IDENTIFYING VARIABLES - STATISTICS
jaybird
schmidt_article.pdf
Journal of Experimental Psychology: Learning, Memory, and Cognition 1994, Vol. 20, No. 4,953-967
Copyright 1994 by the American Psychological Association, Inc. 0278-7393/94/S3.00
Effects of Humor on Sentence Memory
Stephen R. Schmidt
Memory for humorous and nonhumorous versions of sentences was compared. Humorous sentences were better remembered than the nonhumorous sentences on both free- and cued-recall tests and on measures of sentence recall and word recall. These effects persisted when subjects were warned that they were about to read a humorous sentence but were attenuated in incidental learning and limited to within-subjects manipulations. In incidental learning, recall was also scored as a function of subjective ratings of humor. Subjective humor affected memory in both within- and between-subjects designs. Attention, arousal, rehearsal, retrieval, and surprise explanations were explored. Results suggest that humorous material receives both increased attention and rehearsal relative to nonhumorous material.
The emotion evoking qualities of a stimulus have a profound impact on memory processes. This impact has received in- creased attention, leading to the publication of numerous journal articles and to the recent publication of The Handbook of Emotion and Memory: Research and Theory (Christianson, 1992). However, most researchers have focused on the impact of negatively valenced emotions such as trauma and anxiety. The research presented in this article focuses on stimuli leading to the positive emotions associated with humor. Humor plays an important part in education, advertisement, entertainment, and politics. Nonetheless, basic research con- cerning the effects of humor on memory and attention is in short supply (see McGhee & Goldstein, 1983, for a review of research on humor). In fact, there was no coverage of the topic in Christianson's volume. The research presented here at- tempts to answer some of the fundamental questions concern- ing the effects of humor on memory. I begin with a review of previous research, followed by a discussion of potential expla- nations for positive effects of humor on memory. Six experi- ments are then presented in which memory for humorous and nonhumorous versions of sentences was compared, and the results of these experiments are discussed in terms of the potential explanations.
Most research investigating the effects of humor on learning and memory has been conducted in educational settings with less than rigorous experimental control. For example, Kaplan and Pascoe (1977) compared memory for lecture material presented with either serious or humorous examples. Kintsch and Bates (1977) compared memory for topic statements, details, and extraneous remarks and jokes presented during a lecture. Zillmann, Williams, Bryant, Boynton, and Wolf (1980)
This research was supported by a Faculty Research Grant from Middle Tennessee State University. I thank the following people for assistance in selecting materials, testing students, and scoring recall protocols: Dianne Bowden, Charles Durham, Marian Granger, Rich- ard Pearl, Tony Sheppard, Sandy Starrett, Denise Waller, and Alan Williams. I extend special thanks to Constance R. Schmidt and Mark McDaniel for their comments on an earlier version of this article.
Correspondence concerning this article should be addressed to Stephen R. Schmidt, Department of Psychology, Middle Tennessee State University, Murfreesboro, Tennessee 37132. Electronic mail may be sent to [email protected].
looked at the effects of humor contained in an educational television program. In addition, several investigators have looked at the effects of humor in advertising (Duncan, Nelson, & Frontzak, 1984; Gelb & Zinkhan, 1986). Most of these researchers have reported positive effects of humor on memory; however, the nature of these effects is far from clear. Compari- sons between humorous and nonhumorous materials matched in content have not been made. This is important because humorous material often concerns topics that are particularly interesting, meaningful, or relevant to the subjects. For ex- ample, many jokes contain references to sex, dating, politics, and other meaningful topics. In an appropriate test of the effects of humor, the content of the material must be con- trolled, ruling out explanations that are based on interest. In addition, careful control of the experimental context has been lacking in previous experimentation. Furthermore, some stud- ies reported negative effects (Kaplan & Pascoe, 1977), and some studies reported positive effects (Zillmann et al., 1980) of humorous material on memory for surrounding material. Clearly, carefully controlled experimentation is needed before firm conclusions on the effects of humor on memory can be reached.
Numerous explanations for the effects of humor on memory can be offered. First consider the hypothesis that the effects of humor are a result of physiological arousal. McGhee (1983) noted increased heart rate and respiration rate in response to humor and argued that these responses were indicative of activation of the sympathetic nervous system. Numerous stud- ies have demonstrated the positive effects of arousal on memory (see Craik & Blankstein, 1975, for a review). For example, Ellis, Detterman, Runcie, McCarver, and Craig (1971) compared subjects' memory for pictures of nudes to subjects' memory for more typical magazine pictures. Recall and recognition of the nude pictures were greater than memory for the typical pictures. Similarly, high-arousal words, such as rape and vomit, are better remembered than low- arousal words (Maltzman, Kantor, & Langdon, 1966; Walker & Tarte, 1963). Research concerning "flashbulb memories" (Bohannon, 1988; Brown & Kulik, 1977; Christianson, 1989; Pillemer, 1984) has also indicated that emotional responses can enhance memory. In addition, enhanced memory for
953
954 STEPHEN R. SCHMIDT
traumatic stimuli has been reported by Christianson, Loftus, Hoffman, and Loftus (1991) and Heuer and Reisberg (1990).
Most of the studies reported above used arousal associated with negative affect. Isen (1985) argued that the effects of arousal on memory are asymmetrical for positive and negative moods. Thus, much of what we have learned about the effects of arousal on memory may not apply to humor. Nonetheless, one may postulate that the arousal associated with humor is at least in part responsible for enhanced memory for humorous materials. This hypothesis is evaluated below.
A second hypothesis is based on differences in the attention given to humorous and nonhumorous material. Deckers and Hricik (1984) argued that a humorous response was a special case of the physiological orienting response, and the orienting response is often used as an index of increased attention (Isen, 1985). An information processing model of the orienting response was developed by Ohman (1979), and its role in memory was further developed by Schmidt (1991) in his "incongruity hypothesis."
The incongruity hypothesis was developed as a refinement of the "distinctiveness hypothesis" (Eysenck, 1979; Hunt & Elliot, 1980). The concept of distinctiveness has been difficult to define, and a clear operational definition of distinctiveness has not been offered. In contrast, incongruity has been clearly defined. Incongruity results from a mismatch between features present in working memory and features activated by an item. The incongruous item leads to an orienting response and, as a result, receives extra attention during encoding and increased storage of individual-item information (Schmidt, 1991). These processes are thought to be automatic in the sense that they are not the result of an intentional strategy and do not deprive other cognitive processes of attentional resources.
Many theories of humor focus on incongruity as a source of humor (Deckers & Devine, 1981; Suls, 1972). For example, consider the sentence: "People are like tea bags . . . they don't know their own strength until they get into hot water." In most respects, people are not like tea bags, thus the opening phrase presents an incongruity. The humor of the sentence results from resolving this incongruity and understanding how people and tea bags are alike. In addition, humorous material pre- sented in the context of nonhumorous material may lead to incongruity. For example, we may attend equally to the jokes presented in a comedian's monologue, but an isolated joke presented during a lengthy commencement address is incongru- ent with the rest of the speech. As a result, a joke told during commencement may be retained better than the same joke told in a humorous monologue. Thus, within-item and between- item incongruity may contribute to increased attention to, and enhanced memory for, humorous materials.
A third explanation for the effects of humor on memory is that subjects rehearse humorous material more than nonhumor- ous material, and increased rehearsal is responsible for in- creased retention (e.g., Atkinson & Shiffrin, 1968). Unfortu- nately, the rehearsal hypothesis does not specify why subjects rehearse humorous material more than nonhumorous mate- rial. As a result, the circumstances under which humorous materials will receive more rehearsals than nonhumorous materials are left unspecified. Several versions of the rehearsal hypothesis were evaluated in the experiments reported in this
article. To clearly distinguish between rehearsal and increased attention as mechanisms, I make the assumption that rehearsal is a controlled process, requiring intention and cognitive resources. As a result, increased rehearsal of humorous material may occur at the expense of other material presented within a limited time frame.
The effects of humor on memory may result from "surprise" or retrieval strategies. According to the surprise hypothesis, subjects may not expect to read humorous materials as part of an experiment. According to the retrieval hypothesis, subject may have a bias toward retrieving humorous material or toward retrieving humorous material before nonhumorous material.
The mechanisms described earlier need not operate in isolation or to the exclusion of other mechanisms. Thus, for example, physiological arousal may lead to increased attention to an item. Increased attention may sometimes lead to in- creased rehearsal. Similarly, surprise may lead to increased attention. Clearly, separating the impact of these mechanisms on memory for humorous material will be difficult. The main purpose of the series of studies reported below was to establish the fundamental effects of humor on sentence memory. These experiments were not designed to provide critical tests be- tween the potential explanations of these effects. However, some of the data collected were inconsistent with several of these explanations.
Experiment 1: Pilot Study
The purpose of this experiment was to develop a set of materials with which to study the effects of humor on memory and to provide some preliminary findings concerning these effects. To equate the materials on as many variables as possible, I compared memory for humorous and nonhumorous versions of the same sentences. Humorous sayings (e.g., "The only way to keep your good health is to eat what you don't want, drink what you don't like, and do what you'd rather not" Mark Twain, cited in Adams, 1969, p. 164.) were collected by my research team. Each saying was then rewritten to retain the gist of the saying but remove the humor (e.g., "The only way to keep your good health is to eat good food, drink healthy drinks, and do healthy activities."). These sentences were then given to 45 undergraduates who were asked to rate the sentences on five different dimensions. From this group of sentences, I selected 20 sentences for which the humorous and nonhumor- ous versions differed greatly in their rated humor [F(l, 19) = 149.85] but did not differ on rated bizarreness, difficulty, meaningfulness, or familiarity (all Fs < 1.44). A complete listing of the selected sentences appears in the Appendix along with the sources of the sentences. Mean rated humor for the humorous versions of the selected sentences ranged from 3.08 to 4.00 (where 5.00 was the highest humor rating). Mean rated humor for nonhumorous versions ranged from 1.25 to 2.86.
