Escolar Documentos
Profissional Documentos
Cultura Documentos
This paper presents results of a meta-analysis performed using published results of experimental
studies on the effects of humor in advertising. Three categories of response variables were considered
analyses were performed for each response variable to estimate the combined effect sizes of humor in
advertising and to assess the moderating effect of methodological characteristics of the different
studies. The results indicate that cognitive measures are not affected by humor in ads, while attitude
toward the ad, attitude toward the brand, and purchase intention are higher in the presence of humor.
In addition, recall of humorous ads is stronger when humor is measured during the experiment rather
than manipulated and when the sample is exclusively male. Attitude toward the ad is also higher in
1
Despite the famous phrase by Hopkins (1923) that “People don’t buy from clowns”, humor in
advertising is a popular copy strategy and increasingly so. Indeed, 42% of television commercials in
the United States were found to be funny in 1972 (Markiewicz 1974) and 50% in 1996 (Speck and
Elliott 1997). A similar situation was found in Great Britain, where approximately 35% of ads were
considered humorous in 1985 (Weinberger and Spotts 1989), while Hanna, Gordon, and Ridnour
(1994) found that 28% of Japanese commercial had some humorous content. In addition, humorous
ads are often recipients of advertising prizes, such as the Lions in Cannes, the Clio Awards, IPA’s
Research dealing with the effectiveness of humor in advertising has been a steady stream since the
1970s, while research on humor in communication in general, has been performed since a much earlier
time. This research stream has been reviewed by a number of authors (Weinberger and Gulas 1992;
Graby 2001) with an attempt to generalize on some of the experimental results. Unfortunately,
research results vary widely and as such, these generalizations are close to impossible. In many
instances, conclusions cannot be drawn because experimental conditions vary between studies, for
instance gender of the sample, type of product tested, media used, etc. As an alternative to a classical
literature review, this article reports results of a meta-analysis of experimental studies comparing the
The term meta-analysis was defined by Glass (1976, p.3) as “the statistical analysis of a large
collection of analysis results from individual studies for the purpose of integrating the findings”. A
meta-analysis takes into consideration the theoretical foundations of the phenomenon being studied.
Compared to literature reviews, meta-analyses are more systematic, more explicit, more exhaustive,
and more objective. Conventional meta-analysis techniques consist essentially of averaging the
outcomes of available studies, using appropriate weighting to account for differences in sample size. If
considered separately, any one study may be either too small or too limited in scope to come to
unequivocal or generalizable conclusions about the effect of a treatment. Combining the findings
across studies with similar treatment represents an attractive alternative to strengthen the evidence
2
about the treatment effectiveness and thus to increase the power to detect this effect. However, the
main difficulty in integrating the results from various studies stems from the sometimes diverse nature
of the studies, both in terms of designs and methods used. Some are carefully controlled randomized
experiments while others are less well controlled or simply observational studies. Generally, there are
also differences in the sample sizes and more particularly in population characteristics leading to
heterogeneity in the outcomes across studies. Therefore, the use of regression models has been
proposed to deal with this problem (Cooper and Hedges 1994). Thus, when heterogeneity is present,
one objective of a research synthesis is to try to explain the variability across the studies by modelling
the treatment effect as a function of a set of factors that moderate the effectiveness.
Meta-analyses have been used in various scientific fields (e.g. medicine, physics, etc.), but also in
marketing, in areas where a substantive body of research has been compiled. For example, Assmus,
Farley, and Lehmann (1984) performed a meta-analysis on the effectiveness of advertising on sales;
Rao and Monroe (1989) were interested in the meta-results of price effects on buyers’ perceptions;
Sultan, Farley, and Lehmann (1990) used the results of studies on the diffusion of innovations; Brown
and Stayman (1992) looked at attitude toward the ad; Peterson and Jolibert (1995) performed a meta-
analysis on the country-of-origin effects; and Abernethy and Frank (1996) were interested in
advertising information content. Since (??) many experimental studies dealing with the effectiveness
of humor in advertising have been published, it seemed quite appropriate to apply this technique.
