The Irritation Paradox
The Relationship between Irritating Advertisements and Advertising Effectiveness
Wiebe Heeres
Master Thesis: Marketing Management
The Irritation Paradox
The Relationship between Irritating Advertisements and Advertising Effectiveness
June 22, 2015 Wiebe Heeres Noorderstationsstraat 16 9717 KN Groningen wiebeheeres@gmail.com Phone: +31 (0) 649819718 Student no.: 2401320 University of Groningen Faculty of Economics and Business
MANAGEMENT SUMMARY
PREFACE
Five months ago I started writing this thesis with the aim of graduating from the University of Groningen. I feel grateful to have been able to combine this feat with studying a topic close to my own interests: the effectiveness of irritating advertising. This paper as such fittingly concludes an educational journey which always had advertising as a leitmotif, initiating with a Bachelor’s degree in Business Economics and culminating in a Master’s degree in Marketing Management.
Upon reading this thesis I hope that you, the reader, get a clear image of the difference in effectiveness of irritating advertising compared to the nowadays omnipresent humorous forms of advertising. Furthermore, I hope that I have been able to explain the underlying mechanism in an unambiguous way, thereby showing its relevance in the current advertising landscape.
TABLE OF CONTENTS
1.
INTRODUCTION ... 1
2.
THEORETICAL FRAMEWORK ... 3
2.1. AFFECTIVE RESPONSES TO ADVERTISING ... 4
2.2. HUMOR VS. IRRITATION ... 5
2.3. THE NEGATIVITY BIAS ... 7
2.4. LITERATURE SUMMARY AND HYPOTHESES ... 8
3.
METHODOLOGY ... 10
3.1. DATA COLLECTION ... 10 3.2. DESIGN ... 11 3.2.1. MANIPULATION CHECK ... 11 3.2.2. PRETEST ... 12 3.3. PROCEDURE ... 13 3.4. MEASURES ... 14 3.4.1. ADVERTISING EFFECTIVENESS ... 14 3.4.2. INFORMATION-PROCESSING DEPTH ... 154.
RESULTS ... 15
4.1. SAMPLE ... 164.2. MANIPULATION AND COVER STORY CHECK... 16
4.3. DEMOGRAPHICS ... 17
4.4. MAIN DESCRIPTIVE STATISTICS ... 19
4.5. MAIN HYPOTHESIS ... 20
4.6. INFORMATION-PROCESSING DEPTH ... 20
4.7. MEDIATION ... 21
5.
DISCUSSION ... 24
5.1. DISCUSSION OF RESULTS ... 24
5.2. MANAGERIAL IMPLICATIONS ... 27
5.3. LIMITATIONS AND FUTURE RESEARCH ... 28
BIBLIOGRAPHY ... 31
APPENDIX ... 35
1
1. INTRODUCTION
On average, a quarter of each hour in present-day US television programming is dedicated to advertising (Nielsen 2014). In almost half of all those commercials some form of humor is used (Millward Brown 2013). The simple explanation for this observation is that humor is a proven way to attract attention and increase people’s involvement, which in turn is a driver of memorability. Besides, there is the widely accepted belief that “the more they like the ad, the more they like the brand” (Aaker and Bruzzone 1985, p. 47). Humor can be defined using stimulus properties or elicited responses, but first and foremost “in terms of the perceptual response of the audience” (Sternthal and Craig 1973, p. 13). This perceptual response can be regarded as how someone judges or evaluates a stimulus, based on “the psychological state characterized by the appraisal that something is funny, the positive emotion of amusement, and the tendency to laugh” (McGraw et al. 2012, p. 1216). The perceiving audience thus plays a major role in the decision on whether something is funny or not. As a result, humor remains tricky as an advertising tool, since it isn’t guaranteed to work as intended by its sender. For one, when employing mass media like television it is possible for a message to be perceived differently by audiences of different cultures, nationalities or even gender (Alden, Hoyer and Lee 1993; Lee and Lim 2008). Accordingly, an advertisement might not come across as being funny with members of the target audience. Moreover, failed humor can lead to irritation, for example through a joke being over the top or lacking in originality, which is said to be of negative influence on brand attitudes and purchase intentions (Bartos 1981; Duncan and Nelson 1985). Advertisement irritation can be seen as a form of intrusiveness (Greyser 1973) and is here conceptualized as consumers’ perceptions that an ad “is provoking, causing displeasure and momentary impatience” (Aaker and Bruzzone 1985, p. 48). Research has suggested multiple factors that induce it, ranging from unbelievable or confusing situations, exaggerations, unattractive characters, being loud, too extensive or too frequently repeated (Aaker and Bruzzone 1985; Fennis and Bakker 2001).
