INTRODUCTION
During a period of life when health may be more vulnerable and social networks more limited, the physical and psychosocial benefits of humor may buffer the negative changes and enhance the positive changes that occur in the lives of the elderly. Of course, for people to benefit from humor, they must first understand it. It is important to note that comprehension precedes appreciation. Although these processes are often studied in tandem, the cognitive underpinnings of comprehension are not understood well enough to make strong claims regarding its interaction with subsequent processes. Studying humor comprehension in the context of aging, a time when significant cognitive changes generally occur, may clarify the role of cognition in humor comprehension.
Much of humor research is rooted in the philosophical notion that humor arises from a sense of incongruity, a conflict between the expected and the actual (Kant, 1790/1987; Schopenhauer, 1819/1906). More recently, psychologists have proposed the incongruity–resolution theory, a two-stage theory of humor comprehension where the presentation of a humorous stimulus initiates a state of incongruity (i.e., the surprise factor) followed by a reassessment of the joke's punch line by exploring alternate resolutions (Morreall, 1987; Shultz, 1972; Suls, 1972). Both stages are essential for humor comprehension: surprise separates the extraordinary from the ordinary, whereas the selection of the correct resolution (i.e., a resolution that is both surprising and coherent) allows a person to “get” a joke. Successful comprehension of humor occurs upon resolving something that is seemingly incongruous with a logical but less obvious explanation (Suls, 1972).
The Joke and Story Completion Test (Brownell et al., 1983), widely used in humor research, is a measure based on the incongruity–resolution theory of humor. In the first portion of the test, joke stems are presented with four punch line response options. The correct punch line, a logical and humorous answer, involves both the recognition of incongruity and its resolution. The other answer choices fall into one of three categories: straightforward, a logical, nonhumorous answer; humorous nonsequitur, an illogical, humorous answer; or unrelated nonsequitur, an illogical, nonhumorous answer. Selecting any of these three answers indicates the failure to recognize either the incongruity or resolution (or both) of the joke.
Recognizing incongruity and identifying the appropriate resolution in humor are complex mental tasks that rely on the integrity of brain structure and function. Brownell et al. (1983) showed that people with right hemisphere damage made more errors on the Joke and Story Completion Test than normal controls. Other studies found that right hemisphere patients generally had greater deficits in humor comprehension than normal controls and left hemisphere patients (Bihrle et al., 1986; Gardner et al., 1975; Wapner et al., 1981). More recently, studies also suggest that the frontal lobes are important for humor processing (Goel & Dolan, 2001; Shammi & Stuss, 1999, 2003; Uekermann et al., 2006). Comparing activation in response to funny jokes and unfunny jokes, Goel and Dolan (2001) concluded that the medial ventral prefrontal cortex plays an important role during humor processing. Shammi and Stuss (1999) found that people with right frontal damage had deficits in humor comprehension (as measured by the Joke and Story Completion Test), along with deficits in working memory, cognitive flexibility, abstract reasoning, and visual search and scan. People with right frontal damage frequently selected humorous nonsequitur endings, indicating a failure to integrate the surprise and coherence aspects of a joke. Moreover, significant correlations between tests of humor comprehension and tests of cognitive abilities suggest that humor comprehension relies upon these cognitive abilities. Shammi and Stuss concluded that the right frontal lobe may be responsible for many of the cognitive functions used in humor comprehension. Uekermann et al. (2006) added another dimension, theory of mind, to the Shammi and Stuss (2003) paradigm and found similar results. Uekermann et al. suggest that humor deficits may in part be due to deficits in theory of mind, which is typically considered a frontal ability. In sum, these studies would suggest that frontal degradation may result in reduced cognitive abilities and thus reduced humor comprehension.
The frontal aging hypothesis asserts that the frontal lobes are particularly susceptible to the deleterious effects of normal aging and that frontally mediated abilities generally decline with age (West, 1996). Cognitive flexibility, abstract reasoning, and working memory, all frontally mediated abilities (e.g., Eslinger & Grattan, 1993; Parkin & Walter, 1991; Shammi & Stuss, 1999), tend to deteriorate with normal aging (Albert & Moss, 1988; Hess & Slaughter, 1986; Parkin & Lawrence, 1994; Schultz et al., 1980; Shammi & Stuss, 2003). If the cognitive abilities involved in humor comprehension are housed in the frontal regions and thus decline with age, older adults should exhibit greater difficulty understanding humor when compared with younger adults. Research by Shammi and Stuss (2003) support this claim. They did not find age-related differences in humor appreciation but found that older adults were also less able to select the correct punch lines in verbal and nonverbal humor comprehension tests. Correlations between humor comprehension and cognitive tests were fairly strong (r = .44–.59), suggesting that poor cognitive performance may in part account for the decline in humor comprehension among older adults.
