141

© 2018 Elsevier Inc. All rights reserved.

142

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

What are the mental processes involved in “getting a joke” or perceiving something as funny? Is sense of humor related to creativity or other cognitive abilities? Does humor affect cognitive processes (e.g., are we likely to remember humorous information better than serious information? Does experiencing humor cause people to think more creatively?). These sorts of questions are relevant to cognitive psychology, which has been defined as “the study of human mental processes and their role in thinking, feeling, and behaving” (Kellogg, 1995, p. 4). Cognitive psychologists use experimental methods to study how the mind works. Although they recognize that the brain does not function exactly like an electronic computer, they often find it useful to employ a computer analogy in conceptualizing mental processes. Thus, they take an information-processing approach to understand how information is taken in through our sensory organs, encoded, stored, and retrieved from memory, and then used in the comprehension and production of language, problem-solving, creativity, decision making, and reasoning. In short, cognitive psychology is concerned with mental representations of meaning and the mental processes that operate on those representations. Cognitive psychology has made an important impact on the psychological study of humor. We saw in Chapter 2, Classic Theories of Humor, that classic and contemporary theories of humor can be roughly categorized as motivational or cognitive, depending on their primary focus and presumed causal mechanisms (see Table 2.1). Furthermore, the most influential humor theories—incongruity theory and its contemporary descendants described in Chapter 3, Contemporary Theories of Humor—are primarily cognitive theories. That is, they borrow concepts and theory from cognitive psychology to describe the mental processes involved in the perception and interpretation of a stimulus event, identifying perceptions of incongruity as a cognitive mechanism that is minimally necessary for all humor. Cognitive psychology is part of a broader interdisciplinary enterprise known as cognitive science, which also includes some branches of linguistics (the scientific study of language), psycholinguistics (the scientific study of cognitive processes involved in language comprehension and production), neuroscience, and computer science (artificial intelligence (AI)). Over the past two decades, linguists (e.g., Attardo, 1994; Raskin, 1985), psycholinguists (e.g., Giora, 1991), and computer scientists (e.g., Ritchie, 2004) have contributed to comprehensive cognitive theories of humor by further investigating and specifying cognitive underpinnings of humor (e.g., Vaid, Hull, Heredia, Gerkens, & Martinez, 2003). These advances will hopefully stimulate further interest among psychologists in the study of cognitive processes in humor. We have divided this chapter into six parts. First, we consider cognitive methods in the study of humor, followed by an application of those methods to the study of the cognitive processes involved in conversational humor, specifically irony and sarcasm. Third, we address linguistic theories of humor that complement and expand upon incongruity theory and its contemporary descendants described in Chapter 3, Contemporary Theories of Humor. Fourth, we discuss computational approaches to the study of humor that attempt to model linguistic theories using AI. Fifth, we consider the effects of humor on cognition, specifically creativity and memory. Finally, we discuss the conceptualization of sense of humor as a cognitive ability.

THE PSYCHOLOGY OF HUMOR

COGNITIVE METHODS IN THE STUDY OF HUMOR

143

COGNITIVE METHODS IN THE STUDY OF HUMOR Semantic Distance Cognitive psychologists have developed techniques to determine the strength of the semantic association between concepts in memory, and thus a way to investigate how people store knowledge or meaning of concepts in their minds. As we will see, some humor researchers have used these techniques as a way to study the concept of incongruity in humor in a rigorous and quantifiable manner. Charles Osgood and his colleagues (Osgood, Suci, & Tannenbaum, 1957; Snider & Osgood, 1969) pioneered an early approach based on a semantic differential rating scale. Participants would rate numerous words or concepts on a series of scales with endpoints represented by polar opposite adjectives (e.g., hot cold, fast slow, good evil). The researcher would then factor analyze these ratings to identify a smaller number of basic dimensions (factors) that capture most of the variance in participants’ ratings. These factors were assumed to represent the fundamental ways people organize information in their minds. Using ratings of many concepts and many different samples of participants, Osgood and his colleagues repeatedly found that people mentally organize the meaning they attach to concepts using three independent dimensions or factors: Activity (active passive), evaluative (good bad), and potency (strong weak). Osgood conceptualized these factors as three different dimensions in cognitive or mental “space” in which people store words and concepts. Researchers can use the factor loadings of a specific word or concept to identify where the concept is stored in this space. Concepts that are similar in meaning are stored closely together in this hypothetical semantic space, since they have similar loadings on the three factors, whereas those that are quite different in meaning have different loadings and are stored at more distant locations. Thus, one can measure the semantic distance between pairs of words or concepts according to the difference in their loadings on the semantic differential factors. By providing a way of measuring the perceived similarity between concepts, researchers can use the semantic differential method to quantify incongruity in the study of humor (the greater the semantic distance between two words or concepts, the greater the incongruity). In an early study in this vein, Michael Godkewitsch (1974), at the University of Toronto, used the semantic distance between words as a means of measuring incongruity. He asked participants to rate the funniness and wittiness of multiple adjective noun combinations (e.g., “happy child”). He also observed the degree to which participants smiled and laughed. He computed the semantic distance between the words in each pair based on their loadings on the semantic differential factors. As predicted by incongruity theory, participants rated adjective noun pairs funnier to the extent they differed from one another in semantic space. For instance, participants rated semantically distant word pairs (e.g., “hot poet,” “wise egg”) as being funnier than semantically close word pairs (“happy child”). None of the word pairs evoked much amusement. However, supporting incongruity theories of humor, amusement varied systematically as a function of the semantic distance between the word pairs, i.e., how similarly the words loaded on the semantic differential factors. Tim Hillson and Rod Martin also used a semantic distance procedure, the domaininteraction approach (e.g., Trick & Katz, 1986), to examine both incongruity and resolution

THE PSYCHOLOGY OF HUMOR

144

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

of incongruity in simple verbal stimuli (Hillson & Martin, 1994). They hypothesized that word pairs that are quite distant on some dimensions of semantic space (creating incongruity), but also quite close on other dimensions (allowing for resolution of incongruity), are funnier than those that are either distant or close on all dimensions. As humor stimuli, they used simple metaphor-like statements combining two concepts in the form “A is the B of A’s domain” (e.g., “George Bush is the buzzard of world leaders”). The domains used were actors, world leaders, birds, makes of cars, foods, and magazines, and within each domain they used four nouns (e.g., Sylvester Stallone and Woody Allen were two of the actors). Factor analyses of participants’ semantic differential ratings of these nouns and domain names yielded four factors. They identified two of the factors as domain distinguishing (i.e., different nouns within a given domain were found to have very similar factor loadings, whereas nouns from different domains had more distant loadings). They identified the other two factors as domain insensitive (i.e., different nouns within the same domain could have quite different loadings on these two factors). Based on these factor loadings, they computed two types of semantic distance between the nouns: a within-domain distance (using the domain-insensitive factor loadings), and a between-domain distance (using the domain-distinguishing factor loadings). They considered between-domain distance to be a way of operationally defining incongruity (greater distance 5 greater incongruity), and within-domain distance to be a way of operationalizing incongruity resolution (less distance 5 greater resolution). They then created metaphor-like sentences using pairs of nouns from different domains and asked a second group of participants to rate them for funniness. As predicted (and consistent with the findings of Godkewitsch, 1974), the between-domain distance (incongruity) of the noun pairs in each sentence showed a significant positive correlation with the funniness ratings of the sentences. That is, noun pairs with greater between-domain distance were rated as more funny. Also, as predicted, within-domain distance (incongruity resolution) did not correlate with funniness, but did produce a significant interaction with between-domain distance in predicting funniness ratings. Specifically, sentences that were rated as most funny were those that showed both high between-domain distance (incongruity) and low within-domain distance (resolution). To illustrate, a sentence that received a relatively high mean humor rating was “Woody Allen is the quiche of actors.” The between-domain semantic distance between Woody Allen and quiche was large (actors are quite different from foods on some dimensions), but the within-domain distance was small (Woody Allen and quiche are quite similar in some ways within their respective domains). Thus, there is incongruity but also some sort of resolution to the incongruity (i.e., the incongruity “makes sense” in some way). The semantic distance approach did seem to capture some relevant dimensions of humor, as it was capable of systematically predicting funniness ratings of simple verbal material. It could still be a useful method for exploring various additional parameters that may be relevant to some types of humor. However, this technique has several limitations. It provides only a static picture of the organization of semantic meaning, and is therefore not useful for examining the processes whereby cognitive structures (schemas) are activated over time in processing humorous information. It also assumes that cognitive organization is the same in all people and, because mean ratings are averaged across large numbers of participants, it is not amenable to studying individual differences in humor

THE PSYCHOLOGY OF HUMOR

COGNITIVE METHODS IN THE STUDY OF HUMOR

145

comprehension. In addition, it allows only for the study of simple “pseudo-jokes” made up of word pairs, rather than more complex real jokes and other natural forms of humorous material.

