53059852

  • Upload
    mi-ab

  • View
    216

  • Download
    0

Embed Size (px)

Citation preview

  • 7/28/2019 53059852

    1/28

    Humor comprehension, humor production,and insight: An exploratory study

    AARON KOZBELT and KANA NISHIOKA

    Abstract

    The relation between humor comprehension and humor production was exam-

    ined. Unlike earlier studies that used comprehension tasks requiring a produc-

    tive component, here participants simply decided whether cartoons were

    matched or mismatched to presented latent content (i.e., the implied mean-

    ing of the cartoon). To test humor production, the same participants devised

    humorous captions for photographs, which judges reliably rated on funniness.

    Performance on the two tasks showed a signicant positive correlation.Rela-tions between response time, condence, and funniness ratings in the compre-hension task were consistent with an insight view of humor comprehension:

    correct responses were made more quickly, and matched latent content trials

    were identied faster and were rated as funnier than mismatched trials. In the

    humor production task, judged funniness showed a reliable (but rather small)

    positive correlation with response time, offering little support for an insight

    view of humor production. Limitations of the study and possibilities for future

    research on humor production, modeled on creativity research, are discussed.

    Keywords: Humor comprehension; humor production; insight; latent content;

    signal detection theory.

    1. Introduction

    Over the past century, hundreds of studies have investigated many aspects of

    humor (Roeckelein 2002). However, investigations of the cognitive processes

    involved in humor, especially humor production, remain comparatively scarce.What can we say about the mental processes that occur when people generate,

  • 7/28/2019 53059852

    2/28

    376 A. Kozbelt and K. Nishioka

    understand, or appreciate humorous statements? How much do people vary in

    humor comprehension or humor production abilities? How are these related to

    each other or to other aspects of cognition? What is the time course of cognitive

    processes in humor comprehension or humor production? To what extent cansuch ill-dened constructs be reliably and validly measured in the rst place?

    This study addresses such questions.

    Most theoretical and empirical work on cognitive processes in humor has

    aimed to understand humor appreciation or humor comprehension (e.g., Cou-

    turier et al. 1981; Cunningham and Derks 2005; Eysenck 1942; Feingold 1983;

    Feingold and Mazzella 1993; Goldstein 1970; Koestler 1964; Martin 1996;

    Sheehy-Skefngton 1977; Vaid et al. 2003). Humor appreciation is the experi-

    ence of nding something amusing. It is typically operationalized by the inten-

    sity and duration of the mirth response, including smiling and laughing, or

    by subjective funniness ratings provided in response to humorous stimuli. Hu-

    mor comprehension is the process of understanding or getting a joke. It is

    typically assessed by determining whether participants can correctly interpret

    a cartoons meaning, using open-ended or multiple-choice questions. In prin-

    ciple, appreciation and comprehension should be positively associated: one

    should not nd an instance of humor funny if one does not understand it. Psy-

    chometric evidence is largely consistent with this view, nding at least some

    reliable, positive correlations between humor appreciation and comprehension(Byrne 1956; Masten 1986; Wierzbicki and Young 1978; for a review, see

    Kaufman et al. 2008).

    Many humor theorists and researchers have emphasized unexpectedness

    and surprise as important, or even necessary, factors in appreciating or under-

    standing humor. For example, Gestalt theory claims that a sudden restructur-

    ing, or change in interpretation, occurs during humor comprehension (Derks

    1987; Gick and Lockhart 1995; Koestler 1964; Maier 1932). Empirical re-

    search also shows that jokes are rated as funnier when participants take a

    shorter time to appreciate them (Cunningham and Derks 2005; Goldstein

    1970), with response time and funniness ratings showing an essentially nega-

    tive linear relation. Thus, comprehending or appreciating humor does not ap-

    pear to require protracted effort; indeed, the opposite appears to be true.

    This perspective on humor appreciation and humor comprehension suggests

    a link to cognitive psychological research on insightproblem solving (Gick

    and Lockhart 1995; Perkins 1981: 62-64; for a general overview, see Stern-

    berg and Davidson 1995). When an insight occurs, the answer to a difcult

    problem is hit upon quite suddenly, with an associated emotional A-ha! reac-tion. One can think about insight problems in contrast to more mundane, incre-

  • 7/28/2019 53059852

    3/28

    Humor comprehension and production 377

    mental forms of problem solving. For instance, as people work to solve algebra

    word problems, their feeling of warmth, or sense of nearness to a solution,

    increases. In contrast, when a problem requires an insight, feeling of warmth

    ratings remain at until shortly before solution (Metcalfe and Wiebe 1987).Also, just as getting a joke is usually a rapid process, solutions to insight prob-

    lems generally occur quickly, usually in the rst minute of problem solving

    (Lockhart et al. 1988). While taking up to a minute to solve a problem is hardly

    instantaneous, the important feature of insight problem solving is that there is

    not a longer-term speed-accuracy tradeoff, as with problems that can be solved

    incrementally: after about the rst minute, additional working time does

    not benet the problem solver. Thus, although in some cases elaboration

    may be necessary to solve an insight problem even when an appropriate prob-

    lem representation has been achieved (Weisberg and Alba 1981), success or

    failure in insight problem solving seems largely due to the basic choice of a

    proper problem representation. Typically, an insight-yielding representation

    will differ from the default representation, necessitating some kind of re-

    representation if a solution is to be found, as in Gestalt views of humor com-

    prehension. Gick and Lockhart summarized the relation between insight and

    humor comprehension:

    The processes involved in insight share features with those involved in understandingjokes. Typical jokes set up representations of concepts that must be revised before the

    joke can be understood. Application of the revised concept must be fairly automatic;

    otherwise the joke is not funny if too much explanation is required. (Gick and Lockhart

    1995: 224)

    While insight and the processes of humor comprehension and/or appreciation

    have often been associated, little empirical research has studied whether in-

    sights happen when people make jokes. Indeed, in general contrast to the lit-

    erature on humor appreciation and comprehension, relatively few researchers

    have examined the cognitive processes involved in humorproduction: Rocke-

    leins (2002) comprehensive reference anthology of humor research, running

    to nearly 600 pages, includes only a few pages on humor production.

    In many respects, the difculties facing psychologists wishing to research

    humor production are isomorphic to those facing creativity researchers (Derks

    1987; Kaufman et al. 2008; Murdock and Ganim 1993; OQuin and Derks

    1997). Both can be considered forms of ill-dened problem solving, in which

    a person must rst determine that a problem exists, dene or represent the prob-

    lem in a particular way, generate and evaluate potential novel solutions, andelaborate a basic idea into a denitive form. In ill-dened problems there is no

  • 7/28/2019 53059852

    4/28

    378 A. Kozbelt and K. Nishioka

    single correct answer, as there might be for objectively scorable, well-dened

    situations. However, there are usually better or worse solutions, and the quality

    or creativity of a particular solution can be determined based on reliable, consen-

    sual assessment by qualied judges (Amabile 1983). Given these parallels, it issurprising that so few psychological investigations of the cognitive processes

    of humor production have been undertaken, in contrast to creativity, which,

    though still not a completely mainstream psychological topic, has spawned a

    considerable and varied research literature (see, e.g., Sternberg 1999). Interest-

    ingly, however, several investigations (Brodzinsky and Rubien 1976; Fabrizi

    and Pollio 1987; Humke and Schaefer, 1996; Treadwell 1970; Ziv 1976, 1988)

    have found positive associations between the quality of humor production and

    more general measures of creativity.