Evidence supports the notion that humor comprehension is related to at least three aspects of cognition. Using the Joke and Story Completion Test, three studies (Ozonoff & Miller, 1996; Shammi & Stuss, 1999, 2003) have found that cognitive flexibility was associated with a tendency to select answers that require access to more than one interpretation of the humorous material. During the process of punch line selection, cognitive flexibility likely works in tandem with abstract reasoning, the ability to draw connections between seemingly unrelated material. Being able to access alternate explanations only seems useful when a person can abstract the appropriate connection between the beginning of a joke and its punch line (Shammi & Stuss, 1999, 2003). Finally, both of these cognitive processes rely on the ability to store multiple explanations simultaneously. Without short-term memory (STM), proper evaluation and selection of the correct interpretation is difficult (Attardo, 1997; Nerlich & Clarke, 2001).
The goal of the current study was to replicate and extend the work of Shammi and Stuss (2003). One methodological change was to include a nonverbal humor test that is structurally analogous to the verbal humor test to facilitate comparison of results across modality. Because language comprehension is generally retained in older age (Burke et al., 2000), merely using a verbal measure may produce an inaccurate understanding of humor comprehension in old age. Furthermore, replicating previous results using multiple modalities would strengthen the original findings. This study also incorporated measures of verbal and nonverbal STM instead of working memory as in previous research, because there is no manipulation of the temporarily stored material during humor comprehension. Lastly, this study controlled for participants' general intelligence, a factor that previously has been associated with humor abilities (Holt & Willard-Holt, 1995; Miller, 2001).
Based on the evidence suggesting that humor comprehension relies on specific cognitive abilities, all of which are purported to decline with age, older adults are expected to have greater difficulty with tests of cognitive flexibility, abstract reasoning, and STM, and thus, tests of humor comprehension.
METHODS
Participants
Forty-one adults over age 65 (mean age, 78.3; range, = 65–93 years; SD = 6.9) were recruited from the older adult subject pool from the Psychology Department at Washington University in St. Louis. Participants were free of overt cognitive deficits, as assessed by the Short Blessed Test (Katzman et al., 1983) (mean number or errors = 1.5, SD = 2.1). Forty undergraduates (mean age, 19.8; range, 18–22 years; SD = 1.1) were recruited from introductory psychology courses. The groups differed in performance on the vocabulary test from the Wechsler Adult Intelligence Scale [t(79) = 3.04; p < .01], with older adults scoring lower (M = 14.4; SD = 2.6) than younger adults (M = 15.9; SD = 2.1). The older group was 95% Caucasian, whereas the younger group was 70% Caucasian. The groups did not differ on gender composition (28 women in each group) or average number of years of education [old: M = 14.43, SD = 3.01; young: M = 13.70, SD = 1.23; t(79) = −1.43], not significant. Older adults received monetary remuneration, and younger adults received class credit for participation. This study received ethical approval from the Institutional Review Board at Washington University.
Measures
Humor measures.
The Joke and Story Completion Test (Brownell et al., 1983) was used to measure comprehension of verbal humor based on the incongruity–resolution model (Suls, 1972). The joke portion of the test included 16 joke stems, each with four endings categorized as the correct funny ending (FC), straightforward ending (SF), humorous nonsequitur ending (HNS), or an unrelated nonsequitur ending (UNS). Participants were instructed to select the ending that best functioned as the “punch line” of the joke.
Example: A business man is riding the subway after a hard day at the office. A young man sits down next to him and says, “Call me a doctor … call me a doctor.” The businessman asks, “What's the matter, are you sick?”
- The young man says, “I just graduated from medical school.”—FC.
- The young man says, “Yes I feel a little weak. Please help me.”—SF.
- The young man says, “My sister is a nurse.”—UNS.
- The young man pulls out a water gun and squirts the businessman.—HNS.
The story portion of the test measured a person's ability to understand nonhumorous narrative texts. It contained the same stems from the joke completion portion, but participants were asked to complete the story stem with a logical ending instead of a funny punch line. Before the administration of each test section, participants were given an example to ensure their understanding of the task. Scores for both portions of the test may range from 0 to 16. Cronbach's α based on the current sample was .74 for the joke portion and .98 for the story portion.