Semantic Priming More recently, cognitive psychologists and psycholinguists have developed several more sophisticated techniques to study schema activation in real time. These techniques may also be useful for humor researchers who wish to study the way people mentally process humorous information. As we discussed in Chapter 3, Contemporary Theories of Humor, several contemporary cognitive theories of humor employ the concept of scripts to explain how incongruity occurs in jokes and other forms of humor. One method, developed by David Meyer and Roger Schvaneveldt (1971), called the lexical decision task, measures how quickly people classify a string of letters as a word or nonword. In a lexical decision task, a researcher presents participants with a string of letters on a computer screen; participants indicate as quickly as possible whether the letter string is a real word or a nonword (i.e., a random string of letters) by pressing a key associated with each of the options. The reaction time for making this decision is measured in milliseconds. The reaction time, or response latency, for a real word indicates the degree to which that word was activated in memory; faster reaction times mean greater activation. Researchers often use the lexical decision task to measure the effect of priming. The priming effect occurs when exposure to one concept influences the level of activation of other semantically related concepts. Using a lexical decision task, Meyer and Schvaneveldt (Meyer, Schvaneveldt, & M. G. Ruddy, 1975; Schvaneveldt & Meyer, 1973) found that people recognized a string of letters as a real word faster if they had first been exposed to a semantically related word. For instance, participants recognized NURSE faster if they had first seen DOCTOR than if they had first seen BREAD. Psycholinguists have used this lexical decision task to determine the way various scripts become activated while people are reading narrative texts (e.g., Sharkey & Mitchell, 1985). Recently, psychologists have begun to make use of techniques such as the lexical decision semantic priming task in the study of humor comprehension. For example, Jyotsna Vaid and her colleagues (2003) at Texas A&M University used this technique to study schema activation while people read jokes. Based on incongruity theory, they hypothesized that an initial schema (S1) is activated during the joke setup, and a second, surprising or incongruous schema (S2), is activated later in the joke. They were interested in determining the point at which S2 becomes activated, and whether S2 replaces S1, so that only S2 remains active by the end of the joke (the selective attention view), or whether both S1 and S2 remain activated concurrently right up to the end of the joke (the concurrent activation view). This question is relevant to the competing predictions made by incongruity-resolution theories, such as those of Suls (1972) and Shultz (1976), versus concurrent activation theories such as those of Attardo and Raskin (1991) and Wyer and Collins (1992). Vaid and her colleagues presented participants with a series of jokes printed on a computer screen (see Fig. 5.1). Each joke was divided into three segments, with the punch line

THE PSYCHOLOGY OF HUMOR

146

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

Baseline

First half of joke setup

*

A woman walks into a bar with a duck on a leash.

*

2nd half of joke setup

*

*

Bartender says, “Where’d you get the pig? Woman says, “This is not a pig; it’s a duck.”

Joke punch line

*

Bartender says, “I was talking to the duck.”

S1 = “PET” S2 = “SOW”

Presentation of lexical decision probe

Reaction time (milliseconds)

FIGURE 5.1 Vaid et al.’s (2003) sequence of presenting joke segments followed by a lexical decision task. Source: Adapted from “Getting a joke: The time course of meaning activation in verbal humor,” by J. Vaid et al. (2003). 300 250 200 Schema 1

150

Schema 2

100 50 0 Setup 1

Setup 2

Punch line

FIGURE 5.2 Mean reaction times on a lexical decision task for words related to the initial schema (Schema 1) or the surprising schema (Schema 2) following each of three segments of a joke. Source: Results adapted from “Getting a joke: The time course of meaning activation in verbal humor,” by J. Vaid et al. (2003).

forming the third segment. After each segment, participants completed a lexical decision task involving words that were semantically related to either the initial schema (S1) or the second (surprising) schema (S2). If participants exhibited faster reaction times for a given word after a joke segment compared to a baseline test, this indicated that the schema associated with that word had been activated at that point in the processing of the joke. The results revealed that the initial schemas (S1) were activated during the presentation of the first two segments of the jokes, whereas the surprising second schemas (S2) became activated during the second segment. Unexpectedly, however, they found that neither of the schemas showed a pronounced activation in the final lexical decision test occurring immediately after the punch line (see Fig. 5.2). These results were difficult to explain. On the one hand, they seemed to show some support for the concurrent activation view, since S2 was not more strongly activated than S1 by the end of the joke. On the other hand, though, the lack of activation of either schema by that point was inconsistent with either hypothesis. This finding needs to be replicated in further research before firm conclusions can be drawn. Interestingly, the finding that S2 was primed well before the punch line suggests that numerous potential schemas may be activated even before the incongruity is encountered. This finding seems to

THE PSYCHOLOGY OF HUMOR

COGNITIVE PROCESSES IN CONVERSATIONAL HUMOR: IRONY AND SARCASM

147

provide additional evidence, consistent with the findings of Kenny (1955) and Pollio and Mers (1974) discussed in Chapter 2, Classic Theories of Humor, that the recipients of a joke have already anticipated the “true” meaning of a joke well before they hear the punch line, rather than it being unexpected (as suggested by incongruity-resolution theories). However, further research is needed to more fully investigate the priming of schemas at multiple time points during and after the presentation of jokes. Other methods that have been developed for psycholinguistic research on schema activation could also be adapted to address research questions relating to humor. For instance, Stewart and Heredia (2002) used the cross-modal lexical priming task to study schema activation during metaphor comprehension. In the cross-modal lexical priming task, researchers expose participants to auditory information (e.g., a joke or funny narrative) via headphones, and present probe words related to various schemas on a computer screen at precise moments during the auditory presentation. Participants read the probe words aloud as quickly as possible; researchers record how quickly participants can read the probe words. Because words that are semantically related to currently activated schemas are spoken more quickly than those unrelated to activated schemas, this is another way of testing the degree of activation of specific schemas in memory. Another method is the word fragment completion test (e.g., Giora & Fein, 1999), in which participants are instructed to complete a fragmented (partially spelt out) word with the first word they can think of. People more frequently complete fragments with words that are semantically related to recently primed concepts than with words that are unrelated to recently primed concepts. As this brief overview shows, these sorts of techniques hold a great deal of promise for cognitive research on humor, enabling researchers to test specific hypotheses about the time course of schema activation during the processing of humorous texts. More studies are needed to replicate the initial findings of Vaid et al. (2003), to clarify the patterns that have been observed, and to broaden the scope of inquiry. These authors listed a number of unanswered research questions, including the precise timing and duration of activation of the schemas, the role of individual differences in joke processing, the effects of manipulating subjects’ expectations about whether or not they will be encountering humorous materials, the degree to which meaning activation in joke processing is subject to strategic versus automatic control, and the processes involved in different types of humorous texts besides jokes, such as humor occurring spontaneously in conversation (e.g., irony, witticisms). Besides greatly enriching our understanding of the cognitive processes involved in humor, the results of these sorts of investigations should help to address longstanding debates among theorists, such as the debate about incongruity versus incongruityresolution as the basis of humor.

COGNITIVE PROCESSES IN CONVERSATIONAL HUMOR: IRONY AND SARCASM Empirical research on the cognitive underpinnings of humor has largely examined people’s responses to “canned” jokes. This is due, in part, to theoretical considerations. For instance, linguistic theories explicitly address the comprehension of verbal humor such as

THE PSYCHOLOGY OF HUMOR

148

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

jokes. The empirical focus on jokes is also due to methodological considerations. Jokes are self-contained; they do not depend on a specific context for comprehension. Thus, they lend themselves well to empirical research. In contrast, conversational humor depends more on the constantly changing social context and therefore poses greater challenges for theorists and researchers. However, as noted in Chapter 1, Introduction to the Psychology of Humor, most of the humor that we encounter in everyday life is not in the form of canned jokes (R. A. Martin & Kuiper, 1999; Provine, 2000). Indeed, most of the humor we experience through the course of a normal day occurs because of spontaneous interactions with other people in the form of witty retorts, wordplay, banter, teasing, irony, sarcasm, slips of the tongue, practical jokes, and pratfalls (Long & Graesser, 1988; Norrick, 1993, 2003). Accordingly, some theorists have attempted to empirically study spontaneous, conversational forms of humor. For instance, Wyer and Collins (1992) and Wyer (2004) showed how their comprehension-elaboration theory of humor elicitation accounts for many types of witticisms as well as unintentional humor and even nonverbal humor. Norrick (1986) applied his schema conflict theory to a variety of conversational witticisms, including witty retorts, quips, and one-liners. Lippman and Dunn (2000) also conducted a series of experiments on appreciation and memory for puns. Humor scholars in the field of psycholinguistics have particularly studied two types of conversational humor: irony and sarcasm. As mentioned in Chapter 1, Introduction to the Psychology of Humor, irony is a figure of speech that communicates the opposite of what is said (Kreuz, Roberts, Johnson, & Bertus, 1996; Srinarawat, 2005). For example, someone who says “What a beautiful day!” during a bleak and miserable day is actually communicating “what an awful day.” Burgers (2010) defined irony as “an evaluative utterance, the valence of which is implicitly reversed between the literal and intended evaluation” (p. 19). Although irony is not always funny, it can be a source of humor. Sarcasm is a special case or type of irony. Unlike simply communicating the opposite of what is said as in the case of irony, sarcasm conveys a negative evaluation or opinion from a positive statement. Gibbs (1986, p. 3) stated that sarcasm depends for its effect on “bitter, caustic, and other ironic language that is usually directed against an individual.” For example, if someone says, “You’re a fine friend” to someone who has been unkind, this is an ironic statement that is also sarcastic. Much of the research in this area has investigated how people recognize irony and sarcasm in everyday conversation. For example, when we hear someone say something like “You’re a fine friend,” how do we determine whether the statement is meant to be taken literally or as nonliteral irony or sarcasm? Psycholinguist Rachel Giora and her colleagues at Tel Aviv University have proposed a Graded Salience Theory of humor that focuses primarily on irony. Giora (1985, 1995; Giora, Fein, Ganzi, Levi, & Sabah, 2005) suggested that there are implicit rules that people follow while engaging in conversation (“discourse”): (1) all messages should be relevant to the topic of conversation (the relevance requirement); (2) successive messages should be gradually more informative, and not less informative, than preceding ones (the graded informativeness requirement); and (3) any deviation from the first two rules should be “marked” with an explicit semantic connector such as “by the way” or “after all.” When we try to understand the meaning of something another person says during a conversation, we are initially guided by the “graded salience