2 subsequent company success raises the question whether in some cases it might actually pay off to intentionally opt for irritating advertisements. Of course it could also be the case that in the previous example a commercial style that was perceived as being funny actually would have been even more effective, but there seem to be other forces at play. A Dutch TV show, Radar, yearly has over a 100.000 viewers voting for the award of most irritating TV commercial. A dubious honor that was reserved for Zalando even twice, in both 2011 and 2012. Interviews with the initiators of these and other winning campaigns unveil that their ad was rather intended to be funny as opposed to irritating (RadarTV 2015).
3 The relationship between attitudes toward advertising and advertising effectiveness has been studied extensively, but the effect of perceived irritation on advertising effectiveness in particular less so. Moreover, most research on the subject originates from the eighties and early nineties. The current study intends to add to extant research by empirically testing the different effects of humor and irritation in advertising on advertising effectiveness in a modern day environment. Advertising effectiveness is here conceptualized in terms of how well advertising claims are recalled. An extra dimension is added by exploring the mediating role of information-processing depth in this relationship. This provides an addition to current insights, because while its role has been suggested in advertising research (e.g., Aaker and Bruzzone 1985; Ray and Batra 1983), it appears not much empirical testing has been done. Moreover, the introduction of the psychological concept of the negativity bias – even suggesting possible higher effectiveness of irritating advertisements compared to humorous advertisements – means a unique new approach in the field of advertising. As such, the results of this research can potentially aid managers in their decision to consciously opt for irritating elements in their advertising campaigns. The research question formulated for this study is:
Do irritating advertisements have higher advertising effectiveness than humorous advertisements, and is this effect mediated by information-processing depth?
This introduction is followed up by a review on existing literature concerning irritation in advertising and its effectiveness. The role of the negativity bias is further explored with respect to information-processing depth. This literature section functions as a basis upon which the research methodology is built, which will be covered subsequently. After presenting the research results, the discussion section will provide in both concluding remarks and managerial implications.
2. THEORETICAL FRAMEWORK
4 study takes a new approach by including the negativity bias, this concept is further explored in an advertising context. A literature summary at the end introduces the hypotheses of the present study.
2.1. AFFECTIVE RESPONSES TO ADVERTISING
Still widely accepted is the thesis of the “superiority of the pleasant” (Silk and Vavra 1974). This theory serves as an explanation of the effectiveness of humorous advertising and posits that the positive feelings people have about an advertisement will be generalized through a process of classical conditioning and subsequently projected onto the advertising brand. In other words, the positive attitude one has toward an ad will lead to a positive attitude toward a brand (Shimp 1981). Among the first to provide empirical support for this theorem were Mitchell and Olson (1981), who were looking at possible mediating variables in the relationship between product attributes and brand attitude. Their research shows that attitude toward the advertisement mediates brand attitude, but since the study was conducted with fictitious brands the results might be biased because the ads were the only source of information. Nonetheless, a growing body of supporting research emerged in subsequent years, also using datasets from practice. Batra and Ray (1986) propose empirical evidence of a “chain of effects” where affective responses influence the attitude toward the ad, which in turn influences brand attitude, which influences purchase intentions. Just like Lutz, Mackenzie and Belch (1983), they include brand cognitions (perceptions of the advertiser) as a predictor of brand attitudes, and find that the relationship with attitude toward the ad is clearly stronger. Affective responses that influence the attitude toward an advertisement embody the moods and feelings elicited by that advertisement (Batra and Ray 1986). They can take various forms, but can be summarized in warmth, amusement and irritation (Stayman and Aaker 1988). The previous studies however used affective responses as an encompassing independent variable and didn’t differentiate between them. As a result, they didn’t study the effect of irritation specifically. Moreover, while the superiority of the pleasant theory explains the effectiveness of humorous (positive) advertising, it doesn’t explain irritating (negative) advertisements that have proven themselves effective. The classical conditioning principle of the theory implicates that irritation as a negative response also rubs off on the brand and therefore would deteriorate its effectiveness.