Lists containing half humorous and half nonhumorous sentences were presented to other subjects (n = 38) in an intentional memory task. A greater proportion of humorous sentences (M = .35) was recalled than nonhumorous sen- tences (M = .27), F(l, 36) = 7.72, MSe = .1434; F2(l, 9) = 5.97; F'(l, 24) = 3.91. Similarly, a larger proportion of words
HUMOR AND MEMORY 955
was recalled from humorous sentences (M = .22) than from nonhumorous sentences (M = .16), F(l, 36) = 10.13, MSt = .0729; F2(l, 9) = 8.40; F'(l, 25) = 5.08. These results demonstrated that sentence humor is an effective mnemonic variable. When sentences were carefully controlled for sen- tence content and variables known to affect memory perfor- mance, humorous versions of the sentences were recalled better than nonhumorous versions.
Experiment 2
Experiment 2 was designed to replicate the results of Experiment 1 and to determine whether the effects reported in Experiment 1 generalized to a between-subjects design. I included experimental design as a factor by using a procedure described by Erlebacher (1977, 1978), and used by McDaniel and Einstein (1986). This procedure enables one to test directly for interactions between an independent variable and experimental design.
The interest in experimental design stemmed from the following considerations. Walker and Tarte (1963) demon- strated that high-arousal words were recalled better than low-arousal words in both within- and between-subjects ma- nipulations. Thus, a simple arousal explanation predicts benefi- cial effects of humor in both types of designs. In contrast, extra rehearsal of humorous material may occur at the expense of other material on a list. This "differential rehearsal" will only benefit recall of the humorous sentences in a within-list manipulations of sentence humor.
The two sources on incongruity discussed earlier (within- vs. between-sentence) are influenced differently by the manipula- tion of experimental design. Within-sentence incongruity should be unaffected by experimental design and should contribute to enhanced recall of humorous materials in both designs. How- ever, between-sentence incongruity may be limited to the within-subjects manipulation of sentence humor. That is, humorous material may receive increased attention because it is incongruous with a background of nonhumorous material. If the effects of humor are confined to within-subjects manipula- tions, this result would parallel the reported effects of bizarre- ness on sentence memory (McDaniel & Einstein, 1986).
Humorous sentences may benefit from some form of re- trieval advantage over nonhumorous materials. For example, subjects may recall the humorous sentences as a group, and they may recall these sentences before the nonhumorous sentences. If sentence humor influences recall order, then the effects of humor may be due to output interference (Roediger, 1974) or cue overload (Watkins & Watkins, 1975). Both of these phenomena depend on a within-subjects manipulation of humor. As a preliminary test of these retrieval explanations, clustering and output position of humorous sentences were scored in Experiment 2.
The comparison of the effects of humor across experimental design should answer another question concerning the effects of humor on memory. Kaplan and Pascoe (1977) found that humorous lecture material suppressed memory for nonhumor- ous material contained in the lecture. The impact of humorous sentences on memory for nonhumorous sentences in the same list was determined in Experiment 2. A comparison was made
between memory for each type of sentence in the within- subjects design to memory for the same sentences in the between-subjects design.
Method
Subjects and design. Undergraduates from introductory psychology participated for extra credit. In the within-subjects design, 20 subjects studied one list containing half humorous and half nonhumorous sentences. A second group of 20 subjects studied the complementary list containing the other versions of the sentences. In the between- subjects design, 40 subjects studied an all-humorous list and 40 subjects studied an all-nonhumorous list. Design type (within vs. between) was treated as an independent variable, and statistical tests for the main effect of design, and interactions with design, were performed as described by Erlebacher (1977). To properly partition variance in this design, it was necessary to include the same number of subjects in each condition of the between-subjects design as was included in the within-subjects design.
Materials. I used the same sentences as in Experiment 1 and constructed these sentences into four lists. Two mixed lists were constructed that contained half humorous and half nonhumorous sentences. Sentences that appeared in their humorous form in List 1 appeared in their nonhumorous form in List 2. In assigning sentences to list, care was taken to balance the length of humorous and nonhumorous sentences in each list. In List 1, there were 116 content words in the humorous sentences and 106 content words in the nonhumorous sentences. In List 2, there were 111 and 107 content words in the humorous and nonhumorous sentences, respectively. An all-humorous and an all-nonhumorous list were constructed from the same sentence set. The serial position of the two versions of each sentence was constant across all four lists. Sentences were printed on individual pages of booklets for presentation.
Procedure. Students were asked to read 20 sentences in prepara- tion for a memory test. They turned the pages in their booklet and read the sentences in time with prerecorded tones, spaced every 10 s. Following presentation of the sentences, the subjects solved simple arithmetic problems for 5 min. The students were then instructed to turn to their recall sheets and were given 10 min to recall all the sentences. The instructions stressed recall of the exact wording of the sentences. However, the subjects were also instructed to recall any part of a sentence if they could not recall the entire sentence or the exact wording of a sentence.
Results and Discussion
Scoring. Each recall protocol was scored by two indepen- dent raters. Discrepancies between ratings were resolved by a third rater. The number of content words recalled from a sentence was scored, and the probability of word recall was calculated by dividing the total number of content words recalled from a sentence by the total number of content words contained in the sentence. Two criteria for calculating sen- tence recall were used. With a lenient scoring criteria, a sentence was scored as recalled if two or more content words from the sentence were recalled. With a strict scoring criteria, a sentence was scored as recalled if 50% or more of the content words were recalled. A conditional recall score was also obtained by calculating the proportion of content words recalled from each recalled sentence (scored leniently), aver- aged across recalled sentences. When a subject failed to recall any sentences of a given sentence type (humorous or nonhumor-
956 STEPHEN R. SCHMIDT
Table 1 Recall of Humorous and Nonhumorous Sentences as a Function of Experimental Design in Experiment 2
Variable
Sentences (lenient) Sentences (strict) Proportion of words Words per sentence
Sentences (lenient) Sentences (strict) Proportion of words Words per sentence
Sentence type
Humorous
Within-subjects .30 .25 .20 .64
Between-subjects .28 .22 .17 .60
Nonhumorous
.23
.16
.14
.62
.28
.18
.17
.57
ous), the denominator of the conditional probability was zero. In these rare cases, the cell mean replaced the missing value.
I conducted separate analyses of variance (ANOVAs) for each dependent measure. In the first set of analyses, type of experimental design and type of sentence served as factors. Following these tests for the effects of experimental design, separate analyses were conducted on the data from the within- and the between-subjects designs. In the within-subjects analy- ses, type of sentence (humorous vs. nonhumorous), sentence (10 sentences of each sentence type), and list served as variables. In the between-subjects analyses, type of sentence and sentence served as variables. Sentence was treated as a variable so that F' and F2 statistics could be calculated. F' provides an estimate of the degree to which the results generalize to other samples of humorous sentences, whereas F2 provides an index of whether the effects generalize to the complete set of sentences tested in the experiment. Finally, an analysis of the order of sentence recall was performed to measure recall position and clustering in recall. A summary of the results can be found in Table 1. Except where noted, all reported effects were significant at the .05 level.
Lenient sentence recall. A greater proportion of humorous sentences was recalled (M = .29) than nonhumorous sen- tences (M = .26), F(1, 84) = 6.09, MSC = .0105. However, the effect of humor interacted with experimental design, F( 1,84) = 7.70, MSe = .0105. In the within-subjects design, the effect of humor on sentence recall was significant, F(l, 38) = 7.45, MSC = .1316; F2(l, 9) = 11.84; F ( l , 37) = 5.34, with Ms = .30 and .23 for humorous and nonhumorous sentences respec- tively. There was also a main effect of list, F(l, 38) = 9.31, MSe = -1939. In the between-subjects design, no effect of type of sentence was obtained, F{\, 78) = .06, MS, = .0078. These results replicate the positive effects of sentence humor on memory reported in Experiment 1 but indicate that the effects were limited to the within-subjects designs.
Comparisons between individual cells across experimental design were made using Scheffe's method (Winer, 1971, p. 191) and using the error term from the test of the effect of experimental design. Memory for nonhumorous sentences in the mixed lists (M = .23) was suppressed relative to memory for the same sentences in the all-nonhumorous list (M = .28), f(39) = 2.18. Recall of humorous sentences from the mixed list
(M = .30) was not significantly better than recall of humorous or nonhumorous sentences from the lists that were not mixed (Ms = .28), <(39) = 0.87.
Strict sentence recall. The pattern of results was somewhat different when the strict measure of sentence recall was used. Once again, subjects recalled a greater proportion of humor- ous sentences (M = .23) than nonhumorous sentences (M = .17), F(l, 98) = 11.22, MSe = .0120. However, in this case the interaction with experimental design was not signifi- cant, F(l, 98) = 2.40, M5e = .0120. Nonetheless, I conducted separate ANOVAs on the within- and between-subjects' data. In the within-subjects design, there was a significant effect of humor, F(l, 38) = 11.04, F2(l, 9) = 6.92, F'(l, 21) = 4.64, with Ms = .25 and .16 for humorous and nonhumorous sentences, respectively. The effect of sentence type was not significant in the between-subjects design, F(l, 78) = 1.76. Comparisons of individual cells across experimental design revealed that sub- jects recalled humorous sentences in the within-list design (M = .25) better than humorous sentences in the between-list design (M = .22), t(39) = 2.86. Recall of the nonhumorous sentences in the within-list design (M = .16) was not signifi- cantly different from recall of the nonhumorous sentences in the between-list design (M = .18), t(39) = 0.81.