The test of nonverbal humor included 10 captionless four-cell comic strips taken from the “Ferd'nand” series (Rehr, 2005) (see Figure 1 for an example). A paid artist created three additional alternate endings to each of the cartoon strips that mirrored the categorization of the three alternate endings from the Joke Completion Test (i.e., straightforward, humorous nonsequitur, unrelated nonsequitur). Instructions were similar to the verbal humor test. This test did not include an initial example. Scores from this test may range from 0 to 10. Cronbach's α based on the current sample was .68.
Sample “Ferd'nand” cartoon strip (FERD'NAND © PIB Copenhagen/Dist. by United Feature Syndicate, Inc.) with the original and alternate endings below.
Cognitive tests.
In the Alternate Uses Test (Guilford et al., 1960), which measured cognitive flexibility, participants listed up to six different ways (other than its intended function) to use three everyday objects (e.g., button, key, shoe). The Similarities subtest from the Wechsler Adult Intelligence Scale-III (WAIS-III) (Wechsler, 1997) was used to measure abstract reasoning. Participants were presented with 18 two-word pairs and asked how the words are alike. A letter span test, which measured verbal STM, consisted of 18 trials of phonologically dissimilar letter strings that ranged from 2 to 10 letters. There were two trials of each length, and the order of presentation was random to counteract proactive interference. At the end of each trial, participants were asked to recite as many letters as they remembered in the order that they were presented. A grid span test (Hale et al., 1996), which measured nonverbal STM, consisted of 16 randomly ordered trials. Each trial contained two to nine X's, two trials of each length. During each trial, multiple X's appeared, one at a time, within a 4 × 5 grid. At the end of each trial, participants were asked to recall the location of the X's while considering the order of presentation. Both span tasks included four practice trials before testing began. Finally, the vocabulary test from the WAIS-III (Wechsler, 1997) was administered as a measure of general IQ.
Procedure
Upon arrival, participants reviewed and signed a consent form. Subsequently, the experimenter administered the cognitive tests and humor tests in a counterbalanced order. The Joke and Story Completion Test was administered as a paper-and-pencil task, whereas the Cartoon Completion Test was presented in the form of cards to select from. The Vocabulary test and Similarities were administered orally, the span tasks were administered via computer, and the Alternate Uses Test was administered as a paper-and-pencil task. Participants were debriefed (and older participants were paid) following completion of testing.
Data Analysis
Separate 2 × 4 (age group × type of response) mixed design analyses of variance (ANOVAs) were computed for tests of humor and cognition to examine within-group and between-group differences. Bivariate correlations were conducted separately for each age group to examine the relationships among the humor and cognitive measures. Lastly, multiple regression analyses were conducted to examine whether cognitive flexibility, abstract reasoning, and STM explain unique variance in humor test performance after controlling for IQ. All statistical tests were based on an α level of .05.
RESULTS
There were no significant gender effects, so men and women are combined in the analyses that follow. Additionally, because the sample was predominantly Caucasian, effects of ethnicity could not be examined.
Humor Tests
On average, older adults made more errors on the Joke Completion Test than younger adults [t(78) = −2.52; p < .05; ηp2 = .08; see Table 1]. Analyses of the types of errors revealed that older adults were more likely to choose straightforward responses than younger adults [t(78) = −2.43; p < .05; ηp2 = .08]. Rates of selecting humorous nonsequitur endings and unrelated nonsequitur endings did not differ between groups [t(45) = −1.37, not significant; t(79) = −1.92, not significant], respectively. Among the types of errors, older adults were more likely than younger adults to select straightforward responses over humorous and unrelated nonsequiturs [F(1,78) = 6.91; p < .01; ηp2 = .08]. Older adults also were less able than younger adults to select the appropriate logical ending on the Story Completion Test [t(44) = −2.52; p < .01; ηp2 = .08; see Table 1].
Performance on humor tests, by age
Older adults also made more errors than younger adults on the Cartoon Test [t(79) = −3.09; p < .01; ηp2 = .11]. Older adults were more likely to choose straightforward responses than younger adults [t(79) = −3.09; p < .05; ηp2 = .11], but equally likely to select humorous nonsequitur and unrelated nonsequitur endings [t(79) = −1.70, not significant; t(79) = −.58, not significant], respectively. Among the types of errors, older adults were more likely than younger adults to select straightforward responses over humorous and unrelated nonsequiturs [F(1,79) = 5.10; p < .05; ηp2 = .06].