In this paper, we attempted to identify whether humor in advertising was effective by considering the
classical advertising response variables, i.e. cognitive (recall and cognitive responses), affective
(attitude toward the ad and attitude toward the brand), and behavioral (purchase intention) variables. In
addition, contingency variables related to the methodological environment were assessed for their
moderating effect in a statistical regression model. The meta-analysis enabled us to confirm some of
3
The first section of the paper explains the methodology used, including a description of the studies that
were considered for the meta-analysis. The next section presents the results of the analysis; and we
Methodology
Our research methodology follows the steps suggested by Hedges and Olkin (1985) and consists in
defining the field of investigation; establishing objective criteria to include studies in the meta-
analysis; searching and selecting the relevant studies; extracting and coding the results and
characteristics of each study; determining the effect size of humor for each advertising response
Research Field
The research field investigated in this paper is the effectiveness of humor in advertising. Effectiveness
has been defined as cognitive, affective, and behavioral responses to humorous ads. A brief
Cognitive Measures
Three cognitive dependent variables were considered: recall, cognitive responses toward the ad, and
cognitive responses toward the brand. These variables are defined as follows:
Recall: any measure of aided or unaided recall of the brand name, product features, uses, and benefits,
advertising slogan. When both aided and unaided recall were reported in a study, only unaided recall
Cognitive responses toward the ad: any measure of thought-listed responses to the ad. When the
study reported separate results for positive and negative thoughts, these were summed to form a total
score.
Cognitive responses toward the brand: any measure of thought-listed responses to the brand. Again,
when both positive and negative responses were reported, these were summed to form a total score.
4
Affective Measures
Two affective responses were considered: attitude toward the ad (Aad) and attitude toward the brand
(Ab).
Attitude toward the ad: unidimensional measure of attitude toward the ad. When Aad was measured
with multiple items and was not averaged in the study, we used the item most closely linked to the
Attitude toward the brand: unidimensional measure of attitude toward the brand. The same
procedure as in Aad was used for Ab to determine the unique variable to consider.
Behavioral Measure
Inclusion Criteria
The second step in our methodology involved determining on what basis studies should be included in
the meta-analysis. Studies were included if and only if they fulfilled two criteria: (1) they were
(2) they contained the necessary information on at least one of the six response variables mentioned
The search process involved the following steps. First the PROQUEST computer database was
searched on October 19, 2003 for potentially relevant articles published, using simultaneously the key
words “humo*” and “ad*” in the article title. Second, relevant articles selected on the basis of the title
and abstract were retrieved for more detailed evaluation. Third, the bibliographies of relevant articles
were searched for additional references. Fourth, some researchers in the field were contacted by e-mail
for other potentially relevant articles or unpublished reports. Fifth, proceedings from AAA, AMA, and
5
Extraction of the Information from the Selected Studies
Extraction of the relevant information for the meta-analysis was done by two of the authors
following data, if available, was abstracted to compute the effect size for each response variable
considered: sample sizes, mean scores and standard deviations for both the humorous and non-
humorous advertisement groups, and the t-statistic for the statistical comparison of the two groups.
When authors failed to report the size of the humorous and non-humorous samples, but the total
sample size was given, it was assumed that both groups contained the same number of respondents.
In order to identify possible causes for variations in the experimental results across studies, other
variables have been extracted. . They pertain to the methodological characteristics of the studies:
Gender of respondents: this variable has been explicitly tested in various studies and indicated that
males had on average more favorable responses to humorous ads. Consequently, the data extracted for
the meta-analysis included a categorical variable to indicate whether the sample considered in the
randomized experiment was exclusively male, female or mixed. If an article did not report the
composition of the sample, it was assumed to be of both genders. We expected a stronger response for
Measure of humor: humor has been either manipulated or measured in the experiments. Manipulated
humor refers to instances where the authors (or experts) determine a priori which ads are to be
considered as humorous for the experiment. Measured – or perceived – humor indicates that subjects
of the experiment evaluated the degree of humor of the ads, so that the humorous and non-humorous
ads are determined a posteriori. Measured humor seems to be a more effective experimental condition.