Proportion of words recalled. A greater proportion of words was recalled from humorous sentences (M = .18) than from nonhumorous sentences (M = .15), F(l, 97) = 4.83, M5e = .0056. Once again, the interaction with experimental design was significant, F(l, 97) = 4.29, MSe = .0056. In the within- subjects analysis, subjects recalled more words from humorous sentences (M = .20) than from nonhumorous sentences (M = .14),F(1,38) = 9.30, MSe = .06821; F2(l, 9) = 9.18; F'(l, 28) = 5.15. A main effect of sentence list was also observed, F(l, 38) = 15.01, MSe = .1122. In the between-subjects analysis, no effect of type of sentence was obtained, F(l, 78) = .01, MSe = .0051, with Ms = .17 for both the humorous and nonhumorous sentences.
When comparisons were made between individual means, the pattern of results followed the pattern obtained with the lenient measure of sentence recall. Recall of nonhumorous sentences from the mixed-list design (M= .14) was signifi- cantly less than recall of nonhumorous sentences from the between-list design (M = .17), t(39) = 1.80.
Words recalled per sentence recalled. No effect of sentence type was observed in the analysis of the conditional probability of recalling words from recalled sentences, F(l, 113) = 1.15, MSe = .0184. The interaction between experimental design and sentence type was not significant, F(\, 113) = .56, MSe = .0184. In the separate within- and between-subjects analyses, there were no effects of sentence type, F(l, 38) = .087, MSC - .014, and F(l, 78) = 1.38, MSe = .0212, respectively.
Organization in recall. I used the adjusted-ratio-of-cluster- ing (ARC) as a measure of sentence organization. This measure failed to reveal above-chance clustering of humorous sentences (ARC = -.03). The mean percentile output posi- tion of humorous sentences (M = 57.96) was not significantly different from the average output position of nonhumorous sentences (M = 64.56), f(33) = 1.28. These findings rule out the possibility that humorous sentences were recalled as a group or were recalled before nonhumorous sentences.
HUMOR AND MEMORY 957
In summary, the results of Experiment 2 replicated the positive effects of humor on sentence and word recall. How- ever, the effects were a joint function of the experimental design and the memory measure used. In the within-subjects design, humor improved sentence recall (strict and lenient) and word recall. However, there was no effect of sentence humor on recall of words from recalled sentences. In the between-subjects design, there was no indication that sentence humor affected memory.
The fact that the effects of humor were confined to the within-subjects manipulations challenges an explanation on the basis of arousal per se. As noted earlier, other manipula- tions of arousal affect memory in both within- and between-list designs (Walker & Tarte, 1963). Similarly, the analysis of recall order ruled out simple retrieval explanations of this interaction with experimental design. The results were somewhat consis- tent with the hypothesis that humorous sentences received increased attention when presented in the context of nonhu- morous sentences. However, if increased attention is an automatic process, why did the presence of humorous sen- tences have a negative effect on the recall of the nonhumorous sentences? This result is most consistent with the rehearsal hypothesis in which increased rehearsal of humorous sen- tences was at the expense of rehearsal on nonhumorous sentences.
Experiment 3
Experiment 3 had two main purposes: to determine whether the effects of humor generalized to a cued-recall test and to test the hypothesis that the effects of humor on memory resulted from "surprise" at encountering humorous sentences in the context of a nonhumorous experiment.
Einstein and McDaniel (1987) argued that the effects of bizarreness are consistently found in free recall but not in cued recall or recognition. They argued that bizarreness aids access to sentences in memory. Once access is achieved, bizarreness does not aid in sentence reconstruction. If the parallel between bizarre imagery and humor continues, the effects of humor will be confined to measures of sentence access in free recall. To test this hypothesis, I compared free recall and cued recall of sentences as a function of sentence humor.
The comparison between cued and free recall provides another test of retrieval explanations for enhanced memory of humorous sentences. If there is a retrieval bias favoring humorous sentences, that bias may be eliminated by cuing sentence recall. Thus, an interaction between sentence type and type of memory test would provide support for a retrieval explanation of the results.
Hirshman, Whelley, and Palij (1989) argued that subjects are surprised at reading bizarre sentences in a list of common sentences. This surprise was thought to be responsible for the enhanced memory for bizarre sentences. They found that the effects of bizarre imagery could be attenuated if subjects were told that both bizarre and common sentences were to be presented and the words bizarre and common were embedded in the appropriate types of sentences. Presumably, a similar interpretation may apply to enhanced memory for humorous sentences embedded in a list of nonhumorous sentences.
Subjects may be surprised when, during a boring memory experiment, they come upon humorous sentences. The sur- prise hypothesis may provide an explanation for why the effects of humor on memory were confined to within-subjects designs. To provide a test of the surprise hypotheses, I included the experimental variable "labeled versus not labeled" in Experi- ment 3. In the labeled conditions, subjects were told that some of the sentences were humorous and some of the sentences were nonhumorous. Each sentence type was preceded by the appropriate label during presentation. These manipulations should effectively remove any surprise a subject may have concerning the materials. If an effect of sentence humor is found in the labeled conditions, then surprise can be ruled out as an explanation.
Method
Subjects and design. Eighty introductory psychology students par- ticipated in partial fulfillment of a course requirement. Ten subjects served in each cell of a 2 (labeled vs. not labeled) x 2 (free vs. cued recall) x 2 (List 1 vs. List 2) between-subjects factorial. Type of sentence (humorous vs. nonhumorous) was manipulated within sub- jects.
Materials. I used the same sentence sets as those used in Experi- ments 1 and 2. One or two words that were central to the meaning of each sentence were selected as retrieval cues. The cue words were underlined in all sentence sets, as shown in the Appendix. Two sentences had the word salesman as a cue, so one of these sentences was dropped from the material set. A second sentence was dropped to create balanced lists of humorous and nonhumorous sentences. The remaining 18 original sentences were divided into two lists, creating a different set than was used in Experiments 1 and 2. Each list contained 9 humorous and 9 nonhumorous sentences. Humorous sentences in one list were matched with the nonhumorous version of the sentence in the same serial position in the other list. Sentences were printed on individual pages for presentation. In the labeled conditions, the words humorous and nonhumorous were printed in uppercase letters above humorous and nonhumorous sentences, respectively.
Procedure. The procedure was similar to the procedures of Experi- ments 1 and 2. However, before sentence presentation in the labeled conditions, students were told that half of the sentences were intended to be humorous and half of the sentences were intended to be nonhumorous. They were also told that the sentences were labeled, and that they should read the sentence label before reading each sentence. Subjects in the cued-recall conditions were informed that each sentence contained a word that would serve as a retrieval aid. They were further instructed that cue words were underlined, and that the cues would appear on the memory test. As in the previous experiments, students read the sentences in time with prerecorded tones, performed 5 min of arithmetic, and then spent 10 min on the recall test.
Results and Discussion
The measures of recall reported in Experiments 1 and 2 were again used. I calculated separate ANOVAs on each measure with label, type of test, list, type of sentence, and sentence serving as variables. The results of each analysis are reported below. A summary of the results is reported in Table 2.
Lenient sentence recall. The proportion of humorous sen- tences recalled (M = .42) exceeded the proportion of nonhu- morous sentences recalled (M = .35), F(l, 72) = 7.82, MSe -
958 STEPHEN R. SCHMIDT
Table 2 Recall of Humorous and Nonhumorous Sentences as a Function of Labeling and Type of Memory Test in Experiment 3
Variable
Sentences (lenient) Sentences (strict) Proportion of words Words per sentence
Sentences (lenient) Sentences (strict) Proportion of words Words per sentence
No label
Non- Humorous humorous
.36
.29
.26
.74
.52
.36
.31
.59
Free recall .24 .16 .15 .63
Cued recall .45 .27 .24 .55
Label
Humorous
.29
.25
.21
.68
.51
.39
.32
.62
Non- humorous
.22
.18
.14
.60
.47
.28
.26
.55
.2219; F2(l, 8) = 6.73; F'(l, 25) = 4.01, p < .10. Cued recall (M = .49) led to a greater proportion of sentences recalled than free recall (M = .28), F(l, 72) = 46.44, MSe = .3364. No other main effects were observed. There was an interaction between sentence type and list, F(l, 72) = 6.62, MS€ = .2219. Neither the interaction between label and type of sentence, F(l, 72) = .45, nor the interaction between type of test and type of sentence, F(l, 72) = .61, MSt = .2219, was significant.
Strict sentence recall. With the strict measure of recall, essentially the same pattern of results was obtained. The only notable difference was that the magnitude of the effect of humor on memory was larger with the strict scoring criterion than with the lenient scoring criterion, with F(l, 72) = 25.52, M5e = .1450; F2(l, 8) = 8.96; andF'(l, 14) = 6.90.
Proportion of words recalled. The effects of humor on word recall closely followed the effects found with sentence recall. That is, a greater proportion of words was recalled from humorous sentences (M = .27) than was recalled from nonhu- morous sentences (M = .20), F(\, 72) = 21.80, MSe = .0942; F2(l, 8) = 9.61; F'(l, 16) = 6.95. Cued recall (M = .28) exceeded free recall (M = .19),F(1, 72) = 16.48, MS, = .1945. The interactions between sentence type and type of test, F(l, 72) = .49, and sentence type and labeling, F(l, 72) = .40, were not significant, MSes = .0942.