Cognitive Tests
Older adults exhibited greater difficulty with cognitive flexibility [t(79) = 3.63; p = .001; ηp2 = .14], abstract reasoning [t(79) = 4.32; p < .001; ηp2 = .19], and both measures of STM [verbal: t(79) = 6.35, p < .001, ηp2 = .34; nonverbal: t(79) = 7.36, p < .001, ηp2 = .41; see Table 2 for means]. Younger adults scored higher than older adults on the WAIS Vocabulary Test [t(79) = 3.17; p < .01; ηp2 = .11].
Performance on cognitive tests, by age
Relationships Among Humor Tests and Cognitive Tests
For older participants, the number of errors on the Joke Completion Test was significantly associated with scores from Alternate Uses (r = −.46; p < .001), Similarities (r = −.50; p <.01), but not letter span (r = −.13; not significant). The number of errors on the Cartoon Test was associated with scores from Alternate Uses (r = −.53; p < .001), Similarities (r = −.60; p < .001), and grid span (r = −.44; p < .01). There were also significant correlations between Vocabulary and Similarities (r = .65; p < .001), Vocabulary and Alternate Uses (r = .38; p < .01), Alternate Uses and Similarities (r = .58; p < .001), and grid span and Similarities (r = .36; p < .01). For younger participants, the number of errors on the Joke Completion Test and Cartoon Test were not significantly associated with scores from any of the cognitive tests.
Multiple regression analyses were conducted only on the data from the older adult sample, because correlations between cognitive tests and humor tests were nonsignificant in the younger adult sample. Only independent variables that were significantly correlated with the dependent variable were entered into the regression analysis (see Table 3). Performance on Similarities and Alternate Uses accounted for a significant amount of variance in performance on the Joke Completion Test after controlling for IQ [ΔR2 = .12; F(3,34) = 3.35; p < .05]. Neither of the tests on its own had a significant relationship with scores on the Joke Completion Test. Similarly, performance on Similarities, Alternate Uses, and grid span accounted for a significant amount of variance in the Cartoon Completion Test after controlling for IQ [ΔR2 = .26; F(1,35) = 5.95; p < .01], but none of the tests on their own had a significant impact on the Cartoon Completion Test. In other words, after controlling for intelligence, adding cognitive test performance to the model increased the amount of explained variance in humor test performance, yet no test played a unique explanatory role in humor comprehension.
Hierarchical regression analysis predicting humor comprehension in older adults
DISCUSSION
Results from this study replicate and extend previous research findings of age differences in humor comprehension. The Joke Completion Test (Brownell et al., 1983) and the analogous Cartoon Completion Test, created for the current study, produced similar patterns of results, which indicate that age differences exist in both verbal and nonverbal humor comprehension. Specifically, older adults performed worse on humor tests and cognitive tests, suggesting that humor is related to cognition.
Overall, the results of the present investigation are similar to the findings of Shammi and Stuss (1999, 2003). Almost all the correlations between cognitive and humor comprehension measures within the present study are similar in magnitude to those obtained by Shammi and Stuss. Additionally, we found that STM may be as robust of a predictor for nonverbal humor comprehension as working memory. On the other hand, null correlations between verbal humor comprehension and verbal STM suggest that verbal humor comprehension may require a different type of STM from that which is measured by letter span, a task involving the phonological loop. Recent studies have suggested that semantic STM is dissociated from phonological STM (Hanten & Martin, 2000; Martin & Freedman, 2001) and more predictive of verbal problem solving (Haarmann et al., 2003). Similarities between humor comprehension and problem solving processes (Holt & Willard-Holt, 1995; Suls, 1972) suggest that factors related to problem solving may also be related to humor comprehension. Semantic span may be more relevant for problem-solving than phonological span, because this type of memory allows for a more meaningful organization of the memorized material, facilitating the selection of the appropriate memories for correct joke interpretation. Future studies of humor comprehension and cognition could investigate this idea.
If letter span performance was not related to verbal humor, why was grid span performance related to nonverbal humor? It is plausible that processing nonverbal humor may rely more on STM. Processing each frame of a cartoon required the translation of visual elements to an imaginal and a verbal code, whereas the processing of a verbal joke only necessitated a verbal code. Older adults have greater difficulty with picture processing than language processing because of this additional step (Duchek et al., 1995). To “get” the punch line at the end of the cartoon, people had to remember what visual and verbal elements contributed to the meaning of the cartoon. A benefit of this increased difficulty in picture processing is that a nonverbal humor comprehension measure is likely more sensitive to age differences and perhaps offers a more precise depiction of humor comprehension ability in old age. For example, in the present study, based on the joke data alone, we might conclude that age-related deficits in humor comprehension are mild, but the greater between-group difference in the cartoon data suggest that humor comprehension ability may vary depending on presentation modality.