    While the specic underlying cognitive mechanisms for the link between

    humor production and creativity are unclear, a plausible candidate involves

    processes involved in insight problem solving. Indeed, some eminent creators

    describe prototypical insight-like experiences at moments of great creative dis-

    covery, for example, Darwins report of his sudden realization, while reading

    Malthus, that natural selection provided a mechanism for biological evolution

    (Perkins 1981). Researchers have frequently linked insight and creativity (see,

    e.g., Csikszentmihalyi and Sawyer 1995), particularly in the phase of creative

    thinking where an idea rst emerges, rather than in the later protracted effort ofelaborating that idea into a nished product. In the context of creative problem

    solving (and, perhaps, by extension, humor production), insight can be charac-

    terized in terms of its underlying cognitive processes in various non-mutually

    exclusive ways (Schooler et al. 1995). For instance, one possibility is a Gestalt

    process comparable to perceptual gap-lling; another is search through a prob-

    lem space which offers few cues for progress (but which, once found, quickly

    lead to a solution) or which may have been previously dened as to preclude

    discovering a solution. In common with an insight view of humor comprehen-

    sion, such insight processes in creative or productive situations occur rapidly

    and involve modifying a default problem representation of a situation.

    Since humor production represents a main focus of our investigation, we

    now review some previous investigations on the topic. One study is that of

    Derks and Hervas (1988), who examined the relation between the rated funni-

    ness of newly created humorous picture captions, the number of captions pro-

    duced per picture, and the order in which they were produced. They found that

    captions were rated funnier when participants produced ten captions per pic-

    ture, rather than two per picture, and that rated funniness increased with outputorder. Derks and Hervas concluded that a greater quantity of humor production

  • 7/28/2019 53059852

    5/28

    Humor comprehension and production 379

    can lead to higher quality of humor. Using a similar methodology, Derks

    (1987) tested three models of humor production: rather rapid stimulus-response

    associations, restructuring patterns of wholistic concepts toward some guiding

    end, which should take longer for funnier jokes, and computations from a com-plex network of connections without a wholistic goal, which should show very

    variable latencies. Derks found some support for all three models: an initial

    burst of ideas supported the rst model, an even distribution of the best jokes

    throughout the session supported the third model, and the nding that when the

    worst idea occurred, the best one would likely follow it was consistent with the

    second model. The initial burst of the rst model might also be interpreted as a

    function of humor expertise involving pattern matching with a repertoire of

    familiar jokes in a well organized, structured knowledge base in long-term

    memory. Consistent with this view, Siegler (2004) found that, compared to

    novices, expert comic writers made more detailed interpretations of photos for

    which they were to create funny captions, were more likely to create humor by

    mapping a structural relation to make a humorous transformation of their ini-

    tial interpretation of the photo, and tended to activate schemas to write their

    captions. While these investigations are informative in some respects, since

    participants in the studies by Derks (1987) and Derks and Hervas (1988) gen-

    erated numerous captions per picture, the fairly coarse temporal measures in

    these investigations yield little information about another question, namely, thetime course of the production of individual instances of humor and their rela-

    tion to judged humor value.

    Another aspect of humor production was studied by Turner (1980), who

    found that participants who scored high on a self-monitoring scale were judged

    as wittier by members of a discussion group. This nding suggests a metacog-

    nitive component to humor production: perhaps people who are more sensitive

    to whether others will understand and appreciate humor can better evaluate

    jokes likely chances of success. This interpretation has somewhat different

    process implications than the quantity-driven aspect of Derkss (1987) and

    Derks and Hervass (1988) conclusions. In particular, the ability to evaluate

    potential jokes may be as important as the ability to generate jokes in the rst

    place, suggesting a link between humor comprehension and humor production

    abilities. Indeed, some theoretical perspectives (e.g. Attardo 1994; Feingold

    1983) largely equate generative and explicative aspects of humor, arguing that

    if one can explain why something is funny, then one can understand how it is

    produced.

    Other studies have focused on the relation between humor productionand other abilities. For instance, Masten (1986) studied the sense of humor of

  • 7/28/2019 53059852

    6/28

    380 A. Kozbelt and K. Nishioka

    children aged 10 to 14, measuring appreciation (by mirth response and funni-

    ness ratings), comprehension (by two judges ratings of participants open-

    ended explanations of cartoon humor), and humor production (by two judges

    assessments of the quality of humor of newly devised cartoon captions and ti-tles). Masten found a positive correlation between production and comprehen-

    sion and between each of these and appreciation, but only when measured by

    mirth response (not by funniness ratings). While this investigation is one of the

    few to look directly at the relation between humor comprehension and produc-

    tion (as well as appreciation), it is an open question whether results obtained

    with children would generalize to adults. Also, naturally such measures do not

    inform the time courses of any cognitive processes involved in humor.

    In another study, Feingold and Mazzella (1993) examined the relation be-

    tween humor comprehension and production in adults. They used self-report

    ratings of wittiness and several measures of humor production, including car-

    toon captioning, answers to absurd questions, and repartee generation, and

    compared performance on these tests to measures of humor cognition (joke

    knowledge and joke reasoning) and humor communication (willingness to

    share captions created in the humor production test) as well as measures of

    sociability, verbal ability, and scholastic orientation. As in Mastens (1986)

    study, two judges reliably rated the humor of the captions or statement created

    in the humor production tasks. Results showed that humor production waspositively correlated with measures of humor cognition and verbal ability, but

    not to the other constructs, consistent with Sieglers (2004) ndings on schema

    activation in expert comic writers. It also reinforces the idea that humor produc-

    tion and humor comprehension may be related, at least to the extent that Fein-

    gold and Mazzellas measure of humor cognition is a valid measure of humor

    comprehension. However, their operationalization of humor cognition, joke

    knowledge and joke reasoning involved ll-in-the blank questions that actually

    required a considerable amount of production (cf. Masten 1986), rather than

    simple recognition or detection processes. Indeed, Feingold and Mazzella ac-

    knowledged that their humor cognition measure assessed only the ability to

    manipulate incongruities . . . not to recognize them or to evaluate ones pro-

    ductions (1993: 453454). Ideally, a pure measure of humor comprehension

    would not involve any productive component, but only recognition or detec-

    tion of humor in a format that could be objectively scored.

    Other studies of humor production and appreciation (but not comprehen-

    sion) include Koppel and Sechrest (1970) and Khler and Ruch (1996), neither

    of whom found strong evidence for a link between the two constructs. Koppeland Sechrest (1970), in a study of college fraternity brothers, found only a mild

  • 7/28/2019 53059852

    7/28

    Humor comprehension and production 381

    positive correlation between humor production (measured by ratings of par-

    ticipants newly created cartoon captions) and humor appreciation (measured

    by participants own ratings of the funniness of a set of cartoons). Khler and

    Ruch (1996), studying a large sample of adults, found only very low positivecorrelations between humor appreciation and production, as measured by peer-

    ratedperformance criteria (rather than self-reported humor initiation).

    In sum, while investigations of humor production have yielded some valu-

    able initial results, many questions remain. One concerns the time course of

    such processes and whether ner-grained temporal measures would implicate

    sudden, insight-like processes, particularly in humor production, but also in hu-

    mor comprehension or appreciation. Previous research (e.g. Cunningham and

    Derks 2005; Gick and Lockhart 1995; Goldstein 1970) suggests commonalities

    between insight and humor comprehension or appreciation. However, the rela-

    tion between insight and humor production remains unclear, since earlier studies

    examining temporal aspects of humor production (Derks 1987; Derks and

    Hervas 1988) examined them in the context of generating multiple instances

    of humor, rather than the time course of the genesis of an individual humorous

    idea.

    A second remaining question concerns the psychometric relations between

    humor production, comprehension, and appreciation. Earlier research sug-

    gests that humor comprehension is related to both humor production (Feingoldand Mazzella 1993; Masten 1986) and humor appreciation (Byrne 1956;

    Masten 1986; Wierzbicki and Young 1978), but that humor production is

    not strongly related to humor appreciation (Babad 1974; Fabrizi and Pollio

    1987; Khler and Ruch 1996; Koppel and Sechrest 1970; OQuin and Derks

    1997). However, few studies have examined these constructs on the same sam-

    ple of individuals, and studies yielding these results have used a wide variety

    of assessment and measurement techniques, not all of which are necessarily

    ideal.