Words recalled per sentence recalled. Humorous sentences led to greater recall of sentence content (M = .66) than nonhumorous sentences (M = .58), F(l, 72) = 10.66, MSe = .0200. There was a significant effect of type of test, F(l, 72) = 10.45, but no effect of labeling (F = .33, MSts = .0260). No significant interactions involving type of test or labeling were obtained.
It was not possible to calculate the traditional F' when recall was conditionalized on sentence recall. However, the propor- tion of words recalled for each sentence, given that a subject recalled that sentence, was calculated. The data were then collapsed across subjects, and sentences were treated as a random variable and F2 calculated. In addition, the minimum F' statistic (Clark, 1973) was calculated. These analyses provided weak support for the hypothesis that humorous sentences were remembered in greater detail than nonhumor- ous sentences, F2(l, 17) = 4.28, minimum F'(l, 32) = 3.04, both/js between .05 and .10.
In summary, the major findings of this experiment were that the effects of humor on sentence memory occurred indepen- dently of labeling and type of memory test. Telling subjects that the sentence they are about to read is humorous does not diminish the effects of humor on memory. Thus, the effects of humor are not due to the surprise resulting from unexpectedly encountering a humorous sentence. In addition, the effects of humor occurred in both free and cued recall. In the cued-recall test, recall order was controlled. As a result, the average recall position of humorous material was not less than the recall position of nonhumorous material. The positive effect of humor in cued recall thus eliminated output interference as an explanation of those effects.
Experiment 4
Researchers have noted a negative correlation between ratings of humor and recall (Gelb & Zinkhan, 1986). This observation is difficult to reconcile with the results reported above. To further investigate the relation between rated humor and memory in Experiment 4, I asked subjects to rate the sentences on humor before taking the memory test. I was also interested in investigating the effects of humor on inciden- tal memory. Thus, no mention was made of the memory test before the free-recall task. If the effects of humor on memory are the result of an intentional rehearsal strategy, then those effect may be attenuated in an incidental memory task.
Method
Subjects and design. There were 48 subjects in this experiment selected from the same pool as the subjects in Experiments 1-3. A within-subjects manipulation of humor was used. There were two sentence lists, each half humorous and half nonhumorous. Twenty- four subjects were tested on each list.
Materials. I used the same 18 sentences as those in Experiment 3. Unlike the previous experiments, the sentences in this experiment were videotaped. The sentences were recorded in the middle of the screen, with one sentence displayed at a time. Each sentence was displayed for 10 s, followed by a brief tone. The tone signaled the subjects to look up from their answer sheets and read the next sentence.
Procedure. With the following exceptions, the procedure was identical to that of Experiments 1-3. The students were told that the experiment was concerned with the relation between the perception of humor and mathematical abilities. The students were asked to rate how humorous they found each sentence on a 7-point Likert-type scale. A rating of 1 indicated that they did not find the sentence humorous. A rating of 7 indicated that they found the sentence very humorous. The sentences were displayed on a large video monitor in the front of a small classroom. Following the rating task, the students were asked to solve a variety of mathematics problems for 5 min. They were then given the same instructions for the free-recall test that was used in the earlier experiments.
Results and Discussion
I performed analyses similar to those of the previous experiments. In addition, the relation between the subjective ratings of sentences and sentence recall was analyzed. These analyses are reported below and summarized in Tables 3 and 4.
HUMOR AND MEMORY 959
Table 3 Recall and Sentence Ratings of Humorous and Nonhumorous Sentences in Experiment 4
Experiment A
Variable
Sentences (lenient) Sentences (strict) Proportion of words Words per sentence
Humor rating
Sentence type
Humorous
.26
.21
.16
.64
3.89
Nonhumorous
.23
.14
.13
.55
2.73
Lenient sentence recall. The effect of sentence humor on lenient sentence recall was not significant, F(l, 46) = 1.46, MSe = .1338. The mean recall of humorous sentences was .26, whereas the recall of nonhumorous sentences was .23.
Strict sentence recall. With the strict scoring criteria, more humorous sentences were recalled (M = .21) than nonhumor- ous sentences (M = .14), F{\, 46) = 10.34, MSC = .0942. This effect was marginally reliable in the item analyses, F2(l, 8) = 4.51; F'(l, 15) = 3.51,/w < .10.
Proportion of words recalled. Word recall from humorous sentences (M = .16) was greater than recall from nonhumor- ous sentences (M = .13), F(l, 46) = 5.99, MSe = .0460. However, this effect was not reliable in either of the item analyses, F2(l, 8) = 3.08; F ( l , 17) = 2A9,ps > .10.
Words recalled per sentence recalled. The probability of recalling words from recalled humorous sentences (M = .64) exceeded the probability of recalling words from recalled nonhumorous sentences (M = .55), F(l, 46) = 7.30, MSt = .0220. However, this effect was not significant in the item analyses, F2(l, 15) = 3.13, .05 < p < .10; minimumF'(l, 29) = 2.19,p > .10.
Ratings of humor. The first analysis performed on the rating data determined the relation between the ratings of sentences and the experimental definitions of humorous and nonhumorous sentences. This analysis replicated Experiment 1 ratings with a different sample of subjects and with the subset of sentences used in this experiment. Humorous sentences received higher ratings of humor (M = 3.89) than nonhumor- ous sentences (M = 2.73), F(l, 46) = 81.32; F2(l, 8) = 40.72; F'(l, 17) = 27.31. However, unlike the original ratings, there was much overlap in the distributions of rated humor for humorous and nonhumorous sentences. The frequency distri- butions for the ratings are presented in the top portion of Figure 1.
The difference between the subjective ratings and the experimentally defined humor of sentences prompted further
Table 4 Analysis of Sentence Ratings for Recalled Versus Nonrecalled Sentences as a Function of Sentence Type in Experiment 4
Sentence type Sentence status Humorous Nonhumorous Recalled Nonrecalled
Humorous
Non-Humorous
in c
CD O
25
20
15
10
5
0
Experiment 6
4.25 3.73
3.21 2.57
Humor Rating
Figure 1. Frequency distributions of sentence ratings from Experi- ments 4,5, and 6.
analyses. For each subject, sentences were classified as re- called or nonrecalled. The mean humor ratings of recalled sentences were then compared with the mean ratings of nonrecalled sentences. If humor enhances memory, then recalled sentences should have received higher average ratings than nonrecalled sentences. This finding should be indepen- dent of whether or not a sentence was classified as humorous or nonhumorous by the experimenter. One advantage of this type of analysis was that it was not affected by a subject's general bias toward rating sentences as humorous or nonhumor- ous. A 2 x 2 x 2 (List x Sentence Type: humorous vs. nonhumorous x Sentence Status: recalled vs. nonrecalled) ANOVA was calculated. Sentence rating served as the depen- dent variable. A summary of the results is presented in Table 4. As predicted, recalled sentences were rated as more humorous (M = 3.73) than nonrecalled sentences (M = 3.15), ^(1,46) = 9.26, MSC = 1.73. The interaction between sentence type and sentence status was not significant, F(l, 46) = .19, MSe = .9296. This analysis demonstrated that subjects' perception of humor predicted their recall of sentences. Item analyses could
960 STEPHEN R. SCHMIDT
CO o
.a ca .a o
.40
.35
.30
.25
.20
.15
.10
.05
.40
.35
.30
.25
.20
.15
.10
.05
.45
.40
.35
.30
.25
.20
.15
.10
.05
Experiment 4
• Sentences Lenient • Sentences Strict A Words
Experiment 5
• i • i • • i . • . . i • • • . 1 1 • 1 1 1 . . • 1 1 .
Experiment 6
1 1.5 2 2.5 3 3.5 4 4 5 5 5.5 6
Mean Humor Rating
Figure 2. Recall of sentences and words from Experiments 4,5, and 6 divided into six categories of rated humor. Each subject's relative ratings of the sentences were used to sort the sentences into the six groups.
not be performed because the data had to be collapsed across sentences to calculate mean ratings for recalled and nonre- called sentences for each subject.
To further explore the relation between rated humor and recall, I ranked each subject's sentence ratings from high to low. On the basis of sentence rank, the 18 sentences were then placed in 6 groups of 3, and memory performance was evaluated for each group of six sentences by sentence rank. In this manner, I evaluated each subject's performance for those sentences given the highest three ratings down to the lowest three ratings. Researchers studying the relation between arousal and memory have used similar analyses. For example, in a classic study Kleinsmith and Kaplan (1964) sorted non- sense syllables into high- and low-arousal stimuli for each subject on the basis of the subject's galvanic skin response. In the present study, I calculated ANOVAs with humor rating (in six categories) and list serving as variables, and lenient
sentence recall, strict sentence recall, and words recalled serving as dependent variables. Conditional recall could not serve as a dependent variable because a large number of observations were undefined.1 A summary of sentence and word recall, broken down in this manner, is presented in the top portion of Figure 2.
This analysis revealed a main effect of humor on lenient sentence recall, F(5, 230) = 3.59, MSe = -0695, strict sentence recall, F(5, 230) = 6.15, MSe = .0491, and word recall, F(5, 230) = 5.68, MSe = .0292. Thus, higher ratings of sentence humor were associated with higher sentence recall. Item analyses could not be performed because each subject was free to rank the items in a different order.