Brown and Stayman (1992) also considered a manipulated vs. measured variable (for feelings) and
found significant differences in their meta-analysis, with the measured condition yielding higher
results. When measured humor was considered in the studies, the condition for low degree of humor
was assimilated to absence of humor, while a high condition was considered to be the humorous
version. Measured humor was expected to have a larger effect than manipulated humor.
6
Type of product or service examined: the type of product tested was considered as a possible reason
for diverging results across studies. Indeed, both durable and frequently purchased products were
tested in the studies. In addition, as product type determines information processing of individuals, this
tested for product type and found significant differences for this variable (Assmus, Farley, and
Lehmann 1984; Sultan, Farley, and Lehmann 1990; Brown and Stayman 1992; Peterson and Jolibert
1995; Abernethy and Frank 1996). A dichotomous variable indicating the product type was thus
included in the regression model. Based on generally accepted premises, we expected frequently
purchased goods to be more sensitive to humor, as they tend to generate less information processing.
Type of media: since four types of media were present in the sample of studies (radio, TV, magazine,
and flyer), we included this variable in the meta-analysis. Brown and Stayman (1992) considered
media type in their meta-analysis on attitude toward the ad and found some significant differences. A
similar result was found by Abernethy and Frank (1996) in their meta-analysis of advertising
information content. We compared results of studies from electronic broadcasting (television and
radio) with print advertisements (magazine and flyer). No direction of results could be set, as no study
Relevance of humor: while it would have been interesting to test for the type of humor used in the
ads, such information was generally not available. Alternatively, following one of Speck’s typology
(1991), we considered the dichotomy of whether the humor used in the ad had any relevance to the
product/situation or not. No specific direction of results was expected as previous studies show
diverging results for the effect of relevance of humor in ads (Muehling and McCann 1993; Lee and
Mason 1999).
Imbeddedness: some experiments explicitly imbedded the ad in a media, while others used a stand
alone ad. Results of previous studies seem to indicate that imbedded ads are less effective, as
respondents are less attentive to the ad (Brown and Stayman 1992). A dichotomous variable
Type of sample: as Brown and Stayman (1992) reported, subject type is often coded in meta-analyses
as experiments using student samples often result in stronger effects. This variable could, however, not
7
be coded in this analysis as all studies meeting our inclusion criteria report results from student
samples.
If a study evaluated and reported results for different levels of one or more of the independent
variables (e.g., male and female; relevance and irrelevance of humor), the information was extracted
The effect sizes were computed as the difference between the humorous and non-humorous mean
scores of the response variables divided by the pooled standard deviation and adjusted by a
multiplicative factor proposed by Hedges and Olkin (1985) to correct for bias in the effect size
estimates. If standard deviations were not reported, we used the t-statistic to calculate the pooled
We first obtained the combined effect size across studies with its 95% confidence interval (C.I.), for
each of the six response variables, using a random effect model (see Cooper and Hedges 1994, pp.
309-311 for details). If two or more comparisons within a study were derived with the same subjects
for the same response variable, we used only one comparison from the study for the quantitative meta-
analysis, selecting the first comparison done. Therefore, the overall sample size of the combined effect
size is the sum of the sample sizes of each study included in the meta-analysis. We also tested the
hypothesis of homogeneity of the effect sizes across studies (Cooper and Hedges 1994). When this
hypothesis was rejected, we evaluated the moderating effect of the characteristics of the different
studies and subgroups within studies, on the response variables. This was done by using a regression
model. The parameter estimates of the model were computed using the method of moments with the
weighted least squares approach (Cooper and Hedges 1994, p. 311). Because of the small number of
studies included in the meta-analysis for each response variable, we considered separate simple
regression models, one for each of the methodological characteristics extracted from the studies.