5 effective than advertisements that are perceived as being neutral. Over the years, scholars have backed this idea. In research on the relationship between perceptions of audiences and ad impact, Aaker and Stayman (1990) show that commercials perceived as irritating are effective, more so than neutral advertisements. The reason is that attention must be paid before an advertisement has any effect at all, something that irritating advertisements do more so than neutral ones (Ray and Batra 1983). Advertising that is either distinctly humorous or irritating elicits affective responses, which causes people to pay attention and enhances the degree of processing. Neutral commercials tend to go by without people being aware of them more easily. An interesting theoretical construct in this area is the Elaboration Likelihood Model (ELM) of persuasion, as proposed by Petty and Cacioppo (1986). The model proposes two ends of a continuum, where people who carefully process a message take the central route, and people who aren’t able or willing to do so take the peripheral route. Both humor and irritation invoke a form of interest within the perceiver, which can be seen as a motivation to think about the message and process its arguments more thoroughly. Neutral advertising doesn’t trigger this response, and the route taken by the viewer is therefore more dependent on existing involvement with the advertised product or depicted situation. Affective stimuli thus are processed more via the central route of the continuum. Neutral advertising will be processed more peripherally, if picked up at all. Findings by Derbaix (1995) supplement to this idea. Verbal affective responses influence attitude toward the advertisement as well as brand attitude (following immediately after being exposed to the stimuli), whereas facial affective responses (expressions) do not. Verbal affective responses as such require more extensive information-processing than facial responses.
2.2. HUMOR VS. IRRITATION
6 though the law of extremes holds – and both positive and negative affective responses have a positive effect on advertising effectiveness – positive messages are more powerful. There have been studies acknowledging this effect. For example, Edell and Burke (1987) find that positive affective responses resulting from exposure to advertising have a stronger positive impact on advertising effectiveness than negative responses. Stayman and Aaker (1988) come to similar conclusions, in particular for situations where exposure is limited. Second, when an audience perceives an ad to be funny, the resulting positive mood may enhance advertiser credibility, which is said to enhance persuasion (Sternthal and Craig 1973). Perceived irritation on the other hand is suggested to decrease advertiser credibility and erode consumers’ trust in advertised brands (Bartos 1981).
Over the years there have also been studies that point to another perspective: that negative affective responses to advertisements might not necessarily deflect toward the brand. Moore and Hutchinson (1983) find that this can be explained by a difference in immediate and delayed effects, meaning that unfavorable attitudes toward an ad may disappear after sufficient time and result in brand consideration. Exposure to an ad thus creates a sense of familiarity, which in a later stadium can result in brand choice (Zajonc 1980). Chakrabarty and Yelkur (2005) go one step further and question the direct connection between ad irritation and brand attitude. They hypothesize that ad irritation is no significant predictor of brand attitude (as an antecedent for purchase intentions), and find that positive ad-induced feelings and attitude toward the advertiser are. Referring back to the ELM as proposed by Petty and Cacioppo (1986), the explanation might be that through increased processing strong arguments made in the advertisement are accepted despite the irritation caused. This implicates that when managers should purposely opt for irritating advertisements, maybe they don’t hurt their brand attitudes. As a result, they do not necessarily risk losing sales and thus market share. Secondly, research by Fritz (1979) suggests that irritating advertisements are better memorized than positive or neutral advertisements by showing that recall of advertising claims is higher than with non-irritating advertising. Additionally, James and Kover (1992) show that the degree of processing elicited by advertising is higher for people who are irritated, which is explained by suggesting that these people examine the advertisements for points to attack them on.