The results of Experiment 4 provided some additional support for the role of intentional rehearsal in producing the positive effects of humor on memory. That is, under incidental learning conditions the effects of experimentally defined hu- mor were attenuated. However, intentional rehearsal does not provide a complete explanation of the effects of humor. With the strict scoring criteria, more sentences defined by the experimenter as humorous were recalled than nonhumorous sentences. When ratings of humor were considered, recalled sentences were rated as more humorous than nonrecalled sentences, and sentences receiving high ratings of humor were recalled better than sentences receiving low ratings of humor. It is unclear why rehearsal should be tied to rated humor in this way. Perhaps subjects rehearse sentences as part of the rating task. This possibility is explored next. To the extent that rated humor reflects the attention-getting properties of the sen- tences, these results support the role of attention and the incongruity hypothesis.
Experiment 5
Experiment 5 had two main purposes: to further explore the relation between rated humor and memory and to further explore the effects of humor on the conditional probability of word recall given sentence recall.
In Experiment 4, subjects rated sentence humor relative to sentences generally. As a result, the humorous sentences did not receive particularly high ratings of humor, and there was a good deal of overlap in the ratings of humorous and nonhumor- ous sentences (see Figure 1). Because of the large overlap, the manipulation of sentence humor was attenuated, and the effects of experimentally defined humor on memory were attenuated. In Experiment 5, an alternative rating task was used. Subjects were given two sample sentences and they were asked to rate all the sentences in the experiment relative to those sample sentences and relative to other sentences in the experiment. The sample sentences provided a context for subjects to rate the first several sentences in the list. In this way, I hoped to confine their attention to the relative humor of the sentences in the experimental context. This should pull the
1 When a sentence is not recalled, the conditional probability of word recall given sentence recall is undefined. Because sentences were divided into six groups, it was quite common for subjects not to recall any sentences from a group. A large number of missing observations biases the outcome of any summary statistics. For this reason, conditional probability analyses were not completed.
HUMOR AND MEMORY 961
distributions of rated humor for humorous and nonhumorous sentences further apart. I hoped that under these conditions the effects of experimentally defined humor would reemerge.
When the rating task is limited to sentences in the experi- ment, one may assume that subjects select sentences from the list to serve as anchors. These anchors may receive increased rehearsal relative to other sentences in the list. If incidental rehearsal is responsible for the effects rated humor on memory, these effects should appear as good memory for the selected anchors. In the current study, subjects should be equally likely to select low- and high-humor sentences as anchors. Thus, the rehearsal hypothesis predicted that the relation between rated humor and memory would be a U-shaped function. Those items receiving the highest and lowest ratings of humor should be best retained.
According to the incongruity hypothesis, features shared across the sentences in the list should be maintained in working memory. Sentences that contain novel features should lead to incongruity and increased attention, which in turn should lead to enhanced long-term retention. Presumably, a few sentences will be encoded as "really funny" and receive relatively high humor ratings. Only these sentences will be incongruent with, attract more attention than, and be better remembered than, the rest of the list. The incongruity hypoth- esis does not predict that sentences rated extremely low on the humor scale will also be well remembered. As seen in Figure 1, many of the sentences were rated low on the humor scale. Only those sentences at the other end of the distribution stand out from the rest of the list. Perhaps if the sentences ranged from really sad to really funny, the distribution of rated humor would be closer to a normal curve. Under these conditions, both tails of the distribution would be incongruent with the bulk of the list. However, one would not expect this with the current set of sentences.
Experiments 1-4 uncovered marginal effects of humor on conditional word recall. To provide a definitive answer concern- ing the effects of humor on conditional word recall, I designed Experiment 5 to have sufficient power to detect a relatively small effect. Using the estimated effect size from Experiment 4,1 determined that n must be approximately 60 for power to reach the .80 standard. In Experiment 5, data from 62 subjects were obtained.
Another purpose of Experiment 5 was to provide a test of the hypotheses developed by Dixon, Willingham, Strano, and Chandler (1989). They asked college students to read maga- zine articles that contained peripheral advertisements and cartoons. Subjects were then given multiple choice tests on their memory for the content of the article as well as their memory for advertisements and cartoons. In addition, the subjects were asked to rate themselves on their sense of humor. The rating scale was then used to split the subjects into high and low sense of humor groups. The only significant finding was that subjects with a high sense of humor remem- bered the content of the cartoons better than the subjects with a low sense of humor. They argued that sense of humor determines attention to and memory for humorous material. In Experiment 5, subjects were asked to rate their sense of humor. They were then subdivided to determine whether self-reported sense of humor was related to recall.
Table 5 Recall and Sentence Ratings of Humorous and Nonhumorous Sentences in Experiment 5
Variable
Sentences (lenient) Sentences (strict) Proportion of words Words per sentence
Humor rating
Sentence type
Humorous
.25
.18
.16
.60
3.97
Nonhumorous
.22
.14
.13
.53
2.53
Method
Subjects and design. Sixty-five students were sampled from various psychology classes during the summer term in exchange for extra course credit. The subjects were tested in groups ranging in size from 15 to 20. One subject was dropped for failure to complete the rating scale, and two subjects were dropped to balance the number of subjects receiving each list. Thus, I used data from two groups of 31 subjects. Each subject read one of the two lists of sentences in a within-list manipulation of sentence humor.
Materials. The 18 sentences used in Experiments 3 and 4 served as the to-be-remembered items. The 2 sentences from the original list that were not used as experimental sentences (see Sentences 1 and 12 in the Appendix) were used as sample sentences in the instructions. For all conditions, Sentence 1 was presented in its nonhumorous form, and Sentence 12 was presented in its humorous form.
Procedure. The procedure closely followed that of Experiment 4. The only significant difference was in the rating instructions. The experimenter told the subjects that "some of the sentences are intended to be humorous, whereas some of the sentences are not supposed to be humorous." The two sample sentence were then read. The subjects were then instructed, "When you read each sentence, decide how funny the sentence is compared to other sentences in the experiment. Use the sample sentences as well as the other sentences when making your comparisons."
After completing the recall task, the subjects were asked to rate their sense of humor on a scale ranging from 1 (indicating whether they found life in general serious and were difficult to amuse) to 7 (indicating whether they found life in general humorous and could easily find humor in an experience).
Results and Discussion
The data analyses mirrored those in the previous experi- ments. The results are summarized in Table 5 (see also Figures 1 and 2).
Lenient sentence recall. The effect of experimentally de- fined humor on lenient sentence recall was not significant, F(l, 60) = 2.04, MSe = .1587. The proportion of humorous sentences recalled was .25, whereas the proportion on nonhu- morous sentences recalled was .22.
Strict sentence recall. The effect of sentence type on the strict measure of recall was significant, F(l, 60) = 4.50, MSC = .1244, with means equal to .18 and .14 for humorous and nonhumorous sentences, respectively. However this effect was not significant in the item analyses, F2(l, 8) = 2.96, and F'{\, 20) = 2.13. These results were similar to the results reported in Experiment 4.
962 STEPHEN R. SCHMIDT
Proportion of words recalled. The probability of recalling words from humorous sentences (M = .16) exceeded the probability of recalling words from nonhumorous sentences (M = .12), F(l, 60) = 4.03, MSe = .0663. Once again, these results were not significant in the item analyses, F2(l, 8) = 2.79; F ( l , 21) = 2.00.
Words recalled per sentence recalled. Conditional word re- call from humorous sentences was 0.60, whereas recall from nonhumorous sentences was .53. This effect was significant in the traditional ANOVA, F(l, 60) = 5.59, MSe = .0215, but as in the previous experiments, this effect was not significant in the item analyses, F2(l, 14) = 2.83, minimum F'(l, 30) = 1.88, ps > .10.
Ratings of humor. Humorous sentences (M = 3.97) re- ceived higher ratings of humor than nonhumorous sentences (M = 2.53), F(l, 60) = 181.89, MSe = .3538; F2(l, 8) = 45.42; F'(l, 12) = 36.43. The distribution of rated humor is presented in the middle portion of Figure 1. As was predicted, the ratings of humorous and nonhumorous sentences were pulled apart relative to those ratings in Experiment 4.
To explore the relation between ratings of humor and recall, I again performed ANOVAs for recalled and not recalled sentences reported in Experiment 4. Recalled sentences re- ceived higher ratings of humor (M = 4.06) than nonrecalled sentences (M = 2.69), F(l, 60) = 106.49, MSt = 1.083. The absence of an interaction between sentence type and recall status, F(l, 60) = 1.08, MSC = .7644, demonstrated that this pattern of results was obtained for both humorous and nonhumorous sentences.
As in Experiment 4, each subject's ratings of sentences were used to sort the sentences into six groups of rated humor for each subject. Recall was then evaluated for these sentence groups. A summary of the data is reported in middle portion of Figure 2. The effect of sentence type was significant for lenient sentence recall, F(5, 300) = 2.91, MSe = .0592, strict sentence recall, F(5, 300) = 4.54, MSe = .0421, and word recall, F(5, 300) = 5.12, MSe = .0228.
Self-rated humor. Self-reported sense of humor ranged from 2 to 7. Thirty-two subjects rated themselves at 5 and below, whereas 30 subjects rated themselves above 5. This split served as a variable in an ANOVA in which type of subject (high and low sense of humor) and type of sentence (humorous and nonhumorous) served as variables. Two measures of recall were used: probability of lenient sentence recall and the conditional probability of word recall given sentence recall. No main effects or interactions involving subject type were ob- served, all Fs(l, 60) < 2.30. Correlations between self-rated humor and recall of humorous and nonhumorous sentences were also calculated; however, none approached significance (largest r = .20). The hypothesis was also tested by sorting subjects into high- and low-humor groups on the basis of their ratings of the sentences. There were no group differences in memory performance or interactions between group and type of sentence (Fs < 1.0). Thus, Dixon et al.'s (1989) findings of different effects of humor for different types of subjects were not supported.