As a whole, cognition was a significant predictor of humor comprehension in the present study, but no cognitive ability alone had unique predictive power. Significant correlations between cognitive tests suggest that multicollinearity could be one explanation, but collinearity diagnostics in our data were within acceptable range (Cohen et al., 2003). Another possibility is that variable omission led to biased parameter estimates (Cohen et al., 2003). However, the current lack of a formal model of humor comprehension leaves room for more exploratory analyses to clarify the variables that are important in understanding the relationship between humor comprehension and cognition. For example, additional regressions were performed without the predictor with the highest variance inflation factor, Similarities, to compare with the original regression results (refer to Table 4). There is a slight reduction in the proportion of explained variance at step 2, but the removal of this predictor created a better model fit with significant predictors at each level. Explanations may be that (1) these particular humor measures rely more on cognitive flexibility than abstract reasoning; or (2) Similarities and Alternate Uses may not be pure estimates of abstract reasoning and cognitive flexibility, respectively; or (3) some combination of both. Obviously these explanations are speculative, and more research is needed to clarify the relative contribution of specific cognitive processes to humor comprehension and whether these relationships are modulated by the type of humor. Additionally, better measurement of humor comprehension and cognition is needed to characterize their relationship (e.g., use multiple forms of humor tests, create an index for each cognitive ability based upon more than one test).
Hierarchical regression analysis predicting humor comprehension in older adults without similarities
The overall results from this study are consistent with theories of cognitive aging. For example, the present finding that older adults had poorer cognitive and humor comprehension abilities is consistent with the frontal aging hypothesis, which states that the frontal lobes and their respective functions are most vulnerable to aging (Raz, 2000; West, 1996). The findings from this study are also consistent with the inhibitory deficit hypothesis (Hasher & Zacks, 1988), which states that with increasing age comes the tendency to allow irrelevant information into working memory. On both the verbal and nonverbal humor tests, older adults, compared with younger adults, selected more straightforward answers. Even if the correct funny answer registered as a potential answer choice, the straightforward answer choice may have been more salient to older adults due to the answer's semantic relevance to the joke stem based on a literal, or more common, interpretation. This heightened activation of the straightforward answer may have caused interference during answer selection, subsequently resulting in subthreshold activation for the correct answer. Previous studies using the Stroop Test offer confirmatory evidence for age-related inhibitory deficits (e.g., West & Alain, 2000).
Although the present results are positive, they are contingent upon some limitations. Verbal and nonverbal humor tests exhibited adequate reliability in the present sample, but formal study of the instruments' psychometric properties is needed. The lack of normative data is also important, because it is not clear whether the age group differences are attributable to actual skill differences or cohort differences. Third, there are no validity data for these humor tests. Findings that contradict previous studies may be due to sampling or methodological problems. For example, older adults may have performed worse on the Story Completion Test simply due to fatigue. Alternatively, older adults generally have greater difficulty with task shifting (e.g., Axelrod & Henry, 1992; Wecker et al., 2005); thus, it might have been challenging for them to carry out different instructions on the same set of items they just completed with other instructions (the story portion of the test always follows the joke portion). Without these data, interpreting the differences between the current study and Shammi & Stuss (2003) is difficult. Nevertheless, many other studies have used the Joke and Story Completion Test and come to the same conclusion that humor comprehension is impaired in people with cognitive deficits (Bihrle et al., 1986; Brown et al., 2005; Ozonoff & Miller, 1996; Shammi & Stuss, 1999, 2003). Humor tests that are similar to the Joke and Story Completion Test also produce the same results (Uekermann et al., 2006).
Despite its limitations, this study offers additional evidence that there are age differences in humor comprehension, a phenomenon that may be associated with the age-related decline of cognitive abilities. Further study of humor in this population is warranted, because humor may have significant physical and psychological benefits for older adults.
ACKNOWLEDGMENTS
This study was supported by National Institute on Aging training grant no. AG-00030. Thanks to Hiram Brownell for permission to use the Joke and Story Test; Martha Storandt, Sandra Hale, and Joel Myerson for general advice; Janet Duchek for comments on an early draft; and Ashley Johnson and Kendall Smith for assistance with data collection.