THE PSYCHOLOGY OF HUMOR

COGNITIVE PROCESSES IN CONVERSATIONAL HUMOR: IRONY AND SARCASM

149

principle,” which dictates that salient meanings (i.e., the more conventional, common, familiar, or prototypical meanings) are always activated first. If the salient meaning does not match the context (doesn’t make sense), then less salient meanings are activated. Subsequently, there is a contextual integration phase, in which any meanings that have been activated are either retained, or suppressed as irrelevant or disruptive, or permitted to fade. According to Giora (1995, 1998), an ironic statement in a conversation (e.g., someone saying “What a beautiful day!” when the weather is cold and stormy), conforms to the relevance requirement, since it introduces information about the current topic of conversation (the weather, in this example), but it violates the graded informativeness requirement, since it introduces an improbable message with a salient meaning that is either too informative or not informative enough (the discrepancy between the statement and the actual weather). To understand the ironic statement, the listener first activates its salient (literal) meaning (“the weather is beautiful”), but, since this does not make sense in the context, must then activate an “unmarked” interpretation, called the “implicature” (“the weather is terrible”), and both meanings remain activated for comparison. The incongruity between the two activated meanings (the actual stormy weather and the nice weather implied in the statement) causes the irony to be humorous. In addition to explaining irony, Giora (1991) applied her graded salience theory to the understanding of jokes. Although her theory is similar to Raskin’s (1985) script-based theory (which we will discuss in a later section of this chapter), Giora’s theory takes the social context of humor into account, and therefore is more applicable to nonjoke-related humor. Indeed, Norrick (2003) has applied Giora’s theory to various types of conversational witticisms, including puns and amusing anecdotes. Because irony involves activating two meanings, Giora’s theory predicts that people are slower to comprehend ironic statements as compared to nonironic ones. Further, Giora’s theory predicts that both meanings should remain activated after the “true” meaning of the ironic statement has been understood. In contrast to Giora’s theory, other models (e.g., Clark & Gerrig, 1984; Gibbs, 1994; Sperber, 1984) argue that, given enough contextual information, people process an ironic statement in the same way as a literal one. This is known as the processing equivalence hypothesis. From this perspective, people should not take longer to understand an ironic versus a nonironic statement, and only the ironic meaning will be activated. Although some research findings seem to support the processing equivalence hypothesis (e.g., Gibbs, 1986), Giora (1995) reinterpreted these findings in light of her own theory. In addition, Giora and her colleagues have conducted several experiments that support her graded salience theory rather than the processing equivalence hypothesis. Giora, Fein, and Schwartz (1998), for instance, found that participants took longer to comprehend a written statement in a context that cued an ironic interpretation versus a context that cued a literal interpretation, suggesting that the ironic statement required greater information processing. Using lexical decision tasks and word fragment completion tasks, Giora et al. (1998) showed that a context that cued an ironic interpretation of a sentence activated both ironic and literal meanings; however, a context that cued a literal interpretation activated only the literal meaning. Moreover, Giora et al. (2005) noted that statements vary on an irony continuum depending on the size of the discrepancy between the literal statement

THE PSYCHOLOGY OF HUMOR

150

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

and the intended meaning. Giora et al. categorized statements in descending degrees of irony as (1) highly ironic (e.g., “He is exceptionally bright.”), (2) negated overstatement (e.g., “He is not exceptionally bright.”), (3) nonoverstatement (“He is not bright.”), to (4) literal (“He is stupid.”). Giora et al. found that people more easily interpret a statement as ironic to the extent that the discrepancy between the literal statement and the intended meaning is large. These findings are consistent with research suggesting that the use of hyperbole or exaggeration strengthens the evaluative meaning of a statement. Van Mulken and Schellens (2012) refer to hyperbolic words as “intensifiers,” as they magnify or intensify the underlying evaluative connotation of a statement. Other research suggests that the conflict between the processing equivalence and graded salience hypotheses may be resolved by taking social context into account. Albert Katz, a cognitive psychologist at the University of Western Ontario, and his colleagues (Katz, Blasko, & Kazmerski, 2004) summarized a body of research investigating the way people process sarcastic statements when they have been provided with information about the interpersonal context, such as the degree of relatedness and shared knowledge of the participants in a conversation, or the gender and occupation of the speaker. Taken together, these studies showed that the speed with which people recognize statements as sarcasm depends on their prior information about the context. For example, some studies found that, when subjects are told that a statement was made by a male, they take no longer to read sarcastic statements than literal statements (supporting the processing equivalence hypothesis), whereas when the statement is made by a female, sarcastic sentences take longer to read than literal ones (supporting the graded salience hypothesis). These findings suggest that, since people tend to associate sarcasm with males more than females, the sarcastic meaning of an utterance by a male is more readily available during the comprehension process. In contrast, when a woman makes a sarcastic comment, the literal meaning tends to be activated initially, before the sarcastic meaning is accessed, resulting in lengthier processing. Similar differences in processing time were found when participants were given information about the occupation of the speaker. People processed a comedian’s or a factory worker’s sarcastic statements more quickly than a priest’s or teacher’s sarcastic statements. Katz and his colleagues (2004) proposed a Constraint-Satisfaction Model to account for these sorts of findings. According to this theory, different sources of information about the social context (i.e., constraints) provide probabilistic support for different possible interpretations of an utterance (e.g., whether it is meant to be literal versus sarcastic). These constraints operate in parallel while one processes a statement. If the constraints all point in the same direction, competition between alternative interpretations is resolved rapidly, whereas settling on an interpretation takes longer if support for different alternatives is nearly equal. Thus, the social context in which ironic or sarcastic statements are made plays an important role in determining how efficiently they are interpreted. If all indicators point towards a humorous interpretation right from the start, the incongruity of humor can be interpreted very quickly. Other psycholinguistic investigations have provided further evidence of the importance of taking the interpersonal context into account in the interpretation of irony and sarcasm. For example, Penny Pexman and Meghan Zvaigzne (2004), at the University of Calgary,

THE PSYCHOLOGY OF HUMOR

COGNITIVE PROCESSES IN CONVERSATIONAL HUMOR: IRONY AND SARCASM

151

examined the effect of the closeness of a relationship on participants’ comprehension of sarcastic remarks (ironic insults) and compliments. As mentioned earlier, sarcastic remarks or ironic insults convey criticism or insult in a positive literal statement (e.g., saying “You’re a fine friend” when someone has done something unkind), whereas ironic compliments are negative statements that are intended to be taken positively (e.g., saying “Too bad you can’t play baseball” when someone has just scored a home run). Pexman et al. (2004) presented participants with vignettes describing either a close friend or a casual acquaintance making a positive or a negative statement in a positive or negative social context. Participants rated the statements along multiple dimensions. As expected, participants perceived a statement as ironic when the valence of the statement did not match the valence of the context (e.g., a positive statement in a negative context), regardless of whether the statement took place between close friends or casual acquaintances. However, the closeness of the relationship affected the perceived funniness of these ironic statements: irony occurring between close friends, as compared to casual acquaintances, was rated as more humorous, especially if it was an ironic compliment. Participants also perceived irony between close friends (as compared to acquaintances) as friendly teasing and having no impact on their relationship. Interestingly, participants rated ironic compliments as less polite than literal compliments, whereas they rated ironic insults as more polite than literal insults. Pexman et al. concluded from their findings that humor in the form of irony plays a role in building and maintaining close relationships. Similarly, Gibbs (2007) suggested that people use sarcasm directed at members of social out-groups as a method of affirming solidarity among members of an in-group. In addition, the solidarity or closeness of a relationship acts as a cue for how to interpret irony and sarcasm. Thus, social factors, as well as linguistic factors, are important for understanding irony (Pexman et al., 2004; Pexman & Olineck, 2002). Recently, researchers have begun to take a computational approach to better understand the linguistic characteristics of irony and sarcasm that people use when interacting on social media (e.g., Liebrecht, Kunneman, & van den Bosch, 2013; Reyes, Rosso, & Veale, 2012). For instance, using linguistic classification computer programs trained to detect sarcasm in tweets, Liebrecht et al. (2013) revealed many interesting characteristics of sarcasm on Twitter based on a sample of 3.3 million Dutch tweets. Specifically, most tweets contained a positive literal message about a topic of interest (e.g., school, family) and further contained three types of words to signal sarcasm: explicit markers such as “#sarcasme” (#sarcasm in English), “#ironie (irony)” or “#LOL,” “#NOT,” etc., intensifiers or hyperbolic words, and exclamations (e.g., “wow,” “woehoe, “jippie,” etc.). These findings suggest that explicit markers denoted by a hashtag essentially represent the social media equivalent of linguistic and nonverbal markers (e.g., voice intonations, facial expressions) that people use in conversation to signal sarcasm and irony (e.g., Attardo, Eisterhold, Hay, & Poggi, 2003; Bryant & Fox Tree, 2005; Rockwell, 2007). In summary, psycholinguists have recently debated the cognitive processes involved in the comprehension of irony and sarcasm, and this has stimulated a considerable amount of interesting research (see Creusere, 1999, for a review of earlier research). Moreover, as cognitive psychologists have moved beyond the study of jokes to these more conversational forms of humor, their research has increasingly taken the interpersonal context into

THE PSYCHOLOGY OF HUMOR

152

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

account, examining the effects of social, as well as linguistic factors, on cognitive processing of humor. Similar research efforts will hopefully be applied to investigate other types of conversational humor besides irony and sarcasm. Cognitive approaches for assessing schema activation and computational approaches are potentially useful tools for further creative research in this area.