8
The statistical analyses were conducted using SAS statistical software for Windows, release 8.02.
Results
The search strategy in PROQUEST yielded 153 published articles. Based on the title and abstract, 34
were retrieved for more detailed evaluation. Seven of these studies, and five others found from the
bibliographies of the retrieved studies, met the inclusion criteria, meaning that they dealt with
measuring the effectiveness of humor in advertising in experimental conditions and that they
contained the necessary information to perform the meta-analysis. Obviously, a larger number of
studies were found dealing with humor in advertising, but they do not explicitly compare the
effectiveness of a humorous ad with a serious one and could therefore not be used in this study.
Cantor and Venus (1980) manipulated a radio commercial into a humorous and a non-humorous
version for an existing, but unknown in the region, magazine. The commercial was imbedded in
fifteen minutes of actual radio broadcasting. A student sample consisting of 59 males and 58 females
was randomly exposed to one of either version of the commercial. The authors measured memorability
and persuasiveness of the ad. Recall of the ad was found to be higher for the serious version, but no
Lammers et al. (1983) used a serious and humorous version of a radio commercial for industrial sound
sheets. The experiment did not try to imbed the commercial in any actual radio programming. A
student sample of 64 (26 females and 38 males) participated in the study and was randomly exposed to
either the serious or humorous version of the commercial. Measures were taken for cognitive
responses, attitude toward the ad, and recall of brand name, product class, benefits and uses. Results
indicated a gender effect for cognitive responses and attitude toward the ad, with females having
higher couterarguments and a lower attitude toward the ad for the humorous version of the ad. Results
for the entire sample showed that cognitive responses did not vary between the two versions of the
9
commercial; recall of product uses and benefits was lower for the humorous version of the ad; and
attitude toward the ad did not vary between the serious and the humorous ad.
Sutherland and Middleton (1983) used serious and humorous versions of print advertisements for two
product classes, American Tourister suitcases and Fuji cameras. Their study manipulated the type of
humor used, namely nonsensical humor for the American Tourister ad and pun on words for the Fuji
ad. A sample consisting of 107 students was exposed to a serious version of one brand and the
humorous version of the other brand, thereby allowing a between-subject measure of type of humor
and within-subject measure of humor vs. serious. Measures were taken for recall and credibility.
Credibility was measured with a nine-item scale which closely represents the evaluation of the
advertisement (or Aad). Recall was not found to be significantly different for the humorous version
compared to the serious one. Credibility, however, was higher for the serious version in the case of the
American Tourister ad, but higher for the humorous version of the Fuji ad.
Duncan, Nelson, and Frontczak (1984) designed an experiment with a radio commercial for a fictitious
brand of men’s hair care product. Three humorous versions were produced, differing with respect to
where in the message humor appeared. A serious version was used as control ad. A student sample of
149 was randomly exposed to one of the four versions of the commercial, which was embedded in
actual radio programming. Measures were taken for recall and perceived humor. Contrary to previous
studies, data were analyzed between subjects that perceived humor in the commercial with those that
were exposed to the serious version. Findings indicated that recall was higher when subjects perceived
Duncan and Nelson (1985) provide further details on the study described above. The data analysis was
performed for measures of attention paid to the commercial, selling points recalled, positive product
beliefs, attitude toward the ad, attitude toward the product, purchase intention, distraction experienced,
and irritation experienced. Subjects exposed to the humorous version of the commercial were
separated into three groups according to their perception of humor. A MANOVA showed that mean
10
responses differed among the groups. A discriminant analysis indicated that humor in the commercial
enhanced attention paid to the commercial, attitude toward the ad, attitude toward the product, and
decreased irritation.
Sutherland and Sethu (1987) set up an experiment using humorous versions of two existing brands,
namely Parkay margarine and Epeda mattresses, which correspond to a familiar and an unfamiliar
product, respectively. In addition, the authors constructed a non-humorous version of the commercials
based on the real ones. A total of 166 students participated in the study, where both recall and
credibility of the message were measured. None of the ANOVAs provided significant results however.