In summary, Experiment 5 provided a replication and extension of the effects reported in Experiment 4. The effects of experimentally defined humor were attenuated in an inciden-
tal tasks in which subjects were asked to rate the humor of sentences. This result was obtained when open-ended ratings of humor were used (Experiment 4) and when subjects were asked to confine their attention to the relative humor of sentences occurring in the experiment (Experiment 5). Thus, even when humorous and nonhumorous sentences clearly received different ratings of humor, the experimental manipu- lation of humor had a relatively small effect on retention. However, subjects' definition of humor was strongly related to memory, with memory increasing as rated humor increased. The analysis of recall as a function of rated humor provided evidence against the hypothesis that subjects rehearsed an- chors used in the rating task, and that this rehearsal was responsible for increased retention of humorous material.2
Instead, the results support the hypothesis that subject-defined humor affects the amount of attention an individual devotes to a sentence.
Experiment 6
The main purpose of Experiment 6 was to explore the effects of humor on incidental memory in a between-subjects design. Experiment 2 demonstrated that, in intentional learning, the effects of experimentally defined humor were only obtained in a within-subjects design. On the basis of the attention hypoth- esis developed above, one would expect that the same pattern of results would be found in incidental learning. A more interesting prediction is that relative humor in a list should predict relative recall of sentences from the list. From the experimenters' perspective, both the humorous and nonhumor- ous lists in a between-subjects design are homogeneous. However, from the subjects' perspectives each list contains sentences that vary in humor. The amount of attention a sentence receives may still vary with perceived humor, and rated humor should still predict recall. To test these predic- tions, I used a rating task similar to that of Experiment 4, and subjects were given an unexpected recall test. The open-ended rating scale was used so that differences between ratings of humorous sentence from one group of subjects and ratings of nonhumorous sentences by the other group of subjects could be detected.
Method
Subjects and design. Fifty-two subjects participated for extra course credit in a between-subjects design with 26 subjects in each group.
Materials. The same 18 sentences as those from Experiments 3-5 were again used. Two lists were constructed and recorded on video tape. One list contained the humorous versions of the sentences, whereas the other list contained the nonhumorous versions.
2 With some additional assumptions, the differential rehearsal hypothesis may still be able to explain the results of Experiment 5. Subjects may have chosen high- and low-humorous items for anchors, but, given the structure of the list, may have frequently changed their low-humor anchor. Shifting anchors may have reduced the number of rehearsals that items ranked low on the humor scale received. This explanation is obviously post hoc. Also, it fails to explain why subjects should be more likely to change anchors at one end of the rating scale than at the other.
HUMOR AND MEMORY 963
Procedure. The procedure was nearly identical to the procedure used in Experiment 4. Subjects were asked to read each sentence and rate how humorous they found the sentence " . . . relative to other sentences or humorous sayings that you have heard prior to coming to this experiment." Following the rating task, subjects performed 5 min of arithmetic and then were given 10 min to freely recall the sentences.
Results and Discussion
The recall measures were again calculated. Each measure was analyzed in a 2 (sentence type) x 18 (sentence) ANOVA. Sentence type was a between-subjects variable, whereas sen- tence was a within-subjects factor. The results are summarized in Table 6 and in the lower portions of Figures 1 and 2.
Lenient sentence recall. Recall of humorous sentences (M = .25) was not significantly different from recall of nonhu- morous sentences (M = .21), F(l, 50) = 1.54, MSe = .1776. There was a significant effect of sentence, F(17, 850) = 10.52, MSe = .1486, indicating that some of the sentences were consistently more memorable than others. The significant interaction between sentence and sentence type, F(17, 850) = 1.64, MSe = .1486, demonstrated that the humorous versions of one subgroup of sentences were well remembered whereas the nonhumorous versions of a different subgroup of sentences were relatively well remembered.
Strict sentence recall. The effect of sentence type on the strict measure of recall was not significant, F(l, 50) = .83, MSe = .2175, with means equal to .15 and .13 for humorous and nonhumorous sentences, respectively. The main effect of sentence, F(17, 850) = 6.07, and the interaction between sentence and sentence type, F(17, 850) = 2.48, MSes = .1021, were again significant.
Proportion of words recalled. The probability of recalling words from humorous sentences (M = .15) did not exceed the probability of recalling words from nonhumorous sentences (M = .13), F(l, 50) = 1.09, MSC = .1170. The main effect of sentence, F(17, 850) = 9.34, and the interaction between sentence and sentence type, F(17, 850) = 2.11, MSes = .0664, were significant.
Words recalled per sentence recalled. Unlike the previous experiments, there was no indication that word recall from recalled humorous sentences (M = .59) was greater than word recall from recalled nonhumorous sentences (M = .63), F(l, 50) = .47,A/Se = .0
Table 6 Recall and Sentence Ratings of Humorous and Nonhumorous Sentences in Experiment 6
Variable
Sentences (lenient) Sentences (strict) Proportion of words Words per sentence
Humor rating
Sentence type
Humorous
.25
.15
.15
.59
3.59
Nonhumorous
.21
.13
.13
.63
3.06
Ratings of humor. Humorous sentences received higher ratings of humor (M = 3.59) than nonhumorous sentences (M = 3.06), F(l, 50) = 5.85, MSe = 11.32; F2(l, 17) = 8.12; F'(l, 58) = 3.49, p < .10. However, this effect is small in magnitude relative to the Fs reported in Experiment 4 (F = 81.32) and Experiment 5 (F = 181.89). The distribution of rated humor is presented in the lower portion of Figure 1. As can be seen, the distributions for humorous and nonhumor- ous sentences were similar. Comparing this distribution with the distributions in Experiments 4 and 5, one can see that ratings of humor are highly context dependent. It should not be surprising that the effects of humor on memory were equally context bound. In addition, the distributions of rated humor were skewed positively for both types of list used in Experi- ment 6, with relatively few sentences receiving a rating of 7 on the humor scale. Thus, independent of sentence type, sen- tences rated as relatively humorous in a list should stand out relative to other sentences in that list.
The analysis of ratings for recalled and nonrecalled sen- tences reported in Experiments 4 and 5 were again performed. Recalled sentences received higher ratings of humor (Af = 3.84) than nonrecalled sentences (M = 3.17), F(l, 50) = 22.81. The interaction between sentence type and recall status was not significant, F(l, 50) = .08. The MSes were .5211 for both effects. Recall as a function of rated humor was analyzed in the same manner as in Experiments 4 and 5. The results were summarized in the lower portion of Figure 2. The effects of sentence group were significant for lenient sentence recall, F(5,250) = 7.56, MSe = .1690, strict sentence recall, F(5,250) = 2.79, MSe = -1173, and word recall, F(5, 250) = 4.22, MSt = .0788. None of the interactions with type of sentence ap- proached significance, Fs(5,250) < 1.66.
These results indicate that even in a between-subjects manipulation of sentence humor, relative humor of sentences within a list is a good predictor of recall. Recalled sentences received higher ratings of humor than nonrecalled sentences. Recall increased as relative ratings of humor increased. Both of these effects were independent of whether or not a sentence was defined as humorous by the experimenter. These findings reinforce the idea that the effects of humor on incidental memory are context bound and result from the relative attention a sentence receives.
If one wishes to risk between-experiment comparisons, the results provide some additional support for the role of inten- tional rehearsal in memory for humorous materials. In inten- tional learning, increased recall of humorous material oc- curred at the expense of recall of nonhumorous material (Experiment 2). This trade-off did not occur in incidental learning. Recall of nonhumorous material with incidental learning in within-subjects designs (Experiments 4 and 5) did not differ from recall of the same material in the between- subjects design (Experiment 6). For example, compare the strict sentence recall in Experiment 4 with the strict sentence recall in Experiment 6. In Experiments 4 and 6, I used essentially the same procedures, making this comparison sound. In addition, a significant effect of humor was observed in Experiment 4 with the strict measure of sentence recall. Mean recall of humorous and nonhumorous sentences was .21 and .14 in Experiment 4, and .15 and .13 in Experiment 6,
964 STEPHEN R. SCHMIDT
respectively. The enhanced recall of humorous sentences in Experiment 4 was not at the expense of nonhumorous sen- tences.
General Discussion
In previous investigations into the effects of humor on memory, humorous and nonhumorous materials have not been equated on important mnemonic variables. The results re- ported above provided the first convincing evidence of a positive effect of humor on sentence memory. The effects were obtained in free recall, cued recall, in incidental and inten- tional memory, and when subjects were warned about the humorous nature of the sentences.
The effects of humor on sentence memory were a joint function of the memory measure used, the context in which the sentences appeared, whether memory was intentional or incidental, and how humor was denned. First consider how the effects of humor varied with the memory measures used. Essentially the same pattern of results was found with three measures of memory: lenient sentence recall, strict sentence recall, and words recalled. This should not be surprising given that the correlations between these measures ranged from .81 to .95 in the pilot study. In addition, each measure is largely a reflection of sentence access. The effects of humor on reten- tion of sentence detail were less compelling. In Experiments 3, 4, and 5, but not in Experiments 1, 2, and 6, the conditional recall of words from recalled humorous sentences was greater than that from nonhumorous sentences. However, the effect of sentence humor on conditional recall was not significant in any of the item analyses. This suggests that the effects of our manipulation on conditional recall are small at best and may stem from a small set of sentences tested in these experiments. Research with other humorous materials (Schmidt & Willi- ams, 1992) has failed to find significant effects of humor on conditional word recall. At present, there is little support for the hypothesis that humor enhances memory for sentence detail.