LINGUISTIC APPROACHES TO HUMOR Linguistics, which is the study of language, comprises many subfields, including phonology (the study of speech sounds), syntax (grammatical rules that specify the acceptable form of sentences), semantics (language meaning), and pragmatics (rules for appropriate social use and interpretation of language in context). Not surprisingly, linguistic theories of humor address verbal or textual humor, humor that is communicated through language, rather than nonverbal forms like practical jokes or slapstick comedy. Beginning in the mid-1980s, linguists have developed several formal theories of humor (see Attardo, 1994; Kirkmann, 2006 for a review). In this section, we focus on influential linguistic theory and research emerging from the subfield of semantics. Linguists taking a semantic approach to the study of humor examine the way people process narratives or “texts” and interpret them as funny (e.g., Norrick, 1986; Raskin, 1985). The goal of semantic theories is not to delineate and understand the variables that affect how people go about determining whether something is funny or not. Rather, the goal is to develop a model of text analysis that fully and precisely specifies the rules for processing textual information that are necessary and sufficient to produce humor. Thus, linguistic theories complement and contribute to the existing, broader cognitive theories of humor based on incongruity (see Chapter 2, Classic Theories of Humor, and Chapter 3: Contemporary Theories of Humor) not by expanding their scope, but by refining their precision. Because linguists do not examine the actual ways people process information under various conditions, they typically do not conduct experiments on human participants to test their theories. Instead, they use logical reasoning to determine if a theory is internally coherent and whether it accounts for a wide range of text examples (i.e., jokes). Ideally, semantic theories would reach a level of specification that is both precise and broad enough to program a computer to make sense of any humorous text and to distinguish humorous from nonhumorous texts (Hempelmann, Taylor, & Raskin, 2012; Raskin, Hempelmann, & Taylor, 2009). Victor Raskin (1985) introduced a linguistic extension of incongruity theory called the Script-based Semantic Theory of Humor (SSTH). As a semantic theory, SSTH focused on verbal humor, and more specifically jokes that contain a setup and a punch line. SSTH analyzed jokes based on scripts and rules for combining the possible meaning of scripts associated with a given text. Raskin conceptualized the theory’s central concept, script, in a slightly different way from Schank and Abelson (1977). Recall from Chapter 3, Contemporary Theories of Humor, that Schank and Abelson described a script as a special type of schema that relates to knowledge about routine activities (e.g., going to a

THE PSYCHOLOGY OF HUMOR

LINGUISTIC APPROACHES TO HUMOR

153

restaurant) used to encode relevant information. In contrast, Raskin (1985) defined a script in terms of concepts associated with words, specifically: [A]n enriched, structured chunk of semantic information, associated with word meaning and evoked by specific words. The script is also a cognitive structure internalized by the native speaker, and it represents the native speaker’s knowledge of a small part of the world. (p. 81)

Accordingly, scripts contain a person’s complete knowledge of concepts and experiences associated with words. Scripts can contain general knowledge, i.e., an understanding of something that is shared by most native speakers (e.g., going to a restaurant). Native speakers will have similar general knowledge scripts for words they have in common. Scripts can also contain specific or personal knowledge known only to a given individual (e.g., a person’s memory of getting his first car; Hempelmann et al., 2012). According to SSTH, the scripts associated with a text provide the basis for humor. Specifically, Raskin described two necessary and sufficient conditions for one to perceive a spoken or written text as humorous: 1. A text must be compatible, fully or partially, with two different, overlapping scripts. That is, one must be able to interpret the text, at least partially, according to two different scripts. 2. The two scripts with which the text is compatible are opposite on some dimension. Thus, when a person attempts to understand a joke, a mental script is activated to make sense of the events described in the joke setup. However, the punch line of the joke introduces elements that are not compatible with that original script, triggering a switch from one script to another. The punch line makes the listener backtrack and realize that a different interpretation (i.e., an alternative script) was possible from the beginning. For the text to be viewed as humorous, this second, overlapping script must be opposite to the first. There are three general ways in which the scripts may be in opposition to one another: actual vs. nonactual, normal vs. abnormal, or possible vs. impossible. At a more concrete level, script oppositions may be manifested in terms of such pairs as good vs. bad, life vs. death, obscene vs. nonobscene, high stature vs. low stature, clean vs. dirty, intelligent vs. unintelligent, and so on (Ruch, Attardo, & Raskin, 1993). Raskin used the following joke to illustrate how the model works: “Is the doctor at home?” the patient asked in his bronchial whisper. “No,” the doctor’s young and pretty wife whispered in reply. “Come right in.”

According to Raskin’s theory, the first part of this joke evokes a standard “doctor” script (which is presumably stored in the listener’s semantic network) in which a patient presents himself at a doctor’s residence to be treated for an illness that causes him to have a hoarse voice, and is told that the doctor is not there. However, the doctor’s wife’s invitation for the patient to enter the house anyway does not fit with the “doctor” script, so the listener must backtrack and reevaluate the text. The information that the doctor’s wife is young and pretty and that she is inviting the patient into her house when her husband is away activates a different (i.e., “lover”) script. Both the “doctor” script and the “lover”

THE PSYCHOLOGY OF HUMOR

154

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

script are compatible with the text, and these two scripts are opposed to one another on the sex versus no-sex basis. Consequently, the joke fulfills the requirements of the theory and the listener finds it humorous. It is noteworthy that humor requires a connection or “overlap” between competing scripts. Simply presenting two incongruous scripts in a text is insufficient to elicit humor (Triezenberg, 2008). Note also that Raskin’s theory is more consistent with Koestler’s and Apter’s ideas of “bisociation” and “cognitive synergy” than with Shultz’s and Suls’ incongruity-resolution theories because, in Raskin’s theory, both scripts are activated at the same time, rather than one replacing the other. Salvatore Attardo and Victor Raskin revised and extended Raskin’s original SSTH into a broader linguistic theory called the General Theory of Verbal Humor (GTVH; Attardo & Raskin, 1991). GTVH is a semantic theory, but also involves other areas of linguistics such as theory of narrative, discourse analysis, and pragmatics (Attardo, 2001). GTVH represents jokes in terms of six Knowledge Resources (KRs), or hypothetical databases describing the characteristics by which people analyze humorous texts (Attardo, 1994). The six KRs, in order from most concrete to most abstract are: 1. Language (LA): LA describes the actual wording of a joke. 2. Narrative Strategy (NS): NS represents the genre or format of the joke (e.g., riddle or expository text). 3. Target (TA): TA refers to persons or groups ridiculed in a joke. 4. Situation (SI): SI represents the objects, people, places, and other details of the joke text. 5. Logical Mechanism (LM): LM describes the mechanisms or “joke techniques” by which two opposing scripts are brought together in a joke. LMs include such mechanisms as figure-ground reversal, juxtaposition, analogy, parallelism, and faulty reasoning. 6. Script Opposition (SO): SO incorporates the ideas of script overlap and opposition from SSTH and is considered a necessary condition for humor. Attardo (1997) discussed the relationship between the GTVH and incongruity-resolution theories (discussed in Chapter 2: Classic Theories of Humor). He proposed a “three-stage” (setup-incongruity-resolution) rather than a “two-stage” (incongruity-resolution) model of joke comprehension, suggesting that incongruity relates to the SO component, while resolution corresponds to the LM component, and setup refers to the overlap between the two scripts. Also, different from traditional incongruity-resolution theories, Attardo’s formulation suggests that resolution occurs before incongruity; the LM, corresponding to resolution, activates an alternative script, which along with the initial script creates incongruity. Finally, GTVH (like its predecessor, SSTH) assumes that humor arises from the concurrent activation of two incompatible scripts, and is therefore like the views of Koestler (1964), Apter (1982), and Wyer (2004), and different from incongruity-resolution models (e.g., Shultz, 1976; Suls, 1972), which assume that humor emerges from the elimination (resolution) of incongruity. Attardo, Hempelmann, and Di Maio (2002) further developed the concept of LMs, and proposed formulations of the model using graph theory and set theory. Attardo (1998) extended the GTVH to allow for the analysis of humorous texts that are longer than jokes. To do this, he introduced a variety of additional concepts such as jab lines and punch lines, macro- and micronarratives, levels of narratives, strands of lines, stacks of strands, and intertextual jokes (an explanation of these concepts is beyond the scope of the present discussion). Attardo also demonstrated how this more complex model could be applied

THE PSYCHOLOGY OF HUMOR

COMPUTATIONAL APPROACHES TO HUMOR

155

by using it to analyze a segment of a television sitcom. Thus, an attempt has been made to extend the theory so that it can account for spontaneous conversational humor in addition to canned jokes. Finally, Villy Tsakona (2009) demonstrated that one could use GTVH to analyze cartoons that derive humor from their text captions or dialogues. He further suggested that the Language KR accounts for humorous mechanisms such as exaggeration, contradiction, and metaphor in both verbal and strictly visual cartoons, thus expanding the application of GTVH beyond verbal text. Although this brief overview certainly does not do justice to linguistic theories of humor, it should give readers from psychology a sense of the kinds of theories linguists have developed that could serve as a basis for testable hypotheses in psychological research. For example, psychologist Willibald Ruch teamed up with Attardo and Raskin (Ruch et al., 1993) to conduct an empirical study to test propositions of the GTVH. Specifically, they tested the hypothesis that people see pairs of jokes that differ on a more abstract KR dimension as less similar than pairs of jokes that differ on a more concrete or superficial KR dimension. Ruch et al. presented participants with pairs of jokes from three different genres (blonde jokes, chicken jokes, and light bulb jokes) that differed from one another at only one KR level. Thus, two jokes might be identical in every way except that they involved different SOs or different LMs. Participants rated the similarity of the jokes in each pair. In general, the results conformed to the hypothesis. Participants generally perceived greater similarity between jokes that differed at more concrete or superficial KR dimensions. However, similarity ratings of jokes differing in terms of SI and in terms of LM are not in the predicted order, which has prompted much discussion in the literature about the validity of the LM KR (e.g., Davies, 2011; Hempelmann, 2004, 2009; Hempelmann & Attardo, 2011).