7 the present. Just to illustrate: on Dutch television almost forty times as much commercials were broadcasted in 2009 as was the case in 1990 (TNS Nipo and Nielsen 2010). This overkill causes people to become numb and almost immune to advertising, so salience is an issue. As a result, companies need to present their message in ways that stand out. Moreover, as stated earlier, currently almost half of all advertising on TV is using some form of humor. The other half is of - inter alia - informative, pleasant and irritating nature. Irritation aspects differentiate from humor and thus are a way to stand out from most of the advertising clutter.
2.3. THE NEGATIVITY BIAS
8 2.4. LITERATURE SUMMARY AND HYPOTHESES
The findings from literature and in particular the “law of extremes” demonstrate that both humorous and irritating advertising has a higher effectiveness than neutral advertising. However, previous research also shows that ad irritation as an affective response behaves differently with respect to advertising effectiveness than positive affective responses do. Positive responses may rub off on the brand and result in conditioned positive feelings toward a brand. Although there is no consensus among scholars, negative responses may do the opposite. If they do, these effects are suggested to disappear after sufficient time. Furthermore, negative responses may result in deeper information-processing than humorous responses, while leaving a greater impression on a person at the same time. Moreover, irritating advertising stands out more from the advertising clutter, of which almost half is of humorous nature, and therefore has a substantially higher chance to attract attention. This leads to the first hypothesis of this study:
H1: Irritating advertisements have a higher advertising effectiveness than humorous advertisements.
Both distinctly humorous and irritating advertisements elicit affective responses, neutral advertising less so. In other words, people are invited to form opinions about these types of advertising. A prerequisite for this to happen is that they are involved with the advertisement and actively process the stimuli. Since humorous and irritating advertisements are more effective than neutral advertisements, it seems that the relationship between the type of advertisements and advertising effectiveness is mediated by information-processing depth. Additionally, the negativity bias theorizes that negative events have a stronger impact on an individual than a positive event of equal value. Connecting that thesis to the field of advertising allows the assumption to be made that negative events in the form of irritation caused by an advertisements features will be more powerful in involving someone than positive feelings elicited by advertisements of similar caliber. Moreover, it is suggested that people examine advertising that irritates them more extensively in order to find aspects that they can use to disapprove and criticize it with. This brings about the second hypothesis of this research:
9 Previous research has shown that deeper information-processing positively influences brand and claim remembrance, which are prerequisites for advertising having effect. The mechanism can be seen as a form of exposure, resulting in familiarity and eventually brand consideration. This produces the third hypothesis of the current study:
H3: Increased information-processing depth has a positive effect on advertising effectiveness.
Combining the proposed theory that irritating advertisements have a stronger impact on processing depth than humorous advertisements with the theory that information-processing depth influences advertising effectiveness produces the fourth and last hypothesis:
H4: The relationship between type of advertising and advertising effectiveness is mediated by information-processing depth.
The hypotheses as discussed in this section are graphically depicted in the conceptual model in figure 1 below. The relationship between the type of advertisement (irritating vs. humorous) and advertising effectiveness is mediated by the depth of information-processing.
Figure 1. Conceptual model
10
3. METHODOLOGY
This chapter outlines the experiment that was conducted in order to answer the research question as proposed in the introduction and repeated below. First the methods of data collection are covered. Thereafter the design of the experiment is spelled out, along with a description of the accompanying pretest. Subsequently, the experimental procedure is illustrated. Finally, the measures are specified.
Do irritating advertisements have higher advertising effectiveness than humorous advertisements, and is this effect mediated by information-processing depth?
3.1. DATA COLLECTION
A total of 104 respondents took part in the experiment. Since essentially everyone in this modern day society is subject to advertising, the research population wasn’t specifically delineated. Considering the time-frame and scope of this research, a trade-off had to be made with respect to a representative sample and the quality of the experiment. It was considered not feasible to attract enough participants of different backgrounds to a staged lab-experiment where no external factors are of influence. A more viable sample would consist of almost exclusively students with a both scientific and economic background, which could produce distorted results. On the other hand, opting for an online survey allows people to be distracted when participating, which could also produce distorted results. However, people are also pre-occupied when they are exposed to advertising in other situations. Furthermore, it would make it possible to attract a more wide and diverse sample. The experiment was therefore set up as an online survey.