    Addressing these two sets of questions requires considerable methodologi-

    cal care. In general, much humor research is limited by methodological short-

    comings; indeed, some researchers (e.g. Khler and Ruch 1996; Sheehy-

    Skefngton 1977) have explicitly noted the need for much more sophisticated

    techniques for measuring and assessing the construct of humor. Since humor is

    subjective, it is not immediately obvious how to obtain useful measures of

    humor comprehension or production. For instance, a reasonably objective

    measure of the correct interpretation (or at least a widely agreed-upon inter-

    pretation) of a joke must be derived to yield a meaningful measure of humorcomprehension; otherwise, the comprehension measure is just self-report.

  • 7/28/2019 53059852

    8/28

    382 A. Kozbelt and K. Nishioka

    Also, an ideal humor comprehension measure would not include any produc-

    tive component, as was the case with Feingold and Mazzellas (1993) ll-in-

    the-blank questions and Mastens (1986) open-ended explanations of cartoon

    humor. Instead, it would require only recognition or detection processes; other-wise, any correlation with humor production could be an artifact of the way

    humor comprehension was measured. Moreover, even with a valid compre-

    hension measure, care must be taken to distinguish a participants sensitivity

    to humor from any response bias. This can be achieved using a signal detec-

    tion paradigm (e.g. Macmillan and Creelman 1991) that treats hits, misses,

    false alarms, and correct rejections separately. Similarly, to assess the

    humor value of newly produced jokes, one must have reliable and valid

    measures of the funniness of the jokes, as rated by an appropriate sample of

    judges. Ideally, judgments would be made by a fairly substantial sample of

    individuals, rather than a small number (e.g. the two judges used by Feingold

    and Mazzella 1993; Masten 1986). A larger set of judges would reduce con-

    cerns about potential biases in judgment, even if the smaller number of raters

    showed respectable reliability. Finally, to examine the nature and time course

    of the processes of individual instances of humor comprehension and produc-

    tion, it is helpful to have response time measures that are more ne-grained

    than those used in some earlier research (e.g. Derks 1987; Derks and Hervas

    1988).In the present study, we attempt to address all of these methodological points.

    We use a Humor Comprehension Task and a Humor Production Task, both of

    which involve captioned cartoons or photos. The basis for correct answers in our

    Comprehension Task is rooted in Freuds (1953, 1960) idea of latent dream-

    thoughts or latent content (LC). In psychoanalysis, LC usually refers to the

    hidden meaning of a dream, joke, or fantasy, in contrast to the manifest con-

    tent, which involves conscious, overt surface features that are clearly evident.

    Presently, following Erdelyi (1985, 1999, 2006; Erdelyi and Stein 1981), we

    dened LC as the implied meaning or message of the cartoon. According to

    Erdelyis reading of Freud, jokes and humor represent a particularly interesting

    aspect of LC, since jokes are at least somewhat conscious, as well as voluntary

    and playful, in contrast to dreams, daydreams, or fantasies (1985: 170175). In

    this view, LC interpretation is central to humor comprehension and, moreover,

    is non-idiosyncratic: without the interpretation of the latent content there is no

    getting a joke the interpretation is the getting. The interpretation of latent

    content is not chancy. Subjects claiming to get a joke usually come up with the

    same interpretation (Erdelyi in press). Thus, studying jokes also has method-ological advantages.

  • 7/28/2019 53059852

    9/28

    Humor comprehension and production 383

    As an example of the role of LC in humor, Erdelyi cites:

    the comment of an aspish socialite from a bygone era on the goings on at Yale after

    football games: If all the girls at Yale were laid from one goalpost to the other, I

    wouldnt be at all surprised. . . . The remark about Yale would hardly be funny if the

    lady had said bluntly that after football games the girls visiting Yale had sexual inter-

    course with the Yale boys. (Erdelyi 1985: 171173)

    In this view, a crucial part of the humor process is the joker censoring or oth-

    erwise disguising the LC in some way, just as the process of discovering the

    LC is crucial for the comprehension of the listener. Finally, while this particu-

    lar joke involves rather tendentious and taboo LC, the same principle of getting

    a joke by understanding its LC also applies to more innocent instances of hu-

    mor, in which the LC is innocuous. In such cases, comical effects depend on

    the form or style of the joke, rather than on racy content (Erdelyi 1985: 174).

    In sum, our view of LC is not narrowly psychoanalytic, but simply involves

    reading between the lines to tease out the tacit meaning or message of the

    cartoon by putting together information from the cartoon image and its caption.

    Indeed, besides psychoanalytic views of LC, one might also mention humor

    theories claiming that jokes have an obvious and a hidden meaning, for exam-

    ple linguistic theories distinguishing inference and implicature (e.g. Grice 1975).

    To use LC as a means to investigate humor comprehension, we add to cap-tioned cartoons a second caption that is either matched or mismatched to the

    cartoons LC. If participants understand or get the cartoon, they should un-

    derstand the reason for its humor and thus be able to identify matched versus

    mismatched LC. If not, LC identication should be at chance levels. Thus, us-

    ing matched versus mismatched LC represents a much more objective measure

    of humor comprehension than self-report data on getting a joke, assuming

    that the matched versus mismatched LC classication is valid in the rst place.

    Moreover, the data are analyzed using a signal detection model, which allows

    us to distinguish each participants sensitivity to LC, denoted as d', from their

    response bias. Thus a signal detection analysis of sensitivity to LC represents

    a reasonably pure measure of humor comprehension that can then be compared

    to a measure of humor production. Moreover, besides assessing correct versus

    incorrect answers, measures of RT and participants condence in their an-

    swers and funniness ratings of each cartoon will be collected. Previous theory

    and research (e.g. Cunninngham and Derks 2005; Gick and Lockhart 1995;

    Goldstein 1970; Koestler 1964; Maier 1932) suggest that humor comprehen-

    sion and insight share many properties; we expect results in line with this view,and our measures should allow us to identify insight-like processes, for

  • 7/28/2019 53059852

    10/28

  • 7/28/2019 53059852

    11/28

    Humor comprehension and production 385

    in age from mid-20s to mid-50s. Some degree of extraversion and a good sense

    of humor were thought desirable for this study. Therefore, participants were

    initially recruited mainly by referrals of funny, extraverted personal acquain-

    tances of the experimenters, who then referred others in a snowball sample.Participants were ultimately drawn primarily from entertainment elds and

    most were actors, music performers, or writers. While one could argue that

    selecting participants based on extraversion and a sense of humor might lead to

    a restriction of range in humor ability potentially resulting in lower correla-

    tions, we felt that it was more important to select individuals who would be

    comfortable in the task and would likely produce captions judged as funny. As

    will be seen below, our humor comprehension and humor production measures

    did generate sufcient variability to yield reliable correlations.

    2.1.2. Stimuli. Stimuli were single-frame, black-and-white, captioned car-

    toons from The New Yorkermagazine. These were scanned electronically and

    standardized to a width of 6 inches using Adobe Photoshop 5.0.2. The original

    captions were retyped in black 12-pt font underneath the image, so that they

    would be clear and uniform across the cartoons. Underneath the original cap-

    tions, latent content (LC) statements were added in a different 12-pt font, in

    blue. In the debrieng, none of the participants indicated that they were previ-

    ously familiar with any of the particularNew Yorkercartoons used in the study.LC for each cartoon was either correctly matched to the cartoon (i.e. the LC

    actually captured the implied meaning or message of the joke) or mismatched

    to the cartoon (i.e. the LC did not capture the implied meaning or message of

    the joke). For example, in one cartoon, a man and his pet cat are standing next

    to the cats litter box. The man says to the cat, Never, ever think outside the

    box. The matched LC is Inappropriately creative. (The joke, of course, in-

    volves a play on the expression think outside the box as a way of describing

    creativity. Here, the box is literally a (litter) box, and doing things outside the

    litter box that should be done inside creates a problem!) The mismatched LC is

    A man trying to talk rationally to a cat. In another cartoon, a busy execu-

    tive is talking on the phone and consulting his calendar, saying, No, Thursday

    is out. How about never is never good for you? The matched LC is Some

    people dont get the idea. (The joke, of course, is that the executive is exas-

    perated and strongly hinting that he does not want to meet with the other

    person.) The mismatched LC is He also cares for how it ts into the other

    partys schedule. In each of these cases, the mismatched LC does not really

    explain capture the meaning or message of the cartoon, while the matched LCdoes.