Wu, Crocker, and Rogers (1989) tested main and interaction effects of humor, comparative strategy,
and product type in an experiment. A total of eight print advertisements were created, four for facial
tissue and four for athletic shoes. A sample of 360 business students was randomly assigned to one of
the eight advertisements. Four constructs were measured: recall, attitude toward the ad, attitude toward
the brand, and purchase intention. Results showed that only attitude toward the ad was significantly
higher for the humorous version of the ad, regardless of product type.
Chattopadhyay and Basu (1990) used a serious and a humorous version of a television commercial for
a fictitious brand of pens. Eighty students participated in the experiment and were exposed to a 15-
minute television broadcast in which the test ad was imbedded twice. Subjects were randomly
assigned to the humorous or serious version of the ad. Measures were taken for cognitive responses,
attitude toward the ad, attitude toward the brand, brand choice, and purchase intention. Significant
positive effects of humor were found for cognitive responses and attitude toward the ad.
Smith (1993) produced a humorous and a serious version of a print ad for a fictitious brand of life
insurances. Sixty-seven students participated in the experiment and were randomly exposed to either
version of the ad. Measures were taken for perceived humor, cognitive responses, attitude toward the
ad, attitude toward the brand, and purchase intention. Because the humor manipulation in the ad was
11
not successful, the author used perceived humor in the ad to separate the sample into the humorous
and non-humorous version of the ad. Results indicated that the humorous version had a significant
positive effect on attitude toward the ad, attitude toward the brand, and purchase intention.
Zhang (1996) used a serious and a humorous version of a print ad for a fictitious brand of cameras.
Two-hundred and forty students participated in the study and were randomly assigned to either the
serious or humorous version of the ad, imbedded in a magazine-type booklet. Measures were taken for
attitude toward the ad, attitude toward the brand, and purchase intention. Using MANOVA, the author
found a significant effect of humor for attitude toward the ad, attitude toward the brand, and purchase
intention.
Cline and Kellaris (1999) examined the interaction of humor with strength of arguments on attitude
toward the ad and attitude toward the brand. A sample of 122 students was asked to evaluate a
humorous and a non-humorous print advertisement for a fictitious brand of bubble gum. These ads
were shown inserted in the questionnaire booklet. Results indicated no main effects of humor on
Lee and Mason (1999) produced a serious and a humorous version of a print ad for fictitious brands of
computers and computer monitors. A sample consisting of 120 students was exposed to experimental
versions of both the computer and the monitor. Measures were taken for cognitive thoughts, attitude
toward the ad, and attitude toward the brand. Results indicated that humor did not affect the amount of
positive thoughts, but reduced the amount of negative thoughts. In addition, attitude toward the ad and
attitude toward the brand were higher for the humorous ad.
As mentioned above, three types of response variables were considered for the meta-analysis:
cognitive, affective and behavioral responses. Results of the data analysis are presented for each type
of variable.
12
Cognitive Measures
Recall
Table 1 presents the characteristics and the computed effect size of each study (or of each subgroup
within a study) that compared recall of advertisements for humorous and non humorous ads. Recall
was lower for humorous ads in five studies or subgroups within studies (two of them were statistically
significant at the 5% level) and higher in five others (one was statistically significant at the 5% level).
Overall, the estimated combined effect size for recall is positive at 0.04 but not statistically different
from zero (the 95% C.I. for the combined effect size includes 0: -0.34 to 0.44).