11 3.2. DESIGN
Since this study revolves around the hypothesized stronger effectiveness of irritating advertising over humorous advertising, an experiment had to be developed that would include both types. The hypotheses as presented in the previous chapter were therefore tested in a between-group experimental design, manipulating the independent variable type of advertisement. Participants were randomly assigned to either a humorous or an irritating condition, while keeping all other factors constant. This design makes it possible to compare the effects of the two types of advertising on the dependent variable advertising effectiveness. It also provides the possibility to independently measure and compare the effects of both humorous and irritating advertising on information-processing depth. Moreover, having two independent groups made sure that participants were unaware that a comparison between humorous and irritating advertisements was being made.
Two TV commercials from the American car insurance company Geico were chosen to represent the humorous and irritation condition of the independent variable. TV commercials were chosen because they are a form of advertising that everybody is familiar with. Geico being a foreign brand that’s not operating in Europe controls for brand familiarity having an effect on the research results. The commercials are however understandable and come from a culture with a similar taste in humor. Both commercials aired around 2010 and had an equal duration of thirty seconds. The two were part of a single extensive campaign and are therefore structured in exactly the same way. Both start with a man asking whether switching to Geico could really save 15% or more on car insurance. Then a rhetorical question follows which reflects on a subsequent sketch. The humorous commercial features a scene where a former drill sergeant works as a psychiatrist, the preceding question being: “Does a former drill sergeant make a terrible therapist?” The irritating commercial involves a screaming little pig on the backseat of a car, the question here is: “Did the little piggy cry wee, wee, wee all the way home?” Both commercials end with: “Geico, 15 minutes could save you 15% or more”. The fact that both commercials follow the same structure minimizes the chance that the measured difference in effect can be attributed to other factors. At the same time, there are two distinct but equal advertising claims that can be taken away from it: save 15% by switching to Geico in just 15 minutes.
3.2.1. MANIPULATION CHECK
12 as being so. A seven-point bipolar scale consisting of twelve adjective pairs was used to measure ad perception, based on the scale as used by Fritz (1979). Two pairs were altered, to make sure the scale didn’t singularly measure irritation as it did in the study by Fritz, but also whether the commercial was found to be funny and was thus judged as humorous. A reliability analysis on the twelve items was done to check the internal consistency of the scale. Firstly, the analysis reveals a Cronbach’s α of .94, indicating that the scale is internally very consistent. Secondly, it shows that deleting an item wouldn’t result in a higher internal consistency. Lastly, the corrected item-total correlations all being within the .5 and .9 range explain that each item correlates with the overall scale. Furthermore, exploratory factor analysis was done to assert whether the scale was unidimensional. It becomes clear that all items load on the same factor, and that the scale can be used as a single construct measuring ad perception.
3.2.2. PRETEST
A group of 42 people was approached to evaluate the chosen advertisements on the seven-point bipolar scale measuring ad perception, prior to the experiment. Randomly assigned, 22 people were exposed to the
humorous condition and 20 others to the irritation condition. The data are normally distributed (bell-shaped), as confirmed by the Shapiro-Wilk Test of Normality (p=.24). As in this situation two independent groups are being compared on a single independent variable, an Independent T-test is appropriate to see if the groups differ significantly. The results in table 1 show that the means of the scores on the scale differ with each other, the humorous condition tending toward the funny end of of the scale (M=3.33, SD=.84) and the mean of the irritation condition tending toward the irritation end of the scale (M=4.73, SD=.73). For the humorous condition the range is 3.5, whereas it is 2.5 for the irritation condition. The T-test reports that the means differ significantly: t(40)=-5.76, p=.00. As a result the two Geico commercials are considered distinctly humorous and irritating and fit to represent the levels of the independent variable in the experiment. The standard deviations show that in the irritation condition the observations are slightly more centered to the mean. The range indicates that the data are located in a smaller interval for the irritation condition than for the humorous condition. People thus seem to be more unified in their perception of the irritating commercial. In other words, it seems that the humorous commercial was found to be irritating on more occasions than the irritating commercial was found to be humorous.
Table 1. Pretest type of ad and ad perception.