  • 7/28/2019 53059852

    12/28

    386 A. Kozbelt and K. Nishioka

    For each cartoon, matched and mismatched LC statements were produced

    by the experimenters, in consultation with Matthew Hugh Erdelyi and Jacque-

    lyn Bergstein (see Erdelyi 1985, 1999, 2006; Bergstein 2005). Typically, ve

    or so candidate LC statements were generated per cartoon. Over 100 cartoonsand LC statements were reviewed for their funniness, understandability, and

    LC ts. In a pilot study, veparticipants who were otherwise unconnected with

    the study rated whether and how well each candidate LC was matched or mis-

    matched to the cartoons. The cartoons on which judges most strongly agreed

    about matched versus mismatched LC were selected for use in the Comprehen-

    sion Task. The agreed-upon best-matched LC was counted as the correct

    answer for each cartoon. The least well-matched LC served as the mismatched

    version of each of these cartoons.

    In all, 20 cartoons were chosen as stimuli, each one in a matched LC and

    mismatched LC version, for a total of 40 stimuli. These were randomly divided

    into two sets, each containing 20 cartoons, half of which had matched LC and

    half of which had mismatched LC. During the experimental session, no par-

    ticipant saw the same cartoon image more than once.

    2.1.3. Procedure. Each participant received written instruction for the

    Comprehension Task. Participants were told they would see cartoons on the

    computer, with captions in black and LC in blue. LC was dened as the impliedmeaning of the cartoon, which will generally not be explicitly stated in the car-

    toon itself. Participants were asked to determine if the LC matched the cartoon

    or not. To clarify further, they were also told to pay attention to the interpretation

    a LC implies, rather than whether the LC exactly matched the contentof what

    was shown in the cartoon. Participants were told to try not to be inuenced by

    the funniness of the cartoon or LC and to respond to the matched or mismatched

    LC as quickly as possible once they made their decision, since timing was of

    interest. Participants were also told to try to use the full range of the scales across

    all of the cartoons. Finally, participants were not told what proportion of car-

    toons would have matched or mismatched LC. Participants were randomly

    assigned to one of the two sets of stimuli and began the task when they indi-

    cated that they understood the procedure.

    A custom-designed computer program was used to show stimuli and obtain

    responses. Each participant worked individually on a Dell Latitude C840lap-

    top computer. Participants were rst given practice with two cartoons with

    matched versus mismatched LC. To make the task as clear as possible, car-

    toons that were judged as very understandable and easy to answer in the pilottest were used in the practice trials.

  • 7/28/2019 53059852

    13/28

    Humor comprehension and production 387

    Cartoons in the actual Comprehension Task were presented serially in a ran-

    dom order. The cartoon remained visible throughout each trial, during which

    three questions were presented under the cartoon, one at a time. The rst ques-

    tion was: Is the latent content matching the cartoon? Participants responded byclicking a Yes or No button on the screen, and response time was recorded.

    The next question was: How condent are you of your answer? Participants

    responded on a scale of 1 to 10 (1 = not at all condent, and 10 = very con-

    dent).Finally theywere asked: How funny was the cartoon? Participants re-

    sponded on a scale of 1 to 10 (1 = not at all funny, 10 = very funny). After par-

    ticipant answered the third question, the next cartoon appeared. Participants

    continued until they had rated all 20 cartoons in the set.

    2.2. Production task

    2.2.1. Participants. The same people who participated in the Comprehen-

    sion Task participated in the Production Task, immediately afterwards, using

    the same laptop computer.

    2.2.2. Stimuli. The main stimuli were a series of twenty color photographs

    of people in everyday activities or social situations. These served as the basis

    for participants generation of humorous captions. Over 100 copyright-freephotographs were initially sampled from Fotki, a website where people can

    post their own photos and have others look at them online (http://www.fotki.

    com). In a separate pilot study, two acquaintances of the experimenters gener-

    ated multiple funny captions for the candidate photos. Twenty photos that led

    to rapid, relatively easy generation of funny captions were chosen from the

    initial pool and used as stimuli in the Production Task. Photos were standard-

    ized to a height of 7 inches using Adobe Photoshop 5.0.2.

    The Production Task itself was preceded by four example trials in whichparticipants saw photos with humorous captions already included. These were

    selected from a website showing numerous such captioned photos, the Found

    Slides Foundation (http://www.lostandfrowned.com/foundsl2.html). In a sep-

    arate pilot study, four judges viewed 31 captioned photos directly on the web-

    site and completed a survey rating each one on understandability and funni-

    ness. The four captioned photos rated highest on these criteria were chosen for

    the example trials.

    2.2.3. Procedure. Each participant received written instructions for theProduction Task. After the four example trials, participants were told they

  • 7/28/2019 53059852

    14/28

    388 A. Kozbelt and K. Nishioka

    would see 20 photos on the computer, one at a time in a random order, and that

    they should devise one funny caption for each photo. Participants were told to

    try to take up to about one minute per photo to devise a caption; after one min-

    ute, the computer screen would blink as a warning to try to nish soon. Par-ticipants were advised that ideally the caption should be funny both to them-

    selves and to others and that they should pick the rst caption that came to

    mind that they thought was acceptably funny. If participants had more than one

    caption idea for a photo, they were instructed to pick the one they thought fun-

    niest. Participants were told that when they had a pretty good idea of the word-

    ing of their caption, they should type it below the photo and then to click a

    submit button on the screen to make the next photo appear. As before, a

    custom-designed computer program was used to show stimuli and obtain re-

    sponses. In addition to recording the captions themselves, the program re-

    corded several measures of response time: the elapsed time from the appear-

    ance of the photo until the onset of typing and then from the onset of typing

    until the submit button was clicked. We refer to these intervals as pre-typing

    time and typing time, respectively; their sum is total time. Participants

    began the task when they indicated that they understood the procedure, con-

    tinued until they had created captions for all of the photos, and were then

    debriefed.

    2.2.4. Judgment of the captions. A different set of persons rated the funni-

    ness of the captions created in Production Task. Twelve persons (6 female, 6

    male) participated, all native English speakers, all from the New York metro-

    politan area, and ranging in age from early-20s to late 30s. We attempted to

    obtain judges from a similar population as before, since such judges are prob-

    ably more likely to nd the captions humorous.

    Judges received written instructions and were given a set of 20 papers on

    which they were asked to write their ratings. Each page contained all captions

    for a given photo generated by participants in the Production Task, listed in a

    random order. (For each photograph, the order of captions on the page differed

    across judges.) The 20 photos were identied by unique letter or number ID

    codes and were rst shown as small thumbnail images on the computer. Partici-

    pants were instructed to click on an image to enlarge it and then to nd the sheet

    with matching ID code. Participants rated the funniness of each caption on a

    scale from 1 to 8 (1 = not very funny at all, 8 = very, very funny). After rating all

    of the captions on a page, participants proceeded to the next picture on the com-

    puter and repeated the procedure until they had rated captions for all of thephotos. Judges had unlimited rating time but were asked not to over-think why

  • 7/28/2019 53059852

    15/28

    Humor comprehension and production 389

    they found any captions funny. After completing the task, participants were

    debriefed.

    2.2.5. Statistical technique: Rasch analysis. Because ratings in the Judg-ment task are ultimately subjective, one must take care in analyzing them. In

    particular, one must be sure that raters judgments show adequate reliability; if

    they do not, there is little basis for discussing meaningful individual differ-

    ences in humor production ability. Also, one must also acknowledge the pos-

    sibility that different judges might have idiosyncratic senses of humor or at

    least have different standards for what constitutes a funny caption. Taking such

    factors into account should improve the quality of the measure of humor pro-

    duction ability.