Table 1: Recall
The hypothesis of homogeneity of the effect size across studies, for humorous advertisement on recall,
is rejected (p < 0.001). The separate regression models indicate that two of the study characteristics
considered were statistically significant in explaining some of the heterogeneity observed in the
results. More specifically, humorous ads tend to produce a higher recall when humor is measured
rather than manipulated (p=0.002) or when the sample is exclusively male (p=0.034). Note however
that these two independent variables are almost totally confounded, i.e. all studies with measured
Table 2 presents the results when cognitive responses toward the ad are considered as the response
variable. Three effect sizes were negative, that is the cognitive response was lower for humorous ads,
13
and seven were positive. However, all effect sizes were small and none was statistically different from
zero. Therefore, the results were fairly homogeneous across studies (p=0.711 for test of homogeneity)
and the overall estimated effect size for the cognitive response was not statistically different from zero
(95% C.I. for the combined effect size is -0.14 to 0.27). None of the study characteristics was
statistically related to the effect size of humorous ads for the cognitive responses toward the ad. It
should be noted that gender has not been tested in the regression model as there was no variation in the
studies.
Cognitive responses toward the brand have also been studied by a number of authors and have
Four effect sizes are negative and two are positive, but they are all small and none of them are
statistically significant because all the 95% confidence intervals include zero. Hence, the hypothesis of
homogeneity of the effect of humorous ads for cognitive responses toward the brand across studies is
14
not rejected (p=0.530), the combined effect size is not statistically significant and none of the study
characteristics tested in the regression models was statistically related to the effect size.
Affective Measures
Table 4 presents the results of the meta-analysis of attitude toward the ad. Attitude toward the ad was
higher for humorous ads in 14 studies or subgroups within studies (seven of them were statistically
significant at the 5% level) and lower in six others (none of the negative effect sizes was statistically
significant at the 5% level). Overall, the estimated combined effect size for attitude toward the ad was
positive at 0.44 and statistically different from zero (95% C.I. for the combined effect size is 0.16 to
0.73). Hence, humorous ads were more favorably received by the audience compared to non humorous
ads.
The hypothesis of homogeneity of the effect size across studies was rejected (p < 0.001). The separate
regression analyses showed that the measure of humor is significantly related to the effect size
(p=0.034); more specifically, the effects on attitude toward the ad are more important when measured
humor is considered in the studies rather than manipulated humor. Furthermore, the results show that
15
considering whether the ad is imbedded in a media program, also explains variances in results
Attitude toward the brand was higher for humorous ads in 12 studies or subgroups within studies
(seven of them were statistically significant at the 5% level) and lower in four others (one of the
negative effect sizes was statistically significant at the 5% level). Again, the estimated combined effect
size for attitude toward the brand was positive at 0.39 and statistically different from zero (95% C.I.
for the combined effect size is 0.09 to 0.71), indicating that attitude toward the brand for humorous ads
The hypothesis of homogeneity of the effect size was rejected (p < 0.001) and the separate regression
analyses showed that product type was significant at the 10% level (p=0.066). In other words, Ab
tended to be higher for humorous ads when the product was a durable good. No other study
characteristics considered had a statistically significant impact on the variations of results across
studies.
16
Purchase Intention
Table 6 presents the results of the meta-analysis on purchase intention. Two effect sizes were negative,
but not statistically different from zero, and six were positive with four being statistically significant,
indicating significantly higher purchase intention for humorous ads. Overall, the estimated combined
effect size for purchase intention was positive at 0.47 and statistically different from zero (95% C.I.
The hypothesis of homogeneity of the effect size was rejected (p < 0.001). All studies on purchase
intention used imbedded ads; this variable has therefore been excluded from the separate regression
analyses. While the effect sizes varied significantly, none of the study characteristics considered could
explain the heterogeneity in the results; other variables should thus be considered to explain variations
With the exception of attitudes and purchase intention, our results tend to disconfirm what has been
reported in previous attempts to generalize on the effects of humor in advertising. These similarities
We found that recall was not significantly higher for humorous ads than for non humorous ads,
thereby providing a clearer conclusion on the mixed results found in literature reviews (e.g.