COMPUTATIONAL APPROACHES TO HUMOR Is it possible to program a computer to generate and/or understand humor? Although researchers in the field of artificial intelligence (AI) have, for the most part, ignored humor, it can be argued that any attempt to develop a truly intelligent computer system will ultimately need to address the problem of humor. Graeme Ritchie, a Scottish linguist and AI researcher now at the University of Aberdeen, along with his students and colleagues, suggested that AI investigations of humor can not only help to clarify theories of humor, but can also lead to important discoveries about human intelligence, language, problem-solving, and information processing more generally (e.g., Binsted & Ritchie, 1997, 2001; Ritchie, 2001, 2004, 2007). Ritchie (2001, 2004) has advocated an “experimental AI” approach, involving the use of computer programs to test cognitive (and particularly linguistic) theories of humor. In order to adapt a theory to a computer program, the theory needs to be specified in a formal, precise and detailed manner, conforming to the principles of generative linguistics and AI. Thus, AI investigations provide a way of “sniffing out” fuzzy thinking, vague conceptual definitions of central concepts, and faulty logic that might not otherwise be apparent

THE PSYCHOLOGY OF HUMOR

156

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

in theoretical formulations. Unfortunately, according to Ritchie, most of the existing humor theories are too vague and imprecise for computational application. For example, Ritchie (1999) criticized the traditional incongruity-resolution theories (discussed in Chapter 2: Classic Theories of Humor), pointing out that the ideas of incongruity and resolution have not been defined clearly enough, and that different theories define concepts in different ways. In particular, although they appear to be very similar, Shultz’s (1976) theory (which Ritchie refers to as the “surprise disambiguation model”) is, on close analysis, actually quite different from Suls’ (1972) theory (the “two-stage model”). The two theories have different implications and apply to different classes of jokes (see also Ritchie, 2007). One reason for the vagueness and imprecision of many theories, according to Ritchie, is that they attempt to explain too many different types of humor. Ritchie strongly rejects the quest for a “grand theory of humor” at the present time, arguing instead that we need to identify specific subclasses that can be thoroughly characterized and implemented on a computer. Only after we have done this with multiple subclasses can we build up a comprehensive theory that accounts for all kinds of humor. Accordingly, Ritchie has narrowed his focus to verbal jokes, and even more narrowly to certain types of jokes that share specific verbal mechanisms (e.g., punning riddles). Although one could theoretically attempt to develop a program that is capable of processing verbal texts that are fed into it and determine whether they are funny, Ritchie suggests that the more practical place to begin is with programs that apply a given theory to generate humorous texts. Human judges can then determine whether the output of the program is indeed humorous. By observing the behavior of the program (i.e., the types of jokes it produces), one can obtain useful insights into the weaknesses of the theory underlying it. This can then lead to further refinements of the theory and corresponding “tweaking” of the program. Thus, the goal of this sort of programming enterprise is not so much the program itself, but the refinement of the theoretical ideas underlying it. Ritchie has argued that attempts to implement cognitive and linguistic theories of humor in computer programs are beneficial to psychologists as well as linguists by providing a way of testing theories and alerting theorists to weaknesses in their models. Kim Binsted and Graeme Ritchie (1997) have taken this approach in developing a computer program called JAPE (Joke Analysis and Production Engine) that generates a specific class of jokes known as punning riddles. These are question answer jokes that are based on a pun (e.g., What’s the difference between a hairy dog and a painter? One sheds his coat, the other coats his shed.). Binsted and Ritchie began by developing a formal model of the punning mechanisms underlying these types of riddles, identifying a set of symbolic rules about the meaning combinations and textual forms involved. These rules were then built into a program that also has access to a large natural language lexicon (dictionary) of the kind used in AI research generally. This lexicon contains many words, along with information about their phonetic pronunciation, lexical usage, and syntactic meaning. It is important to note that this lexicon does not contain any information that could be conceived as inherently “funny.” Nonetheless, by searching through the lexicon for suitable word pairs that meet the criteria described by the rules, and applying various basic templates of riddle structure, the program can generate a virtually limitless number of novel riddles.

THE PSYCHOLOGY OF HUMOR

COMPUTATIONAL APPROACHES TO HUMOR

157

The following are some examples of the funnier riddles that were generated by JAPE (from Ritchie, 2004): • What do you call a ferocious nude? A grizzly bare. • What do you get when you cross breakfast food with a murderer? A cereal killer. • What’s the difference between leaves and a car? One you brush and rake, the other you rush and brake. • What’s the difference between a horse and a wagon? One bolts and jumps, the other jolts and bumps. Binsted, Pain, and Ritchie (1997) conducted a study to evaluate the output of JAPE, using a sample of 8- to 11-year-old children as judges. They presented these subjects with a random selection of JAPE-produced riddles, human-produced riddles (taken from published joke books), nonsense nonjokes, and sensible nonjokes. The children were asked to determine whether each text was a joke and, if so, how funny it was and whether they had heard it before. The results showed that the JAPE-produced riddles were identified as jokes just as reliably as the human-produced ones, and both were easily distinguished from the nonjokes. Although the JAPE-produced jokes were rated as less funny, on average, than the human-produced jokes, many of the JAPE riddles were rated as being just as funny as those produced by humans. Further analysis of the less funny riddles produced by JAPE may lead to future refinements of the program and, at the same time, a more precise linguistic theory of this type of humor. In addition to the JAPE program, Binsted and Ritchie (2001) analyzed the structure and formal regularities of another class of joke, which they referred to as “story puns,” and offered some suggestions about a possible computational model for their production. Ritchie (2004) also described other computer programs that have been developed by other researchers using a variety of approaches. As one example, Bruce Katz (1993) took a connectionist approach in developing a neural network model of incongruity in humor that attempted also to incorporate concepts of arousal, sexual and aggressive themes, and hedonic tone (i.e., mirth). Ritchie (2004) has also criticized Raskin and Attardo’s GTVH for being too vague for a computer implementation. However, humor researchers have recently begun to use Ontological Semantic Technology (OST) as a way of implementing GTVH in computer programs designed to recognize or detect humor in texts rather than creating jokes (e.g., Hempelmann et al., 2012; Raskin et al., 2009; Raskin, Taylor, & Hempelmann, 2010; Taylor, 2010; Taylor & Raskin, 2012). OST is built on “repositories” of world and linguistic knowledge comprised of a combination of automatic and human computation. The knowledge of these repositories represents the concepts of words and sentences and relationships between them, and it is used by computer programs to disambiguate the different meanings of words and sentences to represent them comprehensively (Hempelmann et al., 2012; Taylor & Raskin, 2012). Thus, when applied to humor, OST represents the meaning of all components of a joke. A semantic analyzer uses these types of repositories to recognize jokes by means of the ontological knowledge of the world. Raskin and colleagues have used OST to reconsider the six GTVH knowledge resources and to refine their definitions and conditions of use (Hempelmann et al., 2012). See Box 5.1 for a description of ontological semantic approaches to computational humor.

THE PSYCHOLOGY OF HUMOR

158

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

BOX 5.1

ONTOLOGICAL SEMANTIC APPROACHES TO C O M P U TAT I O N A L H U M O R B Y D R . C H R I S T I A N HEMPELMANN Christian F. “Kiki” Hempelmann is Associate Professor of Computational Linguistics and Director of the Ontological Semantic Technology Lab (www.tamuc.edu/ontology) at Texas A&M University Commerce. He received his doctorate from Purdue University with a dissertation on computational pun generation in 2003. His research focuses on computational semantics, computational humor, punning, and general humor research. Kiki has published widely on these topics over the last 20 years and regularly presents on them at international conferences and as an invited speaker. He has also been a consultant for the Internet search industry since 2005.

Computational algorithms have been designed for the detection, analysis, and generation of humorous texts. One approach to these tasks is approximating them based on observed patterns in large amounts of humorous vs. nonhumorous text collections called corpora. Computers can be trained to detect these patterns, and even reproduce them to some degree by feeding them, e.g., a few thousand jokes marked as such and a similar number of nonjoke texts of similar length. This approach is referred to by the misnomer “machine learning.” Another type of approach uses more-or-less complicated resources, i.e., lexicons and other knowledge bases and rules operating on the knowledge bases for processing humor computationally. Very simple approaches “generate” jokes by picking words from various lists and inserting them into a joke template. Complex meaning-based approaches, like the Ontological Semantic Theory of Humor (OSTH; Raskin, Hempelmann, & Taylor, 2009), aim to actually process the meaning of texts, including the meaning of their humor. They are based on a model of the world, its event types, objects, their properties, and any other relevant context. These models are called ontologies. All the concepts in rich ontologies are connected to each other. For example, forks are known to be instruments of eat, which is an event type done by any living organism, including any human, who is a subclass of living organism. With thousands of such concepts—with tens of thousands of properties and connections among them—the computer can select the intended meanings of sentences by selecting word meanings that describe possible scenarios in the world that the ontology models. Thus, the meaning of “sulfuric acid can eat through most metals” won’t allow for the sense of eat described above and acid as its agent, but will use the concept labeled corrode with acid as the instrument instead. This process of selecting such intended meaning combinations is called disambiguation and is the most important step in computational semantic (meaning) analysis. The specific task of processing humorous meaning, which is under incremental development (Hempelmann et al., 2012; Taylor & Raskin, 2013; Hempelmann & Petrenko, 2015; see

THE PSYCHOLOGY OF HUMOR

COMPUTATIONAL APPROACHES TO HUMOR

BOX 5.1

159

(cont’d)

also the work by Veale, e.g., 2012), involves a special type of disambiguation. The reason is that humorous texts intentionally carry more than one meaning, and these meanings, usually two of them in a special constellation of opposition, don’t allow for excluding all but one correct one of them. Thus, the words can’t be mapped onto a single set of related concepts. Take the following joke example with multiple puns in it: You can tune a guitar, but you can’t tuna fish. Unless it’s a bass. The first clause conforms with the knowledge of stringed instruments being the theme of “tune” events, but the second clause contains the heterographic (spelled differently) and homophonic (sounding identically) pun-target pair [“tune a”/“tuna”] fish. Humans buy into a relationship between words that are identical or similar in sound, which creates an overlap between tuning fish and the tuna fish subclass of fish. The opposing contrast between stringed instruments and fish should not be disambiguated by a humor-detecting algorithm, but needs to be identified as a potentially humorous contrast. This knowledge needs to be part of the ontological knowledge base. But the joke has another, aligned, pun in it in the homographic (spelled the same) but heterophonic (sounding similar, not identical) puntarget pair “bass” (guitar)/“bass” (fish). The reverse relationship of spelling and pronunciation to the previous pun-target pair is likely to be enhancing the funniness, but clearly beyond what can, or should be, modeled computationally at this stage in the development of the OSTH. Detecting the aligned contrast between another subclass of stringed instrument (bass guitar), and another subclass of fish (bass fish), is already within the purview of ontology-based computational humor. In sum, the type of approach sketched here has lexicons for the languages that it is intended to process, the word meanings of which are defined in terms of ontological concepts. With the help of these resources—lexicons and an ontology—and humor-specific properties like oppositions and script-like groupings in the richly connected ontology, the technology is evolving toward the ability to process various types of natural language humor. This development has taken knowledge-based computational humor from cookiecutter template filling to generating puns to generating jokes with a different mechanism and closer toward the ability of analyzing (in contrast to the more confined task of generating) humor in unrestricted natural language input.