    Fortunately, these concerns can be addressed through the use of Rasch sta-

    tistical analysis (Rasch 1980; Wright and Masters 1982), implemented in the

    WINSTEPS software program (Linacre and Wright 2000). In Rasch analysis,

    the funniness of captions and harshness of judges are each assumed to vary

    intrinsically. Captions and judges are rst separately rank-ordered on these

    criteria. When combined, they should produce a meaningful pattern. For in-

    stance, harsh judges will rate only the funniest captions highly; lenient judges

    will rate many more captions highly, except perhaps the least funny ones. If

    this pattern does not generally hold, the intended variable loses its quantita-tive basis (Wright and Masters 1982: 4) and it is meaningless to talk about

    humor as a measurable construct. However, if such a pattern does hold, the

    Rasch analysis will converge, reecting an underlying uni-dimensional con-

    struct (here, the funniness of the captions).

    The Rasch procedure uses an iterative, maximum-likelihood process that

    minimizes the residuals of the differences between each caption and each judge

    until their positions on the dimension are stable. In particular, along the dimen-

    sion, each judge is located across from a caption that has an equal likelihood of

    receiving a high or a low rating. The unit of measure is the logit, denoted by

    = log / (1 ), which represents the log-odds probability () of a caption

    being rated funny by a particular rater. The Rasch procedure keeps this uniform

    over the range of observations, and a higher score indicates a funnier caption.

    If an underlying dimension exists, the captions and judges positions on the

    dimension will stabilize and the iterative Rasch procedure will converge; if

    such a dimension does not exist, the analysis will not converge. A successfully

    converged Rasch analysis yields interval-scale measures of the funniness of

    each caption and the harshness of each judge. The procedure also yieldsreliability indices for the analysis. The Rasch measures of all of the captions

  • 7/28/2019 53059852

    16/28

    390 A. Kozbelt and K. Nishioka

    produced by each participant can then be averaged, providing an overall ap-

    praisal of each participants humor production ability. The results will be

    roughly comparable to average raw scores, but the Rasch score will better mea-

    sure the underlying construct and achieve a more reliable result, as it maxi-mizes the t between captions and judges.

    3. Results

    From each participant, we collected 20 comprehension measures (360 total)

    and 19 original captions (342 total). Due to a computer error, only 19 of 20

    captions in the Production Task were collected per participant. However, the

    missing photo differed acrossparticipants, and we were still able to collect 14to 18 captions for each photo.

    3.1. Comprehension task

    The results reveal substantial dispersion among participants in correctly iden-

    tifying matched LC. TheM(SD) number correct out of 20 was 14.17 (2.33),

    with scores ranging from 10 to 19 correct out of 20. We also calculated each

    participants sensitivity to LC (d') using a signal detection model. Sensitivitywas measured using the ratio of Hits (correctly identied matched LC) over all

    potential hit items, and the ratio of False alarms (mismatched LC incorrectly

    identied as matched) over all items that should be correctly rejected (for de-

    tails on the computation, see Macmillan and Creelman 1991). For each par-

    ticipant, d'was calculated ve times, using progressively higher thresholds of

    condence ratings (1, 3, 5, 7, or 9 out of 10) to count as committed decisions.

    These were then averaged to yield an overall d'for each participant. The results

    echo the individual differences found using percent correct. TheM(SD) d'was

    1.47 (0.81), with scores ranging from 0.22 to 3.16. Indeed, the raw number

    correct and d'measures were strongly correlated, r(16) = .92,p < .0001.

    Analyses of response times (RT) also yielded some notable ndings. Be-

    cause the distribution of RT scores was positively skewed, RTs were rst trans-

    formed by the natural logarithm. Results showed that, across all captions, ln-

    transformed RT was negatively correlated with participants appraisal of each

    cartoons funniness, r(358) = .16, p = .002, and negatively correlated with

    participants condence in their answers, r(358) = .23,p < .0001, both small

    to medium effect sizes. Thus, when participants were able to comprehend LCmore quickly, they found the cartoon funnier and were more condent of their

  • 7/28/2019 53059852

    17/28

    Humor comprehension and production 391

    answer. Participants funniness ratings were also strongly correlated with their

    condence ratings, r(358) = .43,p < .0001, a medium to large effect size. This

    result probably means that people nd cartoons funnier if they understand

    them better, since lower condence about LC implies that participants did notfully get the jokes.

    Finally, for each participant, we divided responses into correct versus incor-

    rect and matched versus mismatched LC. Paired ttests were used to compare

    each type of response on ln-transformed RT, funniness, and condence; effect

    sizes were estimated by Cohens d. As suggested by Dunlap et al. (1996), the

    original standard deviations of each condition, not the standard deviation of the

    difference scores between conditions, were used in the computation of Cohens

    d. The results of these analyses are shown in Table 1. As can be seen, correct

    responses were made signicantly more quickly than incorrect responses, with

    a medium effect size. Correct and incorrect responses did not reliably differ in

    condence or funniness.

    More reliable results were found for comparisons of matched versus mis-

    matched latent content. In particular, matched LC cartoons were answered sig-

    nicantly faster and were judged as funnier than mismatched LC cartoons.

    Participants condence ratings for matched LC cartoons were marginally sig-

    Table 1. Within-subjects comparisons of all three dependent measures on correct versus incor-

    rect responses and matched versus mismatched latent content trials

    Correct vs. Incorrect Responses

    Correct Incorrect

    Measure M SD M SD t p Cohens d

    ln (RT) 2.50 0.32 2.63 0.26 2.39 .029 0.45

    Condence 7.82 1.09 7.59 1.04 1.30 .210 0.22

    Funniness 6.67 1.57 6.72 1.71 0.26 .798 0.03

    Matched vs. Mismatched LC

    Matched LC Mismatched LC

    Measure M SD M SD t p Cohens d

    ln (RT) 2.48 0.29 2.60 0.29 2.58 .020 0.41

    Condence 7.92 1.15 7.59 1.05 1.99 .063 0.30

    Funniness 7.08 1.38 6.29 1.77 4.04 .001 0.50

    Note: All df= 17. Cohens dcalculated using the original standard deviations of each condition,

    not the standard deviation of the difference scores between conditions.

  • 7/28/2019 53059852

    18/28

    392 A. Kozbelt and K. Nishioka

    nicantly higher than those for mismatched LC cartoons. Here funniness

    showed a medium effect size; RT and condence each showed small to me-

    dium effect sizes. Note also that since matched LC cartoons were rated as fun-

    nier than mismatched LC cartoons, conceivably participants could have usedfunniness information to make their response, rather than only relying on their

    basic comprehension of the cartoons and their assessment of the LCper se.

    Alternatively, the matched LC statements could simply have helped partici-

    pants make better sense of some of the cartoons, boosting funniness ratings via

    increased comprehension and indirectly contributing to the funniness advan-

    tage for matched LC cartoons. These alternatives cannot be decisively distin-

    guished by the present dataset and should be pursued in future research.

    These data also allow a participant-level comparison of humor comprehen-

    sion scores with degree of humor appreciation, which was operationalized by

    averaging each participants funniness ratings over all of the trials in the Com-

    prehension Task. Across all trials, funniness was marginally positively corre-

    lated with number correct, r (16) = .47, p = .052, and with d', r (16) = .46,

    p = .058. The basic pattern of results held when funniness ratings for matched

    versus mismatched items were separately correlated with humor comprehen-

    sion (measured by number correct or by d') and when funniness ratings for

    correct versus incorrect responses were separately correlated with humor com-

    prehension (measured by number correct or by d'). The resulting eight correla-tions, all positive, were nearly all marginally reliable and ranged from r= .40,

    p = .102, to r= .47, p = .049. Thus, humor appreciation and humor compre-

    hension appear to be at least mildly positively related.