17
Weinberger and Gulas 1992, p.36). We can furthermore explain some variation of the heterogeneity in
results, as we found that studies in which humor was measured as opposed to those that manipulated
humor had a higher effect on recall. While gender (i.e. male samples) seems to have a positive effect
on recall of humorous ads, this variable was however confounded with the measure of humor in our
meta-analysis. Further original experimental studies are needed to elucidate the moderating effect of
Similarly, comprehension lead to conflicting results when surveying the literature. The meta-analysis
clearly showed however an absence of effect on cognitive responses either toward the ad or the brand,
in other words, humorous ads do not tend to produce more cognitive responses than non-humorous
ads.
The meta-analysis further indicated that affective responses, attitude toward the ad and attitude toward
the brand, were significantly higher for humorous ads compared to non humorous ads. This result is
not surprising as a humorous ad tends to be viewed as distractive, original, leading to a more favorable
attitude toward the ad. It also confirms conclusions published in literature reviews. In addition, there is
some evidence that experiments using measured humor tend to produce stronger results, but this
should be taken with caution as only a limited number of studies have used measured humor rather
than manipulated humor. Similarly, when the ad was imbedded in a media, responses were stronger
As for attitude toward the brand, we showed that durable goods tend to produce stronger effects than
frequently purchased goods, which is contrary to what was expected. This counterintuitive result could
be explained by the fact that humor tends to be less expected for durable goods and may therefore
18
Finally, purchase intention was found to be significantly higher for humorous ads, however the limited
number of studies measuring this response prevented us from identifying study characteristics that
While most of our results confirm what has been previously published, our conclusions have a
statistical validity beyond the literature review. Also, we were able to point to some study
characteristics that account for heterogeneity in results, something which previous authors have only
Nevertheless, our study is not without limitations. Firstly, we should point to the fact that only a very
limited number of studies met our inclusion criteria, i.e., a randomized experiment and publication of
the relevant information to compute the effect size on the response variables. The small number of
studies, however, has a limited impact on the level of significance of the aggregated results; indeed, a
meta-analysis takes into consideration the sample sizes of the individual studies and combined them to
obtain a more powerful estimate of the effect. This is easily seen by noting that the width of the 95%
confidence intervals of the combined effect sizes in Tables 1 to 6 is smaller than the with of the 95%
C.I. for the individual studies. But, the small number of experimental studies included in our meta-
analysis had a more important impact on our search for moderating factors that could explain the
Secondly, we had to make some difficult choices with respect to the measure of attitude (toward the ad
and toward the brand), as this variable has often been considered as a multidimensional variable. Only
“true” liking was considered as attitude, thereby ignoring other variables which are closely related to
the concept.
Thirdly, our study suffers from limitations inherent to any meta-analysis (or literature review, for that
matter). In particular, there is probably a publication bias in so far as mainly studies reporting some
19
significant results are generally published and available, but also that some studies suffer from
methodological problems which we did not account for in the data analysis.
Clearly, this study did not consider all relevant moderating effects, in particular source credibility – a
factor which has often been considered as a determinant of advertising effectiveness – or type of
humor, because of the small number of studies and lack of objective information on these study
characteristics. These limitations should however be considered as opportunities for further research in
References
Abernethy Avery M., Frank George R. The information content of advertising: a meta-analysis.
Journal of Advertising 1996; 25 (2): 1-17.
Assmus Gert, Farley John U., Lehmann, Donald R. How advertising affect sales: meta-analysis of
econometric results. Journal of Marketing Research 1984; 21 (February): 65-74.
Brown Steven P., Stayman Douglas M. Antecedents and consequences of attitude toward the ad: a
meta-analysis. Journal of Consumer Research 1992; 19 (June): 34-51.
Cantor Joanne, Venus Pat. The effect of humor on recall of a radio advertisement. Journal of
Broadcasting 1980; 24 (1): 13-22.
Chattopadhyay Amitava, Basu Kunal. Humor in advertising: the moderating role of prior brand
evaluation. Journal of Marketing Research 1990; 27 (November): 466-476.