Although computational models such as JAPE and OST appear to be quite promising, Ritchie (2001, 2004) acknowledged that they are still at a very early stage of development. The implementation rules underlying these programs are not tied to any real hypotheses about humor in general, and it is not clear how to generalize from computational models to other forms of humor. In addition, a complete computational model of humor will ultimately require the development of truly intelligent systems with a vast foundation of encyclopedic knowledge coupled with sophisticated reasoning abilities. Nonetheless, Ritchie contends that steps can be taken toward this goal by breaking the problem into

THE PSYCHOLOGY OF HUMOR

160

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

smaller chunks, identifying specific classes of humor, and developing rigorous formal descriptions that can be implemented using existing technology. “The overall message,” states Ritchie (2001, p. 132), “is that endeavoring to develop computational models of humor is a worthwhile enterprise both for AI and for those interested in humor, but we are starting from a very meager foundation, and the challenges are significant.” To be psychologically relevant, however, it is important that the computer simulations carry out the tasks in the same way that humans are assumed to do. For example, although computer chess programs can outplay most of the best human players, they operate very differently than human chess players, and are therefore not a very good test of cognitive theories of human chess-playing. Similarly, it is not entirely clear that programs like JAPE generate humor in the same way that humans do. Ritchie’s recommendation for more narrowly focused theories applied to discrete types of humor may also be a useful suggestion for psychological humor research, although this arguably depends on the goals of the individual researcher. If the goal is to identify general characteristics of humor that distinguish it from other human activities, then broader, more general theories may be appropriate. On the other hand, if the goal is to describe in detail how people cognitively process specific forms of humor, then greater progress will likely be made with research aimed at testing specific hypotheses derived from narrowly focused theories. However, for the purposes of understanding psychological aspects of humor, it may not be as necessary to make such fine-grained distinctions (e.g., distinguishing between several different classes of puns), and psychologists may find it useful to partition the humor domain (“carve nature at its joints”) in different ways than do AI researchers. In any case, for the psychologist, advances in AI research on humor may be viewed as a rich source of potential hypotheses for further experimental research.

EFFECTS OF HUMOR ON COGNITION In this chapter, thus far, we have examined cognitive processes that are involved in humor comprehension. We now turn to a discussion of the possible effects of humor on other cognitive phenomena, focusing particularly on creativity and memory.

Creativity Many theorists and researchers have noted a close relationship between humor and creativity. Koestler (1964) considered humor, along with scientific discovery and artistic creation, to be forms of creativity, all of which involved the common process of bisociation (see Chapter 2: Classic Theories of Humor). Just as humor theorists see elements like incongruity, surprise, and novelty as necessary elements of humor, creativity theorists see them as defining elements of creativity (e.g., Besemer & Treffinger, 1981). Both humor and creativity involve a switch of perspective, a new way of looking at things (O’Quin & Derks, 1997). Indeed, many creativity researchers consider humor to be essentially a type of creativity. Consequently, some measures of creative ability or creative personality include assessments of humor among their items (e.g., G. A. Davis & Subkoviak, 1975; Torrance,

THE PSYCHOLOGY OF HUMOR

EFFECTS OF HUMOR ON COGNITION

161

1966). O’Quin and Derks (1997), however, suggested that although there are close theoretical links between the two, creativity and humor should be considered as two separate but partially overlapping constructs. In the next section, we consider the role of creativity as an underlying trait that fosters humor production ability. Here, though, we focus on the reverse relationship, addressing the question, “does exposure to humor cause people to be more creative in their thinking?” There are at least two possible mechanisms by which humor may be expected to affect creativity stated here as hypotheses: 1. The Flexible Thought Hypothesis: The flexible thought processes and activation of multiple schemas involved in the processing of incongruities in humor could facilitate the flexible and divergent thinking required for creativity (Belanger, Kirkpatrick, & Derks, 1998). 2. The Positive Emotion Hypothesis: The positive emotion (i.e., mirth) associated with humor could reduce tension and anxiety, encouraging thinking, and enhancing the ability to relate and integrate divergent material (Isen, Daubman, & Nowicki, 1987). Several experiments have provided considerable evidence that exposure to humor increases people’s creative potential. Israeli psychologist Avner Ziv (1976) compared scores of 10th grade students on two tests of verbal creativity after they had either listened to a recording of a popular comedian or engaged in a nonhumorous activity. Compared to the controls, those in the humor condition scored higher in total creativity as well as specific components of creativity: fluency, flexibility, and originality. In the 1980s, psychologist Alice Isen and her colleagues conducted a series of studies demonstrating facilitative effects of positive emotion on creativity (Isen et al., 1987; Isen, Johnson, Mertz, & Robinson, 1985). They assessed creativity using a variety of methods including Mednick and Mednick’s (1959, 1962) Remote Association Test (RAT), in which participants must identify a concept that links two seemingly unrelated words. Although Isen and her colleagues framed their studies in terms of the effect of positive mood on creativity, they often induced positive mood by exposing participants to comedy films. Their findings revealed that exposure to comedy resulted in greater creativity compared to exposure to emotionally neutral or negative stimuli. These findings are consistent with the Positive Emotion Hypothesis because nonhumorous methods of inducing positive emotions also enhanced creativity. Thus, it seems that the creativity-enhancing effects of humor are likely due, at least in part, to effects of mirth (i.e., the emotional component of humor) on cognition rather than to a more cognitive mechanism (e.g., by activating multiple schemas, humor produces more flexible, less constrained thinking). More research, however, is necessary to fully test the mediating role of positive emotions. Recently, researchers have begun to consider the possibility that different humor styles can differentially affect creativity (Chang, Chen, Hsu, Chan, & Chang, 2015; Chen, Su, & Ye, 2011). Chang et al. (2015), for instance, tested two competing hypotheses about the mechanisms by which humor styles might relate to creativity. First, the Positive Emotion Hypothesis described above is predicated on findings that momentary positive affect can enhance creativity (see Baas, De Dreu, & Nijstad, 2008 for a review) and that people high in the two adaptive humor styles (affiliative and self-enhancing) tend to experience greater

THE PSYCHOLOGY OF HUMOR

162

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

positive affect. Thus, people high in affiliative and self-enhancing humor styles should exhibit greater creativity because they experience more positive affect. Alternatively, the intrapersonal variability hypothesis proposes that people who possess opposing or conflicting traits tend to exhibit more creativity. Consistent with this hypothesis, Kim, Zeppenfeld, and Cohen (2013) found that people high in internal conflict completed a task in a more creative manner than those low in personal conflict. Junior high school students in Taiwan completed Martin et al.’s (203) humor styles questionnaire, a measure of creative attitudes (divergent thinking), and a version of the Chinese Creative Thinking Test in which participants were given 10 minutes to use a Chinese character to complete as many different drawings as possible. Supporting the intrapersonal variability hypothesis, Chang et al. found that people who were high in all four humor styles (“general humor endorsers”) exhibited greater creativity than those who were high in only the positive humor styles (“positive humor endorsers”), those who were high in only negative humor styles (“negative humor endorsers”), or those low in all four humor styles (“humor deniers”). Thus, it seems that diverse humor styles could be more essential for creativity than positive affect or positive traits. Researchers have also examined the relationship between creativity and individual differences in disposition toward ridicule and being laughed at. Chan, Chen, and Lavalee (2013) examined the relationship between gelotophobia, gelotophilia, and katagelasticism, and creativity. They found that gelotophobia, the fear of being laughed at, correlated negatively with individual differences in creativity and, as a result, negatively related to performance on a creativity task (gelotophobes tend to be less creative by nature and thus exhibit less creativity on a given task). Gelotophilia, the joy of being laughed at, in contrast, positively correlated with creative disposition and creative task performance (gelotophilies tend to be more creative by nature and thus exhibit greater creativity on a given task). Lastly, they found no significant relation between katagelasticism (the joy of laughing at others) and creativity. In summary, it seems clear that humor affects creativity. First, momentary exposure to humorous stimuli enhances creative thinking. Second, people who engage in both positive and negative styles of humor to a higher degree in daily life tend to be more creative. The mechanism by which humor affects creativity is less clear, however. Current research suggests that humor might increase creative thinking by inducing positive emotion or affect. However, research has yet to explicitly demonstrate the mediating role of positive affect. In addition, it appears that the positive emotion hypothesis does not account for the relationship between humor styles and creativity. Perhaps different mechanisms underlie the effect of momentary exposure to humorous stimuli on creativity and the way the chronic use of humor in daily life affects creativity. Further research is necessary to more fully investigate these questions.

Memory Does humor enhance memory? More specifically, is humorous material remembered better than nonhumorous material? Many studies have found that humor does indeed