Humorous (0) Irritating (1)
Mean 3.33 4.73
SD .84 .73
13 To conclude the pretest, a question was included asking whether respondents were already familiar with the brand. Only 4.8% (2 of 42) had heard of Geico before, thus implying that the results of the experiment wouldn’t be influenced substantially by brand familiarity and prior knowledge of the brand and/or its advertising campaigns.
3.3. PROCEDURE
Respondents were approached with a link to the experiment (appendix A) on an online platform (Qualtrics). After a short introduction, they were shown a fragment of the nature documentary “Great Migrations: Born to Move” by National Geographic, followed by three 30 second commercials. Before forming this set-up, a testing round among 18 people was carried out. In this initial experiment, the documentary fragment lasted 3 minutes, and five 30 second commercials were included. It became clear that this longer video resulted in a survey that took too much time and allowed participants to lose focus. As a result participants didn’t provide very useful responses. Based on this experience the documentary was shortened to 1.5 minutes and two commercials were excluded. An even shorter video wouldn’t give enough room for humorous and/or irritating advertising to actually stand out. Via Qualtrics, a timer was set which revealed the continue button only after 3.5 minutes, in order to make sure that people wouldn’t click continue before the video was over. Participants were randomly assigned to either the humorous or irritating Geico commercial. The other two commercials were filler ads of neutral nature. They aired in 2015 during the Super Bowl and are assumed as neutral based on their low scoring on an influential yearly ranking system (Ad Meter 2015). The order of commercials was randomized to guard for possible order effects, like for example the last or first ad in the video being remembered relatively better.
14 whether people were aware of the actual research goal. This allows to exclude respondents who indicated they were aware from analysis to see if this results in distorted results.
3.4. MEASURES
To test the hypotheses, the video including the documentary and the three commercials was followed with a questionnaire. Both the effect of the different advertising types on the dependent variable advertising effectiveness and the depth of information-processing were measured.
3.4.1. ADVERTISING EFFECTIVENESS
Testing advertising effectiveness directly in terms of actual purchase behavior was not feasible considering the time and scope of this study. Following Fritz (1979), unaided claim recall was utilized as a stand-in measure of the effectiveness of the advertisements. Claim recall allows to see how well each advertisement worked and subsequently allows to make inferences about the effect of irritating and humorous advertising. As such, after being exposed to the commercials the participants were asked to write down as many advertising claims as they could remember. Naturally, claim recall of the target ad - the Geico commercial - was of interest here. As stated before two distinct claims could be remembered. Since both commercials contained the exact same two claims, a simple summation of claims constitutes the measurement of the dependent variable advertising effectiveness. For example, a respondent who writes down that switching can save 15% remembers one claim, while another who also adds that this takes just 15 minutes remembers two. When the claims of the irritating commercial were remembered most compared to the humorous commercial, H1 can be accepted.
15 down as much brands and product-types as possible for the three commercials. Even though this question has a function as a filler question it also allowed for some interesting statistics, namely how high brand recall and product type recall was for the target ad in the humorous and irritation conditions. Furthermore, recoding the data into a third variable measuring either brand recall, product recall or both provides insight in how much participants were able to remember.
A question measuring mood at the time of participation was also included. Using this as a covariate allows to control for possible effects that this variable has on the relationship between type of advertising and claim recall. This question was set up exactly the same as in a study by Sar, Nan and Myers (2010): as a two item seven-point bipolar scale, ranging from sad to happy and from bad to good. Its Cronbach’s α of .88 indicates internal consistency.
3.4.2. INFORMATION-PROCESSING DEPTH
The depth of information-processing for the Geico ad was measured with a three item seven-point Likert scale, based on a four item scale as used by Wang (2006). The scale asked participants how much attention was paid, how engaging the commercial was and to which extent it involved them, where 1 = “completely disagree” and 7 = “completely agree”. A reliability analysis on the scale gives a Cronbach’s α of .85, which indicates that the three items are consistent within the scale. Deleting one of the three items would only result in a lower α. The corrected item-total scores range between .7 and .8, which signifies that the items individually correlate with the overall scale. Additional exploratory factor analysis reveals that the scale doesn’t have any underlying dimensions and can therefore be used as a single construct.