Second, consider the effects of context on memory for humorous materials. In within-lists manipulations of sentence humor, humorous sentences were better retained than nonhu- morous sentence. In the between-list manipulations, memory for sentences in an all-humorous sentence list did not exceed memory for sentences in an all-nonhumorous sentence list. Rated humor was influenced by sentence context, leading to a smaller difference in rated humor in the between-list than in the within-list manipulation of sentence type. Nonetheless, subjects gave higher average ratings of humor to sentences in the all-humorous sentence list than to sentences in the all-nonhumorous sentence list. Thus, in the between-list de- sign perceived humor was successfully manipulated without the manipulation having an impact on memory performance.
The effects of humor on sentence retrieval were also influenced by the nature of the encoding task. In intentional memory, sentences rated as normatively more humorous were recalled better then sentences rated as less humorous. This effect was attenuated in incidental tasks in which subjects rated humor. Nonetheless, the strict sentence recall measure showed effects of humor in both Experiments 4 and 5.
In several of the experiments, the relation between humor and memory was explored with humor defined subjectively rather than experimentally. With the subjective definition, the relation between humor and memory was stable, appearing in both within- and between-list manipulations of humor and appearing in incidental memory.
Beyond demonstrating the effects of humor on memory, the research reported in this article narrows the field of theoretical explanations for these effects. One potential explanation is that retrieval processes favored the recall of humorous mate- rial at the expense of nonhumorous material in the same list. However, humorous sentences were not recalled before nonhu- morous sentences. In fact, there was no evidence that organiza- tion in recall was based on sentence type. Furthermore, the effects of humor on sentence recall were the same for cued and free recall. These findings rule out simple retrieval explana- tions of the enhanced recall of humorous material.
A second potential explanation of the results is that subjects were surprised to find humorous materials in a memory experiment. A direct test of the surprise hypothesis was provided in Experiment 3. Subjects were told that some of the sentences were humorous, and they were warned when humor- ous and nonhumorous sentences were about to be read. Contrary to the surprise hypothesis, humor still enhanced memory.
A third potential explanation is that humor elicits physiologi- cal arousal in the form of a defensive reaction (McGhee, 1983), and this arousal is responsible for enhanced memory. However, simple arousal explanations suggest that humor would enhance memory in both within- and between-subjects manipulations (Walker & Tarte, 1963). The results of Experi- ments 2 and 6 contradicted this prediction by demonstrating that the effects of sentence humor were confined to within- subjects comparisons. A possible rejoinder to this argument is that the manipulation of humor per se is context bound, so that in a between-list manipulation of humor, potential arousal differences between the all-humorous sentence condition and the all-nonhumorous sentence condition disappear. The ratings data collected in Experiment 6 challenged this interpretation by demonstrating that differences in perceived humor survived the between-list manipulation.
Another potential explanation for the results reported above is a differential rehearsal hypothesis. According to this hypothesis, subjects intentionally rehearsed humorous mate- rial at the expense of nonhumorous material. There were various sources of support for this hypothesis. First, with intentional instructions, recall of humorous material had a negative impact on recall of nonhumorous material from the same list (Experiment 2). Second, the effects of experimentally manipulated humor were attenuated with incidental learning instructions (Experiments 4 and 5). And third, the negative effects of humor on recall of nonhumorous material disap- peared in incidental learning (cf. Experiments 4 and 6). However, rehearsal fails to provide a complete explanation of the effects of humor on memory. If rehearsal occurs only in intentional learning, then rehearsal cannot explain the fact that rated sentence humor predicted recall in incidental learning. Alternatively, rehearsal may occur in incidental learning as part of performing the rating task. However,
HUMOR AND MEMORY 965
according to this version of the rehearsal hypothesis, both high- and low-humor anchors should receive increased re- hearsal and thus be well retained relative to other items in the list. Contrary to this prediction, only sentences rated as highly humorous were recalled better than other items in the list.
Another explanation of these results is that humorous sentences receive more attention than nonhumorous sen- tences. According to this hypothesis, the amount of attention an item receives is highly context dependent. One explanation for such attentional shifts is the incongruity hypothesis (Schmidt, 1991). Incongruity results from a mismatch between active cognitive structures and newly presented material. There were at least two potential sources of incongruity in the above experiments. First, as noted in the introduction, incon- gruity may arise within a humorous sentence. However, this source of incongruity should not be context bound and should have enhanced recall of humorous material independent of context. The experiments reported above provided little sup- port for this hypothesis. A second source of incongruity derives from comparisons across sentences and is thought to be responsible for the increased attention received by humorous sentences.
Assume that the content of recent sentences is maintained in active memory during the presentation of each new sen- tence. Incongruity should arise to the extent that a new sentence is different from previous sentences. A particularly humorous sentence should stand apart from previously pre- sented sentences that were not as humorous. This would not happen for all the sentences defined by the experimenter as humorous. But in a within-list manipulation of humor, it should be more likely to occur for sentences defined as humorous than for sentences defined as nonhumorous. In the between-list manipulations of sentence humor, a few of the sentences were perceived as very humorous. This was appar- ently true for both the humorous and nonhumorous lists of sentences and is reflected in the ratings data summarized in the lower portion of Figure 1. The sentences receiving a 6 or a 7 on the rating scale stood apart from other sentences in the list, received extra attention, and were well remembered.
The attention hypothesis has two major shortcomings. First, it fails to explain the negative effects of sentence humor on the recall of nonhumorous sentences reported in Experiment 2. Second, it does not explain why the effects of experimentally defined sentence humor were attenuated when incidental learning tasks were used. Perhaps the most complete explana- tion of these results is a combination of the attention and the differential rehearsal hypotheses. The amount of attention a sentence receives may be a direct function of how humorous the sentence is relative to recent sentences. This process is described by the incongruity hypothesis. With intentional instructions, a second process may be involved. Subjects may devote more time to rehearsing the sentences that received increased attention.
There are important practical implications of the results reported above. Humor can be a significant part of the educational process (Powell & Andresen, 1985). However, humor may enhance memory for the humorous material at the expense of other information contained in a lecture or text- book chapter. Thus, humor must be integrated with the course
content (Kaplan & Pascoe, 1977). In addition, constant use of humor will have little effect on retention because such humor will no longer be incongruous with the background context. Humor should be used to illustrate important concepts and not incidental or peripheral material. Similar advice can be offered to advertisers. Advertisers should embed their humorous commercials in a series of nonhumorous commercials. The product itself should be integrated with the source of the humor. Otherwise, the humorous part of a commercial may suppress memory for the rest of the commercial content.
In summary, humor enhances sentence memory. This effect was found in free and cued recall, in incidental and intentional learning, and on a variety of measures of sentence access. However, the effects of sentence humor on memory were greatly influenced by the context in which the sentence appeared. The results were inconsistent with simple arousal, surprise, and retrieval explanations. The most complete expla- nation was that humorous items command increased attention as described by the incongruity hypothesis. In addition, with intentional learning instructions, humorous sentences were rehearsed at the expense of nonhumorous sentences.
References
Adams, A. K. (1969). The home book of humorous quotations. New York: Dodd, Mead.
Atkinson, R. C , & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation: Advances in research and theory, (Vol. 2, pp. 89-195). New York: Academic Press.
Bohannon, J. N. (1988). Flashbulb memories of the space shuttle disaster: A tale of two theories. Cognition, 29, 179-196.
Braude, J. M. (1964). Braude's treasury of wit and humor. Englewood Cliffs, NJ: Prentice Hall.
Brown, R., & Kulik, J. (1977). Flashbulb memories. Cognition, 5, 73-99.
Christianson, S.-A. (1989). Flashbulb memories: Special, but not so special. Memory & Cognition, 17, 435-443.
Christianson, S.-A. (Ed.). (1992). The handbook of emotion and memory: Research and theory. Hillsdale, NJ: Erlbaum.
Christianson, S.-A., Loftus, E. F., Hoffman, H., & Loftus, G. R. (1991). Eye fixations and memory for emotional events. Journal of Experimen- tal Psychology: Learning, Memory, and Cognition, 17, 693-701.
Continuum. (1988, October). Omni, pp. 52-53. Continuum. (1989, June). Omni, pp. 38-39. Clark, H. H. (1973). The language-as-a-fixed-effect fallacy: A critique
of language statistics in psychological research. Journal of Verbal Learning and Verbal Behavior, 12, 335-359.
Craik, F. I. M., & Blankstein, K. R. (1975). Psychophysiology and human memory. In P. H. Venables & M. J. Christie (Eds.), Research in psychophysiology (pp. 388-417). London: Wiley.
Deckers, L., & Devine, J. (1981). Humor by violating an existing expectancy. The Journal of Psychology, 108, 107-110.
Deckers, L., & Hricik, D. (1984). Orienting and humor responses: A synthesis. Motivation and Emotion, 8, 183-204.
Dixon, P. N., Willingham, W., Strano, D. A., & Chandler, C. K. (1989). Choice of humor as a mediator during incidental learning of humor-related material. Psychological Reports, 3, 851-855.
Droke, M. (1956). The speaker's handbook of humor. New York: Harper.
966 STEPHEN R. SCHMIDT
Duncan, C. P., Nelson, J. E., & Frontzak, N. L. (1984). The effect of humor on advertising comprehension. In T. C. Kinnear (Ed.), Advances in consumer research (pp. 432-437). Chicago: Association for Consumer Research.