THE PSYCHOLOGY OF HUMOR

EFFECTS OF HUMOR ON COGNITION

163

serve as a memory aid for material presented in educational settings (e.g., Kaplan & Pascoe, 1977; Kintsch & Bates, 1977; Ziv, 1988), for advertisements (e.g., Duncan, Nelson, & Frontzak, 1984; Krishnan & Chakravarti, 2003), and for material presented in experimental research settings (e.g., Chambers & Payne, 2014; Lippman & Dunn, 2000; Schmidt, 1994; Schmidt & Williams, 2001). Not all research, however, has demonstrated a positive “humor effect” (see Weinberger & Gulas, 1992, for a review). Thus, researchers have sought to better understand the underlying mechanisms or psychological processes by which humor affects memory. There are several mechanisms (framed below as hypotheses) by which humor might work on the encoding or initial processing of information to affect memory: 1. Rehearsal Hypothesis: People might engage in more rehearsal of humorous versus nonhumorous information during initial processing (Slamecka & Katsaiti, 1987). 2. Arousal Hypothesis: Humor might induce an emotional response or physiological arousal, which itself positively affects memory (e.g., Burke, Heuer, & Reisberg, 1992). 3. Contextual Surprise Hypothesis: People might be surprised to encounter humorous information in a specific context and thus pay greater attention to it. 4. Incongruity Hypothesis: People might attend more and think more about humorous material because it involves an incongruity whose resolution requires extensive information processing. Empirical studies have systematically tested these underlying mechanisms. Steven Schmidt, a psychologist at Middle Tennessee State University, for instance, conducted a series of six experiments to examine the effects of humor on memory for sentences (Schmidt, 1994). Schmidt constructed humorous and nonhumorous sentences that were equivalent in other important qualities: they were equally bizarre, and thus contextually surprising, equally difficult to understand, and equally meaningful. Schmidt found that when humorous and nonhumorous sentences were presented in separate lists, participants recalled each type of sentence equally well. However, participants recalled humorous sentences better than nonhumorous sentences when the list of sentences contained both humorous and nonhumorous sentences. In fact, when the list contained both humorous and nonhumorous sentences, participants recalled humorous sentences better and nonhumorous sentences worse compared to conditions in which each type of sentence was presented alone. Thus, participants remembered humorous sentences better at the expense of remembering nonhumorous sentences. Schmidt (1994) ruled out the contextual surprise hypothesis to account for the memory advantage for humorous material when presented with nonhumorous material because the degree of surprise potential was held constant between them. Schmidt further suggested that the findings were consistent with the rehearsal hypothesis. Schmidt and Williams (2001) further examined the effects of humor on memory, testing the rehearsal and incongruity hypotheses. Schmidt and Williams presented participants with original cartoons, cartoons with altered captions that removed both incongruity and humor, or “weird” cartoons with altered captions that retained incongruity but not humor. They found that participants recalled the gist of original cartoons better than nonhumorous or weird cartoons. Participants, however, did not differentially recall specific details of the cartoons such as the exact wording of the captions. Further, participants recalled the

THE PSYCHOLOGY OF HUMOR

164

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

gist of original cartoons better than the weird and nonhumorous cartoons following “intentional learning” instructions to maximize rehearsal and “incidental learning” instructions to minimize intentional rehearsal. These findings rule out the incongruity hypothesis because participants recalled the gist of original cartoons better than weird cartoons that also contained incongruity. Further, they recalled the gist of weird and nonhumorous cartoons about equally. These findings also argue against the rehearsal hypothesis because the effect of humor on memory occurred in conditions of incidental and intentional learning. Based on the findings of Schmidt (1994) and Schmidt and Williams (2001), Schmidt (2002) proposed a new explanation by which encoding and retrieval processes work together to create a memory advantage for humorous material under certain conditions. Schmidt expanded on Waddill and McDaniel’s (1998) spontaneous elaboration hypothesis, that unusual or distinctive information in a specific context requires the activation of more background knowledge in order to interpret and understand it. Thus, unusual information receives more elaborate information processing or encoding. Schmidt (2002) added that to the extent that humorous material is relatively uncommon, it would likely receive greater elaboration during initial encoding. Schmidt referred to this as the “context-dependent elaboration hypothesis.” In addition, McDaniel, DeLosh, and Merritt (2000) suggested that people more easily retrieve distinctive information relative to more common information. Thus, humorous material might be afforded both an encoding and a retrieval advantage that work together to give it a memory advantage when it is distinctive (e.g., presented together with nonhumorous material). Schmidt (2002) replicated previous findings showing that participants remember the gist of humorous cartoons better than nonhumorous or weird cartoons, and that this effect occurred when participants read a list containing each type of cartoon. These effects also occurred when participants did not expect a memory test. These results corroborate previous findings, further ruling out the rehearsal and incongruity hypotheses. Schmidt also measured participants’ heart rates to examine the role of physiological arousal in the humor effect on memory. Contrary to the arousal hypothesis, participants did not show heart rate acceleration upon initially processing humorous cartoons versus the other two types of cartoons. Thus, physiological arousal does not appear to be the critical mechanism responsible for the effect of humor on memory. However, there is some evidence that arousal might enhance memory effects. Peter Derks and his colleagues at the College of William and Mary used experimental procedures like those of Schmidt (1994) to examine potential memory effects of “tendentious” (i.e., sexual and aggressive) humor compared to nontendentious humor (Derks, Gardner, & Agarwal, 1998). They partially replicated Schmidt’s findings of the memory-enhancing effects of humorous material, and found a strong effect for tendentiousness, indicating that emotionally arousing elements such as sex and aggression further enhance the memory advantage for humorous material. Schmidt’s (2002) results did reveal a greater heart rate deceleration in response to the humorous cartoons compared to the others, which Schmidt suggested was associated with ease of discrimination. He suggested that during the encoding process, the humorous cartoons were “marked” as different from the others (p. 136). Overall, then, Schmidt’s (2002)

THE PSYCHOLOGY OF HUMOR

EFFECTS OF HUMOR ON COGNITION

165

findings support his context-dependent elaboration hypothesis. It appears that the combination of encoding processes that differentiated humorous material from other material combined with retrieval processes that favor humorous material provides the most complete explanation for these and previous findings. Carlson (2011) noted that Schmidt’s (2002) context-dependent elaboration hypothesis does not propose that there is something special about the perception of humor that facilitates memory. Rather the perception of distinctiveness is critical, and humorous material is often distinctive. Thus, according to the context-dependent elaboration hypothesis, people should recall nonhumorous material as well as humorous material provided it is equally distinctive in a given context. However, Carlson also found that people recall humorous material better than nonhumorous, distinctive material (i.e., inspirational material), suggesting that there is indeed something special about the perception of humor that facilitates memory. Summerfeldt, Lippman, and Hyman (2010) offered an alternative to Schmidt’s (2002) context-dependent elaboration hypothesis by more explicitly delineating how humor facilitates encoding and retrieval of information. They argued that to recall information such as a line in a poem or song, one must preserve other characteristics such as the rhythm and rhyme scheme (Rubin, 1995). An accurate reconstruction of a line in the poem or in a song is constrained and thus aided by the characteristics of rhythm and rhyme. Humorous material such as jokes could be constrained by restrictions imposed by the punch line, rhyme scheme, set up expectations, joke context, etc., that function together as memory cues to help people reconstruct the joke. So, while humorous material might indeed be distinctive in each setting (Schmidt, 2002), it also might come with unique constraints for accurate reconstruction that serve as effective memory cues. Summerfeldt et al.’s constraint explanation might account for why humor seems to facilitate memory for the gist of humor material but not specific details. It seems likely that constraints of a punch line and humor context would serve as memory cues for the essence or general substance of a joke, cartoon, or statement rather than specific wording or other details. Similarly, the constraint explanation might account for an interesting finding that humor in advertising leads to better recall of an advertisement, but worse recall for a specific product (e.g., Hansen, Strick, van Baaren, Hooghuis, & Wigboldus, 2009). Perhaps humor can act as a constraint that aids the accurate reconstruction of the general essence of an advertisement as a whole, but not the specific product promoted in the advertisement. In summary, empirical research provides convincing evidence that people recall the general essence of humorous information better than nonhumorous information when both are presented in the same context. Researchers have proposed several mechanisms by which humor might affect the initial encoding of information to facilitate memory. However, it appears that humor probably works on both encoding and retrieval processes. Because humorous information is unusual or distinct in many contexts, people elaborate on it more during the initial processing and thus more easily retrieve it than more common information. In addition, humorous material appears to contain its own restrictions or constraints that aid in the memory of the general essence of the material.

THE PSYCHOLOGY OF HUMOR

166

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

HUMOR AS A COGNITIVE ABILITY As we discussed in Chapter 4, The Personality Psychology of Humor, humor scholars have conceptualized individual differences in humor in terms of humor appreciation and the use of humor in daily life. Relevant to the cognitive psychology of humor, scholars also have considered individual differences in humor production ability: one’s aptitude or skill for saying, doing, or creating things (e.g., making jokes, telling stories) that others find funny (O’Quin & Derks, 1997). People who are gifted with this creative talent are presumably the amateur comedians who keep their friends “in stitches” and are the “life of the party,” while the supremely talented few may become professional comedians and comedy writers. Traditionally, researchers have assessed humor production ability by giving people an opportunity to make up something funny. Perhaps the most common method involves asking people to generate captions for cartoons, often New Yorker cartoons (e.g., Babad, 1974; Brodzinsky & Rubien, 1976; Ziv, 1980). Ko¨hler and Ruch (1993), for instance, developed the Cartoon Production Punchline Test (CPPT) that includes three categories of caption-removed cartoons derived from Ruch’s (1992) 3 WD test of humor appreciation: incongruity-resolution, nonsense, and sexual. Researchers have also asked people to create funny captions for TAT cards (Day & Langevin, 1969), to generate witty word associations (Hauck & Thomas, 1972), and to make up funny presidential campaign slogans (Clabby, 1980). Recently, Howrigan and MacDonald (2008) developed an innovative new humor production task requiring participants to complete a blank resume for an odd or funnylooking person depicted in a photo by generating responses to typical resume categories or prompts (e.g., name, occupation, about me, my typical day, etc.). Nusbaum et al. (2017) also developed a new assessment of humor production ability called the Jokes Stems Task, which requires participants to write something funny to complete the set up of a story. For instance: Imagine that one of your classes this semester is incredibly boring, and you’re trying to convey just how boring this class is to one of your friends. So you say “Seriously, this class is so boring. . .” Please complete the phrase “Seriously this class is so boring. . .” with something funny.