    In sum, the results of the Comprehension Task show considerable individual

    variation in sensitivity to LC, a fairly strong positive correlation between rated

    funniness and condence, shorter RTs for correct versus incorrect responses,

    and faster and funnier ratings for cartoons with matched LC, compared to mis-

    matched LC. Participants who were more sensitive to LC also found the car-

    toons marginally funnier, and this effect held across trial type and response

    type.

    3.2. Production and judgment tasks

    As noted above, captions created in the Production Task were rated in the Judg-

    ment Task, and the rating data were entered into a Rasch statistical analysis.

    The iterative Rasch procedure did converge, indicating a stable underlying di-

    mension of humor production ability. Caption and judge reliabilities were .67and .99, respectively, indicating substantial agreement between raters in the

  • 7/28/2019 53059852

    19/28

    Humor comprehension and production 393

    Judgment task. These numbers are equivalent to Cronbach a and KR-20

    (Linacre and Wright 2000: 100) and indicate the extent to which captions reli-

    ably separate into different levels of funniness and judges reliably separate into

    different levels of harshness. Thus, these results provide a relatively rm basisfor examining meaningful individual differences in humor production ability.

    As expected, participants average Rasch Production scores were highly cor-

    related with their raw scores, r(16) = .94,p < .0001. TheM(SD) raw funni-

    ness scores for each participants captions were 2.76 (0.45), ranging from 1.77

    to 3.56. The M (SD) Rasch funniness scores for each participants captions

    were 42.46 (2.10), ranging from 37.79 to 45.24.

    Thus, there appear to be considerable individual differences in the average

    funniness of the captions produced by different participants. Is there any rela-

    tion between caption funniness and the amount of time taken for production?

    Recall that the Production task yielded three measures of RT: pre-typing time,

    typing time, and total time. The different RT measures gave somewhat varied

    results. As with Comprehension RT, each Production RT distribution was pos-

    itively skewed, so each was transformed by the natural logarithm. The correla-

    tion between funniness and ln-transformed pre-typing time was not signicant,

    r (340) = .06, p = .27. However, the correlations between funniness and ln-

    transformed typing time and between funniness and ln-transformed total time

    were reliable: respectively, r(340) = .11, p = .04, and r(340) = .14, p = .01,both small effect sizes. Since it is likely that participants continue to think

    about and rene their caption idea while they are typing, total time is probably

    the best measure of response time. There thus seems to be a slightly positive

    relation between the judged funniness of a caption and the time taken to create

    it. However, these small effects probably should not be over-emphasized.

    3.3. Comprehension and appreciation vs. production

    The nal sets of analyses examine the relations between each participants

    average performance in the Production Task and their LC performance and

    funniness ratings in the Comprehension Task. When each participants num-

    ber correct on the Comprehension Task was correlated with average Rasch

    Production scores, the correlation was positive and statistically reliable, r(16) =

    .62, p=.007, a large effect size. Note also that when raw scores on both

    measures were compared, the correlation remained signicant, r (16) = .51,

    p = .03, again with a large effect. A very similar result was obtained when d'was compared to Rasch Production scores, r(16) = .49,p = .04, again a large

  • 7/28/2019 53059852

    20/28

    394 A. Kozbelt and K. Nishioka

    effect.Interestingly, whend'was compared to raw Production scores, the cor-

    relation dropped substantially, r(16) = .31, p = .21, but remained a medium

    effect size. This somewhat discrepant nal result reinforces the utility of Rasch

    statistical analysis for establishing a sensitive measure of humor productionability. Overall, these results indicate that people who are better at getting

    jokes (i.e., in correctly identifying matched versus mismatched LC) appear to

    be better at making jokes.

    In contrast, no evidence for a correlation between humor appreciation and

    humor production was found. Specically, for participants funniness ratings

    in the Comprehension Task versus raw ratings of their captions in the Produc-

    tion Task, r(16) = .22,p = .38. For participants funniness ratings in the Com-

    prehension Task versus Rasch ratings of their captions in the Production Task,

    r(16) = .19,p = .44.

    4. Discussion

    The basic ndings are straightforward. Substantial individual differences were

    found in humor comprehension and humor production, and these two abilities

    were positively correlated. Judges rating newly created humorous captions re-liably agreed on the captions funniness, providing a rm foundation for this

    comparison. In itself, the basic relation between comprehension and produc-

    tion is consistent with earlier views and empirical ndings (Attardo 1994;

    Feingold and Mazzella 1993; Masten 1986) and thus may not be completely

    unexpected. However, demonstrating this relation with a comprehension task

    that involves no productive component (in contrast to Feingold and Mazzella

    1993; Masten 1986), as well as the large effect size observed in three of the

    four comparisons, is notable. More generally, this investigation demonstrates

    that humor comprehension and humor production abilities are both amenable

    to reliable, objective measurement. In particular, the use of a LC matching task

    to assess humor comprehension and Rasch measurement to rene humor pro-

    duction judgments appear to be methodologically advantageous aspects of the

    present study. Such methodological advances can help lay the groundwork for

    future research into the cognitive processes involved in humor (cf. Khler and

    Ruch 1996).

    A measure of humor appreciation, derived from the funniness ratings given

    by participants to cartoons in the Comprehension Task, was also examined.Humor appreciation was consistently (though only marginally) positively cor-

  • 7/28/2019 53059852

    21/28

    Humor comprehension and production 395

    related with both measures of humor comprehension, in line with theoretical

    arguments and other reports (Byrne 1956; Masten 1986; Wierzbicki and Young

    1978). In contrast, humor appreciation showed no reliable relation with humor

    production, again consistent with numerous other ndings (Babad 1974;Fabrizi and Pollio 1987; Khler and Ruch 1996; Koppel and Sechrest 1970;

    Masten 1986).

    In both the Comprehension and Production tasks, data from other measures

    (e.g. RT, condence, and funniness ratings) were also informative. In the Com-

    prehension Task, we found a fairly strong positive correlation between funni-

    ness and condence, shorter RTs and higher condence for correct versus in-

    correct responses, and faster and funnier ratings for cartoons with matched LC,

    compared to mismatched LC. These results support an insight view of hu-

    mor comprehension (e.g. Cunningham and Derks 2005; Gick and Lockhart

    1995; Goldstein 1970; Koestler 1964;Maier 1932), in which getting a joke is

    a fast, condent process with no speed-accuracy trade-off. Indeed, taking ad-

    ditional time to consider a joke was adversely associated with accurate humor

    comprehension.

    In contrast, RT data failed to support the notion that humor production is

    also an insight-like process. Such a conclusion would have been supported by

    a signicant negative correlation between RT and the judged funniness of each

    caption. However, two of the three RT measures showed reliable (thoughsmall) positive correlations with judged funniness. This suggests that to at least

    some extent, the more a person crafts a humorous caption, the better. The fact

    that typing time, rather than pre-typing time, drove the relation with funniness

    reinforces this view. However, as mentioned above, these effects, though sta-

    tistically reliable, are small.

    Despite this basically null result, our data do not conclusively rule out the

    idea that humor production, like humor comprehension, might have some sud-

    den, insight-like properties. Specically, the present methodology might not

    have been sensitive to the manner or timescale in which such processes may

    operate. Alternatively, one could collect dynamic feeling of warmth ratings

    (Metcalfe and Wiebe 1987) or verbal protocols (Siegler 2004) while partici-

    pants work on generating humorous statements or photo captions. Temporally

    sampling the process of humor production more densely might yield evidence

    for spikes of insight that represent the sudden onset of a humorous idea.

    Such a pattern might be more diagnostic of insight-like processes (likely in-

    volving abrupt changes of representation, as in humor comprehension) than

    looking for general speed-accuracy tradeoffs over the entire duration humorproduction trials, each of which can last up to a minute or more. Such an

  • 7/28/2019 53059852

    22/28

    396 A. Kozbelt and K. Nishioka

    approach would likely yield more detailed information about the nature and

    time course of cognitive processes in humor production.