Einstein, G. O., & McDaniel, M. A. (1987). Distinctiveness and the mnemonic benefits of bizarre imagery. In M. A. McDaniel & M. Pressley (Eds.), Imagery and related mnemonic processes: Theories, individual differences, and applications (pp. 78—102). New York: Springer-Verlag.
Ellis, N. R., Detterman, D. K., Runcie, D., McCarver, R. B., & Craig, E. (1971). Amnesic effects in short-term memory•. Journal of Experi- mental Psychology, 89, 357-361.
Erlebacher, A. (1977). Design and analysis of experiments contrasting the within- and between-subjects manipulation of the independent variable. Psychological Bulletin, 84, 212-219.
Erlebacher, A. (1978). The analysis of multifactor experiments de- signed to contrast the within- and between-subjects manipulation of the independent variables. Behavior Research, Methods, & Instrumen- tation, 10, 833-S40.
Eysenck, M. W. (1979). Depth, elaboration, and distinctiveness. In L. S. Cermak & F. I. M. Craik (Eds.), Levels of processing in human memory (pp. 89-118). Hillsdale, NJ: Erlbaum.
Gelb, B. D., & Zinkhan, G. M. (1986). Humor and advertising effectiveness after repeated exposures to a radio commercial. Journal of Advertising, 15, 15-20.
Heinlein, R. A. (1973). Time enough for love. New York: Putnam. Heuer, F., & Reisberg, D. (1990). Vivid memories of emotional events: The
accuracy of remembered minutiae. Memory & Cognition, 18, 496-506. Hirshman, E., Whelley, M. M., & Palij, M. (1989). An investigation of
paradoxical memory effects. Journal of Memory and Language, 28, 594-609.
Hunt, R. R., & Elliott, J. M. (1980). The role of nonsemantic information in memory: Orthographic distinctiveness effects on retention. Journal of Experimental Psychology: General, 109, 49-74.
Isen, A. M. (1985). Asymmetry of happiness and sadness in effects on memory in normal college students: Comment on Hasher, Rods, Zacks, Sanft, and Doren. Journal of Experimental Psychology: Gen- eral, 114, 388-391.
Kaplan, R. M., & Pascoe, G. C. (1977). Humorous lectures and humorous examples: Some effects upon comprehension and reten- tion. Journal of Educational Psychology, 69, 61-65.
Kintsch, W., & Bates, E. (1977). Recognition memory for statements from a classroom lecture. Journal of Experimental Psychology: Human Learning and Memory, 3, 150-159.
Kleinsmith, L. J., & Kaplan, S. (1964). Interaction of arousal and recall interval in nonsense syllable paired-associate learning. Journal of Experimental Psychology, 67, 124-126.
Maltzman, I., Kantor, W., & Langdon, B. (1966). Immediate and delayed retention, arousal, and orienting and defensive reflexes. Psychonomic Science, 6, 445-446.
McDaniel, M. A., & Einstein, G. O. (1986). Bizarre imagery as an effective memory aid: The importance of distinctiveness. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 54-65.
McGhee, P. E. (1983). The role of arousal and hemispheric lateraliza- tion in humor. In P. E. McGhee and J. H. Goldstein (Eds.), Handbook of humor research. Vol. I. Basic issues. New York: Springer-Verlag.
McGhee, P. E., & Goldstein, J. H. (1983). Handbook of humor research. Vol. I. Basic issues. New York: Springer-Verlag.
Ohman, A. (1979). The orienting response, attention, and learning: An information-processing perspective. In H. D. Kimmell, E. H. van Olst, & J. E. Orlebeke (Eds.), The orienting reflex in humans (pp. 4 4 3 ^ 7 1 ) . Hillsdale, NJ: Erlbaum.
Pillemer, D. (1984). Flashbulb memories of the assassination attempt on President Reagan. Cognition, 16, 63-80.
Powell, J. P., & Andresen, L. W. (1985). Humor and teaching in higher education. Studies in Higher Education, 10, 79-90.
Prochnow, H. V., & Prochnow, H. V. (1969). A treasury of humorous quotations for speakers, writers, and home reference. New York: Harper & Row.
Roediger, H. L., HI. (1974). Inhibiting effects of recall. Memory & Cognition, 2, 261-269.
Schmidt, S. R. (1991). Can we have a distinctive theory of memory? Memory & Cognition, 19, 523-542.
Schmidt, S. R., & Williams, A. R. (1992, November). Memory for the Far Side. Paper presented at the Annual Meeting of the Psycho- nomic Society, St Louis, MO.
Suls, J. M. (1972). A two-stage model for the appreciation of jokes and cartoons: An information-processing analysis. In J. H. Goldstein and P. E. McGhee (Eds.), The psychology of humor (pp. 81-100). New York: Academic Press.
Walker, E. L., & Tarte, R. D. (1963). Memory storage as a function of arousal and time with homogeneous and heterogeneous lists. Journal of Verbal Learning and Verbal Behavior, 2, 113-119.
Watkins, O. C , & Watkins, M. J. (1975). Buildup of proactive inhibition as a cue-overload effect. Journal of Experimental Psychol- ogy: Human Learning and Memory, 104, 442—452.
Winer, B. J. (1971). Statistical principles in experimental design (2nd ed.). New York: McGraw-Hill.
Zillmann, D., Williams, B. R., Bryant, J., Boynton, K. R., & Wolf, M. A. (1980). Acquisition of information from educational televi- sion programs as a function of differently paced humorous inserts. Journal of Educational Psychology, 72, 170-180.
Ziv, A. (1988). Teaching and learning with humor: Experiment and replication. Journal of Experimental Education, 57, 5-15.
Appendix
Sentences Used in Experiments 1-5
Sentences 1 and 12 were not used in Experiments 3, 4, and 5. Underlined words served as retrieval cues in Experiment 3.
Humorous Sentences
1. A psychiatrist recently reported that half of his patients went to him because they weren't married, while the other half went to him because they were. (Braude, 1964, p. 164)
2. No matter how thin you slice it, it's still baloney. (Alfred E. Smith in Adams, 1969, p. 255)
3. Suburbia is where the developer bulldozed out the trees, then named the streets after them. (Bill Vaughn, cited in "Continuum," Omni, 1989, p. 39)
4. The only way to keep your good health is to eat what you don't want, drink what you don't like, and do what you'd rather not. (Mark Twain, in Adams, 1969, p. 164)
5. An elephant is a mouse built to government specifications. (Heinlein, 1973, p. 244)
6. There are three ways a man can wear his hair: parted, unparted, and departed. (Braude, 1964, p. 17)
HUMOR AND MEMORY 967
7. If at first you don't succeed, you're probably not related to the boss. (Braude, 1964, p. 66)
8. Golf is a lot like taxes—you drive hard to get to the green and then wind up in the hole. (Braude, 1964, p. 85)
9. What is called "congestion" on the subway is called "intimacy" in a night club. (Braude, 1964, p. 149)
10. Nature is amazing; who would have thought of growing a fly swatter on the rear end of a cow? (Braude, 1964, p. 142)
11. The hydrogen bomb has made one great contribution to democ- racy: with it all men are cremated equal. (Braude, 1964, p. 50)
12. Confucius say, "Salesman who cover chair instead of territory always on bottom." (Droke, 1956, p. 224)
13. Two can live as cheap as one, but it costs twice as much. (Frank Sullivan, in Prochnow & Prochnow, 1969, p. 341)
14. It's not true that life is one damn thing after another—it's one damn thing over and over. (Edna St. Vincent Millay, in Adams, 1969, p. 205)
15. I got a bill for my surgery—now I know why those doctors were wearing masks. (James H. Boren, cited in "Continuum," Omni, 1989, p. 38)
16. The easiest way to teach children the value of money is to borrow some from them. (Braude, 1964, p. 146)
17. Some people are like blotters—they quickly soak it all in, but get it all backwards. (Droke, 1956, p. 442)
18. People are like tea bags . . . they don't know their own strength until they get into hot water. (Braude, 1964, p. 215)
19. A salesman who isn't fired with enthusiasm should be fired—with enthusiasm. (Droke, 1956, p. 218)
20. A great many open minds should be closed for repairs. (Toledo Blade, cited in "Continuum," Omni, 1988, p. 52)
Nonhumorous Sentences
1. A psychiatrist recently reported that half of his patients were not married, while the other half were married.
2. No matter how you look at it, it's still baloney. 3. In suburbia the developer bulldozes out the trees, and the names
of the streets come from trees.
4. The only way to keep your good health is to eat good food, drink healthy drinks, and do healthy activities.
5. The government often builds an elephant when it only needs a mouse.
6. Men can wear their hair with or without a part, unless they are bald.
7. People who are related to the boss often succeed the very first time.
8. Golf and taxes are a lot alike—in golf you try to get to the green and in taxes you try to save money.
9. People are just as close together on a congested subway as in an intimate night club.
10. Nature is amazing in that it supplies a cow with a tail with which to swat flies.
11. The hydrogen bomb is democratic in that everyone is equally likely to be cremated.
12. Confucius say, "Salesman who spend a lot of time in their chairs are not very hard workers."
13. Unlike the common saying, two people cannot live as cheaply as one.
14. It's not true that life is one damn thing after another—it's full of the same problems again and again.
15. I got a bill for my surgery—those doctors were like robbers with the prices they charged.
16. Borrowing money from children is an easy way to teach them the value of money.
17. Some people can soak up material as quickly as a blotter—but then get it all backwards.
18. People don't know how strong they are until they get into serious trouble.
19. A salesman who isn't fired with enthusiasm should be quickly terminated.
20. A great many open minds do not function properly, and thus should be fixed.
Received March 4,1992 Revision received August 9,1993
Accepted August 17,1993 •