In each of these humor production tests, the experimenters or independent judges rate the funniness of participants’ responses, yielding a score for humor production ability. Given the conceptual similarities between humor and creativity noted earlier, it is perhaps not surprising that researchers have found that humor production ability is strongly related to creativity (e.g., Brodzinsky & Rubien, 1976; Clabby, 1980; Fabrizi & Pollio, 1987). Brodzinsky and Rubien (1976), for instance, found that participants who generated funnier captions for cartoons also scored higher on the RAT (Mednick & Mednick, 1959, 1962) described above. Importantly, the RAT is thought to measure “convergent creative thinking” (Lee, Huggins, & Therriault, 2014; Smith, Huber, & Vul, 2013). Convergent creative thinking is involved in coming up with optimal strategies to solve problems (Cropley, 2006). Thus, funnier people tend to be better at seeing problem-solving strategies that lead to effective solutions to problems. O’Quin and Derks (1997) conducted a meta-analysis of this research and found an average correlation of 0.34 between humor production ability

THE PSYCHOLOGY OF HUMOR

HUMOR AS A COGNITIVE ABILITY

167

and creativity as measured by the RAT. O’Quin and Derks concluded that, although creativity and humor production do involve similar mental processes, they are nonetheless distinct. Whereas humorous productions are typically creative, individuals can be creative without being funny. Although humor production ability and creativity are clearly related, the nature of that relationship remains unclear based on previous research. It is possible that there is a direct relationship between creativity and humor production, such that creativity is a critical cognitive ability involved in humor production. However, as O’Quin and Derks (1997) noted, humor production ability and creativity might be related due to the common influence of a third variable such as intelligence. Indeed, intelligence and creativity are related to one another (Jauk, Benedek, Dunst, & Neubauer, 2013; Silvia, 2008; 2015), and humor production ability positively correlates with intelligence (e.g., Greengross & Miller, 2011; Greengross et al., 2012; Howrigan & MacDonald, 2008; Koppel & Sechrest, 1970). Thus, it is possible that because of the conceptual overlap between creativity and intelligence, neither predicts humor production ability on its own when taking the other into account. Accordingly, Kellner and Benedek (2017) conducted a study to better delineate the unique roles that creativity and intelligence play in humor production. Kellner and Benedek devised their own humor production test (HPT) based on Ko¨hler and Ruch’s (1993) CPPT. The HPT consisted of visual cartoons and asked participants to come up with funny captions for the cartoons. As in the CPPT, the HPT reflected three different types of cartoons: incongruity resolution, nonsense, and sexual. To better understand how intelligence relates to humor production, they assessed intelligence conceptualized as fluid intelligence and crystallized intelligence (Horn & Cattell, 1966). Fluid intelligence refers to the ability to think and reason abstractly; whereas crystallized intelligence refers to knowledge acquired from prior learning and experiences. The two forms of intelligence work together in daily life. For instance, when taking a statistics test, you would use fluid intelligence to devise problem-solving strategies; you would use crystallized intelligence to recall and execute the right formulas. Kellner and Benedek also assessed creativity conceptualized as divergent creative thinking rather than convergent thinking. Divergent creative thinking involves spontaneous, unconstrained generation of ideas or solutions to a problem (e.g., brainstorming; Cropley, 2006). Kellner and Benedek replicated previous findings showing that humor production ability correlated at a statistically significant level with creativity and overall intelligence: participants who came up with funnier cartoon captions also scored higher in intelligence and creativity. Interestingly, Kellner and Benedek’s study adds to previous research showing that proficiency in divergent creative thinking, as well as convergent creative thinking, seem to be critical skills underlying the ability to be funny. Additional analyses showed that both intelligence and creativity played unique roles in explaining humor production when taking the other into account. Both intelligence and creativity are important, independent components of humor production. Finally, Kellner and Benedek also showed that humor production ability was more strongly related to crystallized intelligence than with fluid intelligence. This finding is consistent with those of Greengross and Miller (2011) who also found that humor production ability correlated more strongly with a measure of crystallized intelligence than with a measure of fluid intelligence. Together these findings more clearly delineate how intelligence relates to humor ability. It appears that acquired

THE PSYCHOLOGY OF HUMOR

168

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

knowledge, particularly in the subtleties of language, plays a critical role in being funny (p. 56). In addition to creativity and intelligence, various personality traits also appear to underlie humor production ability. For example, Robert Turner (1980) examined the association between humor production ability and self-monitoring, a personality trait having to do with the degree to which people are sensitive to environmental cues of social appropriateness and regulate their behavior accordingly. Turner assessed humor production ability in two ways. Participants performed the usual cartoon caption completion task, and a second task in which they created 3-minute comedy monologues to describe miscellaneous objects (i.e., tennis shoes, a wristwatch, and a box of crayons). For each task, judges rated the funniness of participants’ responses. As predicted, higher self-monitors produced funnier responses on both tasks. Turner suggested that attending to social cues and reactions enables high self-monitors to develop the skill to say and do funny things. In contrast, low self-monitors, because they are less attentive to others, do not learn as readily from those responses and therefore do not develop as much skill at producing humor. Consistent with these results, other research has found a positive correlation between self-monitoring and a self-report measure of humor production ability (Bell, McGhee, & Duffey, 1986). Recall from Chapter 4, The Personality Psychology of Humor, that openness to experience (one of the Big Five personality traits) also positively relates to humor production ability (Nusbaum et al., 2017). People higher in openness to experience are better than others at being funny. Overall, then, humor production is a skill that people develop over the course of their lives that requires crystallized intelligence and creativity in the form of both convergent and divergent thinking. Furthermore, certain personality traits, namely self-monitoring and openness to experience, appear to contribute to the development of the ability to produce humor.

SUMMARY AND CONCLUSION Cognitive psychology has made an important impact on the psychological study of humor. In fact, the most influential comprehensive theories of humor can be described as cognitive theories: they borrow concepts and theory from cognitive psychology to describe the mental processes involved in the perception and interpretation of a stimulus event, identifying perceptions of incongruity as a cognitive mechanism that is minimally necessary for all humor. In this chapter, we discussed important contributions to the cognitive psychology of humor from a broad network of related disciplines including linguistics, psycholinguistics, and computer science. Indeed, many important theoretical advances in recent decades have originated in linguistics. We first described some of the methods cognitive psychologists have developed to measure the strength of semantic association between concepts in memory, and thus investigate how people store knowledge or meaning of concepts in their minds. In the mid-1950s psychologists developed self-report questionnaires to measure the semantic distance between concepts, and then in the 1970s they began to take advantage of computer technology to develop several more sophisticated techniques (e.g., the lexical decision task) to

THE PSYCHOLOGY OF HUMOR

KEY CONCEPTS

169

study schema activation in real time. We then discussed how psychologists and psycholinguists have used these techniques to investigate the cognitive mechanisms (e.g., perception of incongruity, incongruity resolution) that underlie the perception of jokes and other forms of conversational humor such as irony and sarcasm. In our consideration of linguistic approaches to the study of humor, we noted that linguists have contributed to the cognitive psychology of humor by developing models of text analysis that fully and precisely specify the rules for processing textual information that are necessary and sufficient to produce humor. Ideally, semantic theories would reach a level of specification that is both precise and broad enough that theorists can represent them computationally, i.e., use the theory to program a computer to make sense of any humorous text and to distinguish humorous from nonhumorous text. The most influential linguistic theories have been Raskin’s (1985) script-based semantic theory of humor and Attardo and Raskin’s (1991) general theory of verbal humor. The cognitive psychology of humor also includes investigations of the ways humor affects or relates to other cognitive processes such as creativity and memory. It is apparent that exposure to humorous stimuli and endorsement of each of the four humor styles relate positively to creative potential. It is also clear that people show better memory for the general essence of humorous information than nonhumorous information when both are presented in the same context, and that humor probably facilitates memory by aiding both encoding and retrieval of information. Finally, relevant to the cognitive psychology of humor, scholars have conceptualized humor (particularly humor production) as a cognitive ability. Research suggests that the development of humor production ability (the ability to be funny) requires crystallized intelligence and creativity in the form of both convergent and divergent thinking. Furthermore, certain personality traits, namely self-monitoring and openness to experience, appear to contribute to the development of the ability to produce humor. The field of humor research studies many different questions that are relevant to the cognitive psychology of humor. As noted throughout this chapter, there are many research questions and hypotheses derived from a variety of theories that warrant empirical investigation. Further research on questions related to the cognitive psychology of humor might not only provide a better understanding of the ubiquitous phenomena of humor, but might also shed light on other more basic questions of interest to psychologists, such as the interface between cognition and emotion, comprehension of ambiguous meaning, and cognitive aspects of nonverbal as well as verbal interpersonal communication. Research questions relating to cognitive aspects of humor could form the basis of a good many Masters and Ph.D. theses for years to come.

KEY CONCEPTS • • • • •

Perceptions of incongruity Activity factor Evaluative factor Potency factor Within-domains distance

THE PSYCHOLOGY OF HUMOR

170 • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

5. THE COGNITIVE PSYCHOLOGY OF HUMOR

Between-domains distance Lexical decision task Priming effect Cross-modal lexical priming task Word fragment completion test Irony Graded Salience Theory Processing equivalence hypothesis Intensifiers Constraint-Satisfaction Model Script-based Semantic Theory of Humor General Theory of Verbal Humor Knowledge resources JAPE (Joke Analysis and Production Engine) Flexible Thought Hypothesis Positive Emotion Hypothesis Intrapersonal Variability Hypothesis Rehearsal Hypothesis Arousal Hypothesis Contextual Surprise Hypothesis Incongruity Hypothesis Humor Production Ability Cartoon Production Punchline Test Joke Stems Task Convergent creative thinking Fluid intelligence Crystallized intelligence Divergent creative thinking Self-monitoring Openness to experience

CRITICAL THINKING 1. As discussed in the current chapter, humor often generates an increase in individuals’ creativity. However, the mechanism by which this increase is caused is not yet fully known. What do you think would be a good explanation for the relationship? Why? 2. The use of technology to study humor has been a prevalent theme in the field of cognitive psychology. How else could technology or social media platforms be used to study styles of humor? What types of humor would they best be applied to? How might these topics/studies change the field of cognitive psychology? 3. Vaid and colleagues made use of priming tasks to study humor comprehension and schemas. The results of their 2003 study were inconclusive: the initial schemas were activated during the first two joke segments, the second schemas were active during the

THE PSYCHOLOGY OF HUMOR

CRITICAL THINKING

171

second segment, but neither schema was active at the punch line. Why might these results have occurred? What variable(s) might mediate this effect? 4. In the last section of this chapter, humor is described as a cognitive ability, one that is affected by intelligence, self-monitoring, and openness to experience. If humor can be affected by these things, then could humor also affect these traits? Why or why not? Based on your opinion, formulate a hypothesis and then design a study to test your hypothesis. Include conditions, treatments, measures, participants, and predicted results.

THE PSYCHOLOGY OF HUMOR