Cline Thomas W., Kellaris James J. The joint impact of humor and argument strength in a print
advertising context: a case for weaker arguments. Psychology and Marketing 1999; 16 (1): 69-
86.
Cooper Harris, Hedges Larry V. The Handbook of Research Synthesis. New York, Russell Sage
Foundation, 1994.
Duncan Calvin P., Nelson James E. Effects of humor in a radio advertising experiment. Journal of
Advertising 1985; 14 (2): 33-40, 64.
Duncan Calvin P., Nelson James E., Frontczak Nancy T. The effect of humor on advertising
comprehension. Advances in Consumer Research 1984; 11: 432-437.
Glass Gene V. Primary, secondary, and meta-analysis of research. Educational Researcher 1976; 5
(November): 3-8.
Graby Françoise. Humour et comique en publicité. Paris: Editions EMS, 2001.
Hanna Nessim, Gordon Geoffrey L., Ridnour Rick. The use of humor in Japanese advertising. Journal
of International Consumer Marketing 1994; 7 (1): 85-106.
Hedges Larry V., Olkin Ingram. Statistical Methods for Meta-Analysis. New York: Academic Press,
1985.
Hopkins Claude. Scientific Advertising and My Life in Advertising. (1923) reprinted by Advertising
Publications Inc, Chicago, 1966.
Lammers H. Bruce, Leibowitz Laura, Seymour George E., Hennessey Judith E. Humor and cognitive
responses to advertising stimuli: a trace consolidation approach. Journal of Business Research
1983; 11 (2): 173-185.
Lee Yih Hwai, Mason Charlotte. Responses to information incongruency in advertising: the role of
expectancy, relevancy, and humor. Journal of Consumer Research 1999; 26 (September): 156-
169.
Markiewicz Dorothy. Effects of humor on persuasion. Sociometry 1974; 37 (3): 407-422.
20
Muehling Darrell D., McCann Michelle. Attitude toward the ad: a review. Journal of Current Issues &
Research in Advertising 1993; 15 (2, Fall): 25-58.
Peterson Robert A., Jolibert Alain J.P. A meta-analysis of country-of-origin effects. Journal of
International Business Studies 1995; 26 (4): 883-900.
Rao Akshay R., Monroe Kent B. The effect of price, brand name, and store name on buyers'
perceptions of product quality: an integrative review. Journal of Marketing Research 1989; 26
(August): 351-357.
Smith Stephen M. Does humor in advertising enhance systematic processing? Advances in Consumer
Research 1993; 20: 155-158.
Speck Paul S. The humorous message taxonomy: a framework for the study of humorous ads. Current
Issues and Research in Advertising 1991; 13: 1-44.
Speck Paul S., Elliott Michael T. Predictors of advertising avoidance in print and broadcasting media.
Journal of Advertising 1997; 26 (3): 61-76.
Sultan Fareena, Farley John U., Lehmann Donald R. A meta-analysis of applications of diffusion
models. Journal of Marketing Research 1990; 27 (February): 70-77.
Sutherland John C., Middleton Lisa A. The effect of humor on advertising credibility and recall.
Proceedings of the Convention of the American Academy of Advertising 1983; 17-21.
Sutherland John C., Sethu Sudha. The effect of humor on television advertising credibility and recall",
Proceedings of the Convention of the American Academy of Advertising 1987; R3-R8.
Weinberger Marc G., Gulas Charles S. The impact of humor in advertising: a review. Journal of
Advertising 1992; 21(4): 35-59.
Weinberger Marc G, Spotts Harlan E.Humor in US versus U.K. TV advertising. Journal of
Advertising 1989; 18 (2): 39-44.
Wu Bob T.W., Crocker Kenneth E., Rogers Martha. Humor and comparatives in ads for high- and
low-involvement products. Journalism Quarterly 1989; 66 (Autumn): 653-661, 780.
Zhang Yong. Responses to humorous advertising: the moderating effect of need for cognition. Journal
of Advertising 1996; 25 (1): 15-32.
21