    Such future work should also take into account the possibility that a fast re-

    sponse could be the result of rapid associations due to humor expertise, that is,pattern matching with a repertoire of familiar joke schemas in memory (Derks

    1987; Siegler 2004). Ruling out this possibility is not trivial, either method-

    ologically or conceptually, but modifying Derkss (1987) method of having

    participants produce multiple captions per photo might be one informative ap-

    proach. Specically, expert pattern matching should mostly happen early in the

    response window; in contrast, sudden, insight-like new associations could

    arise at any time. Eliminating participants early contributions and then focus-

    ing on feeling of warmth ratings leading up to later captions to the same picture

    might yield some evidence that would link insight and humor production.

    Moreover, producing written humor, as in the present study, may simply re-

    quire more revision and polishing than is possible when producing spoken

    humor (Feingold and Mazzella 1993: 440). Indeed, there is evidence that even

    some insight problems require considerable elaboration to reach a solution

    (Weisberg and Alba 1981). Thus, spoken humor production, including sponta-

    neous humor that occurs in interpersonal interactions, may show more evi-

    dence of insight-like processes than written humor production; this is another

    potential direction for future research. Other types of tasks besides picture orcartoon captioning would also reveal the extent to which the present results

    would generalize.

    While the present ndings are suggestive, there are limits to their interpreta-

    tion. For instance, one might question the generalizability of the results to a

    broad population, given the selection criteria of the present study, where par-

    ticipants were chosen based on extraversion and likely success at the humor

    production task. If anything, restriction of range should have diminished the

    correlation between production and comprehension. While some basic cogni-

    tive differences may exist between our participants and others, the fact that

    present results are consistent with numerous other reports increases condence

    that this is not the case. However, broader objections about attempts to mea-

    sure humor in settings that may not be ecologically valid (e.g. Babad 1974) are

    still relevant here, as with a great deal of humor research.

    Another limitation centers on understanding the reasons for the correlation

    between humor comprehension and humor production. One possibility is that

    cognitive processes in humor rely on domain-specic knowledge or abilities.

    In this view, the ability to understand jokes may be a prerequisite to producinggood jokes. Funny people might have ner assessment skills for both other

  • 7/28/2019 53059852

    23/28

    Humor comprehension and production 397

    peoples jokes and their own jokes and thus may use the same humor compre-

    hension skill to produce or select better jokes of their own (Turner 1980). They

    may also prefer more humorous entertainment forms, associate with more

    funny friends, or have richer humor schemas or a richer reserve of jokes.Sieglers (2004) expert-novice comparison suggests that comic writers possess

    such knowledge and schemas and can appropriately activate them to produce

    humor, and this is partly consistent with the ndings of Derks (1987). Feingold

    and Mazzellas (1993) ndings that humor cognition measures positively pre-

    dict humor production are likewise consistent with this view. Alternatively, a

    more general cognitive aptitude like verbal intelligence or the cognitive exi-

    bility to achieve insights through re-representation might also provide a com-

    mon explanation for the observed advantages of some persons in both humor

    comprehension and humor production. Indeed, a number of investigations

    have found positive associations between humor comprehension and verbal

    intelligence (e.g. Couturier et al. 1981; Feingold 1983; Feingold and Mazzella

    1993; Masten 1986; Wierzbicki and Young 1978), and a few have reported

    positive correlations between humor production and verbal intelligence (Fein-

    gold and Mazzella 1993; Koppel and Sechrest 1970; Masten 1986).

    Ultimately, the explanation for the positive correlation between humor com-

    prehension and humor production need not be unitary, and it would not be

    surprising if both domain-specic knowledge and a more general cognitiveaptitude contribute to this relation. Along these lines, Feingold and Mazzella

    (1993) reported that the overall correlations between humor production versus

    humor cognition and verbal ability, respectively, were .38 and .41, while hu-

    mor cognition and verbal ability were themselves positively correlated, r= .55,

    allp < .001. Future work in this mode could further inform the basis of indi-

    vidual differences in humor abilities by partial correlational analyses that

    would control for individual differences in factors like verbal or general intel-

    ligence or insight problem solving ability (see Schooler and Melcher 1995).

    Another potentially fruitful direction, not pursued in this report, is to de-

    velop links between the accurate measurement of humor appreciation, compre-

    hension, and production with evolutionary theory, specically Darwinian sex-

    ual selection. If humor comprehension and production are stable, ability-based,

    and possibly heritable individual difference variables, then they may be related

    to other cognitive skills, intelligence, and creativity, and thus reect genetic

    tness (Kaufman et al. 2008, Miller 2000). From a sexual selection viewpoint,

    there should also be selection pressure for males to be better producers of hu-

    mor and women to be better appreciators of humor, a hypothesis supported inrecent work by Bressler et al. (2006).

  • 7/28/2019 53059852

    24/28

    398 A. Kozbelt and K. Nishioka

    Finally, future work might examine the extent to which the ability to detect

    and understand LC functions broadly across domains, that is, not just in jokes

    or humor. These domains might include art, poetry, literature, advertising, in-

    terpreting narratives in psychotherapy sessions, or other twilight phenomena(Erdelyi 1985: 145209). In any case, the present investigation provides a

    promising basic methodology and set of statistical procedures that can serve as

    the basis for inquiry into the nature of humor production, its relation to humor

    comprehension and appreciation, and some of their underlying cognitive

    mechanisms.

    Brooklyn College of the City University of New York

    Notes

    Correspondence addresses: [email protected]; [email protected]

    * Special thanks to Matthew Hugh Erdelyi and Jacquelyn Bergstein, who provided valuable ad-

    vice about latent content. James Liu wrote the computer programs used for stimulus display

    and data collection, and Felix George greatly helped with data entry and organization. Paul

    Bruening and Michael Magee helped pilot test some of the materials. Two anonymous referees

    provided very helpful critiques.

    References

    Amabile, Teresa M. 1983. The social psychology of creativity. New York: Springer.

    Attardo, Salvatore. 1994.Linguistic theories of humor. Berlin & New York: Mouton de Gruyter.

    Babad, Elisha Y. 1974. A multi-method approach to the assessment of humor: A critical look at

    humor tests.Journal of Personality 42. 618631.

    Bergstein, Jacquelyn. 2005. Recognition hypermnesia. New York: City University of New York

    dissertation.

    Bressler, Eric R., Rod A. Martin & Sigal Balshine. 2006. Production and appreciation of humor as

    sexually selected traits.Evolution and Human Behavior27. 121130.

    Brodzinsky, David M. & Janet Rubien. 1976. Humor production as a function of sex of sub-

    ject, creativity, and cartoon content. Journal of Consulting and Clinical Psychology 44. 597

    600.

    Byrne, Donn. 1956. The relationship between humor and the expression of hostility. Journal of

    Abnormal and Social Psychology 53.8489.

    Cohen, Jacob. 1988. Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum.

    Couturier, Lance C., Richard S. Manseld & Jeanette M. Gallagher. 1981. Relationships between

    humor, formal operational ability, and creativity in eighth graders. The Journal of Genetic Psy-

    chology 139. 221226.

  • 7/28/2019 53059852

    25/28

    Humor comprehension and production 399

    Cunningham, William A. & Peter Derks. 2005. Humor appreciation and latency of comprehension.

    Humor: International Journal of Humor Research 18. 389403.

    Csikszentmihalyi, Mihalyi & Keith Sawyer. 1995. Creative insight: The social dimension of a

    solitary moment. In Robert J. Sternberg & Janet E. Davidson (eds.), The nature of insight,

    329363. Cambridge, MA: MIT Press.Derks, Peter. 1987. Humor production: An examination of three models of creativity. The Journal

    of Creative Behavior21.325326.

    Derks, Peter & Dedreck Hervas. 1988. Creativity in humor production: Quantity and quality in

    divergent thinking.Bulletin of the Psychonomic Society 26. 3739.

    Dunlap, William P., Jose M. Cortina, Joel B. Vaslow & Michael J. Burke. 1996. Meta-analysis of

    experiments with matched groups or repeated measures designs. Psychological Methods 1.

    170177.

    Erdelyi, Matthew H. 1985. Psychoanalysis: Freuds cognitive psychology. New York: W. H.

    Freeman and Company.

    Erdelyi, Matthew H. 1999. The unconscious, art, and psychoanalysis. Psychoanalysis and Con-

    temporary Thought22. 609626.Erdelyi, Matthew H. 2006.The unied theory of repression.Behavioral and Brain Sciences 29.

    499511.

    Erdelyi, Matthew H. & Judy B. Stein. 1981. Recognition hypermnesia: The growth of recognition

    memory (d') over time with repeated testing. Cognition 9. 2333.

    Eysenck, Hans. 1942. Appreciation of humor: An experimental and theoretical study.British Jour-

    nal of Psychology 32. 295309.

    Fabrizi, Michael S. & Howard R. Pollio. 1987. Are funny teenagers creative? Psychological Re-

    ports 61.751761.

    Feingold, Alan. 1983. Measuring humor perception: Revision and construct validation of the Humor

    Perceptiveness Test.Perceptual and Motor Skills 56. 159166.

    Feingold, Alan & Ronald Mazzella. 1993. Preliminary validation of a multidimensional model ofwittiness.Journal of Personality 61. 439456.

    Fotki.com (n.d.). http://www.fotki.com (accessed 2 March 2010).

    Found Slides Foundation (n.d.). http://www.lostandfrowned.com/foundsl2.html (accessed 2

    March 2010).

    Freud, Sigmund. 1953 [1900]. The interpretation of dreams. In James Strachey (ed. and trans.),

    Thestandard edition of the complete psychological works of Sigmund Freud, Vols. 4 and 5,

    1630. London: Hogarth Press.

    Freud, Sigmund. 1960 [1905].Jokes and their relation to the unconscious. In James Strachey (ed.

    and trans.), Thestandard edition of the complete psychological works of Sigmund Freud, Vol. 8,

    1258. London: Hogarth Press.

    Gick, Mary L. & Robert S. Lockhart. 1995. Cognitive and affective components of insight. InRobert J. Sternberg & Janet E. Davidson (eds.), The nature of insight, 197228. Cambridge,

    MA: MIT Press.

    Goldstein, Jeffrey H. 1970. Humor appreciation and time to respond. Psychological Reports 27.

    445446.

    Grice, H. Paul. 1975. Logic and conversation.In Peter Cole & Jerry L. Morgan (eds.),Syntax and

    semantics, Volume 3: Speech acts, 4158. New York: Academic Press.

    Humke, Christiane & Charles E. Schaefer. 1996. Sense of humor and creativity. Perceptual and

    Motor Skills 8.544546.

    Kaufman, Scott B., Aaron Kozbelt, Melanie L. Bromley & Geoffrey L Miller. 2008. The role

    of creativity and humor ability in human mating. In Glenn Geher & Geoffrey L. Miller (eds.),

    Mating intelligence: Theoretical and empirical insights into intimate relationships, 227262.Mahwah, NJ: Erlbaum.

  • 7/28/2019 53059852

    26/28

    400 A. Kozbelt and K. Nishioka

    Koestler, Arthur. 1964. The act of creation. New York: Dell.

    Khler, Gabriele & Willibald Ruch. 1996. Sources of variance in current sense of humor invento-

    ries: How much substance, how much method variance? Humor: International Journal of

    Humor Research 9.363398.

    Koppel, Mark A. & Lee Sechrest. 1970. A multitrait-multimethod matrix analysis of sense andhumor.Educational and Psychological Measurement30. 7785.

    Linacre, John M. & Benjamin D. Wright. 2000 [1991]. A users guide to WINSTEPS. Chicago:

    MESA Press.

    Lockhart, Robert S.,Mary Lamon&Mary L. Gick. 1988. Conceptual transfer in simple insight

    problems.Memory & Cognition 16. 3644.

    Macmillan, Neil & C. Douglas Creelman. 1991. Detection theory: A users guide. Cambridge:

    Cambridge University Press.

    Maier, Norman R. F. 1932. A Gestalt theory of humor.British Journal of Psychology 23.6974.

    Martin, Rod A. 1996. Situational Response Questionnaire (SHRQ) and Coping Scale (CHS): A

    decade of research ndings.Humor: International Journal of Humor Research 9. 251272.

    Masten, Ann S. 1986. Humor and competence in school-aged children. Child Development57.461473.

    Metcalfe, Janet & David Wiebe. 1987. Intuition in insight and non-insight problem solving.Mem-

    ory & Cognition 15. 238246.

    Miller, Geoffrey L. 2000. The mating mind. New York: Random House.

    Murdock, Mary C. & Rita M. Ganim. 1993. Creativity and humor: Integration and incongruity.

    Journal of Creative Behavior27. 5770.

    OQuin, Karen & Peter Derks. 1997. Humor and creativity: A review of the empirical literature. In

    Mark A. Runco (ed.), The creativity research handbook, vol. 1, 227256. Cresskill, NJ: Hamp-

    ton Press.

    Perkins, David N. 1981. The minds best work. Cambridge, MA: Harvard University Press.

    Rasch, George. 1980 [1960].Probabilistic models for some intelligence and attainment tests. Chi-cago: University of Chicago Press.

    Roeckelein, Jon E. 2002. The psychology of humor: A reference guide and annotated bibliography.

    Westport, CT: Greenwood Press.

    Schooler, Jonathan W., Marte Fallshore & Stephen M. Fiore. 1995. Epilogue: Putting insight into

    perspective. In Robert J. Sternberg & Janet E. Davidson (eds.), The nature of insight, 559587.

    Cambridge, MA: MIT Press.

    Schooler, Jonathan W. & Joseph Melcher. 1995. The ineffability of insight. In Stephen M. Smith,

    Thomas B. Ward & Ronald A. Finke (eds.), The creative cognition approach, 97133. Cam-

    bridge, MA: MIT Press.

    Sheehy-Skefngton, Alan. 1977. The measurement of humor appreciation. In Antony J. Chapman

    & Hugh C. Foot (eds.),Its a funny thing, humour, 447449. Oxford: Pergamon Press.Siegler, John N. 2004. Funny you asked: Re-representation and the cognitive processes involved

    in creating humor. Dissertation Abstracts International Section A: Humanities & Social Sci-

    ences 64(11-A). 3965.

    Sternberg, Robert J. (ed.). 1999.Handbook of creativity. New York: Cambridge University Press.

    Sternberg, Robert J. & Janet E. Davidson (eds.). 1995. The nature of insight. Cambridge, MA: MIT

    Press.

    Treadwell, Yvonne. 1970. Humor and creativity.Psychological Reports 26. 5558.

    Turner, Robert G. 1980. Self-monitoring and humor production.Journal of Personality 48. 163

    172.

    Vaid, Jyotsna, Rachel Hull, Roberto Heredia, David Gerkens & Francisco Martinez. 2003. Getting

    a joke: The time course of meaning activation in verbal humor. Journal of Pragmatics 35.14311499.

  • 7/28/2019 53059852

    27/28

    Humor comprehension and production 401

    Weisberg, Robert W. & Joseph W. Alba. 1981. An examination of the alleged role of xation in

    the solution of several insight problems.Journal of Experimental Psychology: General110.

    169192.

    Wierzbicki, Michael & Richard D. Young. 1978. The relation of intelligence and task difculty to

    appreciation of humor. The Journal of General Psychology 99. 2532.Wright, Benjamin D. & Geofferey N. Masters. 1982.Rating scale analysis: Rasch measurement.

    Chicago: MESA Press.

    Ziv, Avner. 1976. Facilitating effects of humor on creativity.Journal of Educational Psychology

    68. 318322.

    Ziv, Avner. 1988. Using humor to develop creative thinking.Journal of Children in Contemporary

    Society 20. 99116.

  • 7/28/2019 53059852

    28/28