4. RESULTS
16 4.1. SAMPLE
To assure unbiased results, the sample consisting of 104 respondents was checked for anomalies prior to analysis. First, the individual durations of the respondents’ participation in the experiment were checked. Since the video had a playing time of 3½ minutes and in general participation duration was close to 10 minutes, cutoff points were applied at 4½ and 20 minutes. Experiments completed in a shorter timespan can reasonably be assumed to have been answered carelessly or even untruthfully, while experiments taking longer than twenty minutes can provide biased answers. The data of three respondents was deleted as a result. Second, the data were checked for obvious nonsensical answers, leading to omitting the results of one person stating sheep, cows and chickens were observed in the documentary. Finally the data were checked for outliers. Three respondents chose exclusively extreme answers on the semantic scale measuring advertisement perception. Their answers on the open questions as well as the scale measuring information processing-depth reveal that these extreme values were chosen because they didn’t see the Geico commercial. The data of the three opting for extreme answers were omitted from the final analysis, following the reasoning that they bias the results. However, all tests were performed including these respondents as well, making sure that leaving these results out didn’t influence the main outcomes on the various tests. Similarly, seven other respondents chose exclusively neutral answers. The data of these respondents was left in since it isn’t possible to determine that they didn’t watch the video. Also, these data do not strongly bias the results. Furthermore, performing the main analyses without these respondents as well shows that exclusion doesn’t result in different outcomes. In summary, the data of 7 of the 104 respondents were omitted from analysis for the following reasons, resulting in a dataset of n=97:
- 2 respondents completed the experiment in under 4½ minutes.
- 1 respondent took longer than 20 minutes to complete the experiment. - 1 respondent provided unmistakably nonsensical answers.
- 3 respondents chose exclusively extreme answers on the various scales.
4.2. MANIPULATION AND COVER STORY CHECK Respondents were asked to rate the Geico commercial on a twelve item scale measuring advertisement perception. This question was included as a manipulation check. Using an
Table 2. Manipulation check type of ad and ad perception.
Humorous (0) Irritating (1)
Mean 3.66 4.74
SD .92 1.12
17 Independent samples T-test in the same manner as was done in the pretest is appropriate, since the data are normally distributed (bell-shaped). This is confirmed by the Shapiro-Wilks Test of Normality not being significant (p=.17). Again the commercials are judged as being significantly different. The T-test results as depicted in table 2 show very similar statistics compared to the preT-test. Unlike in the pretest, the humorous condition the mean lies slightly closer to the center of the seven-point scale (M=3.66, SD=.92). The irritation condition is perceived almost identical (M=4.74, SD=1.12). In other words, compared to the pretest the humorous condition is perceived a little less funny, whereas no difference is observed in the irritation condition. The range for the humorous condition is 4.08, and 4.50 for the irritation condition. As was the case in the pretest, the T-test here reveals that the irritating commercial was found to be significantly more irritating than the humorous one: t(95)=-5.23, p=.00. As a result the manipulation can be considered successful, an important aspect considering the value of the results of the experiment.
The last question of the questionnaire asked whether respondents were aware of the actual goal of the research while participating in the experiment. This query followed after the debriefing that cleared up the deception used in the cover story. On a yes or no question, 9.3% (9 out of 97 respondents) indicated that they knew what was actually being researched. It is possible however that they just answered ‘yes’ to come across as being smart. In any case, to be sure this didn’t influence the outcomes of the analyses done on the data, all were done excluding the respondents who indicated that they knew the research objective. No outcomes were found to be different.
4.3. DEMOGRAPHICS
The 97 respondents forming the final dataset can be described on the basis of three demographic statistics as depicted in table 3 on the next page. For starters, it becomes clear that the ratio male/female
lies at around 60/40.
Considering age, the youngest respondent is 14 and the eldest
Sample Humorous (0) Irritating (1)
18 60 years old, with a mean of 33.21. The median lying slightly above 27 indicates that the bulk of the sample is aged between 24 and 29 years old. Furthermore, the sample is largely well educated, considering the fact that 33% indicated to have been educated on a scientific level and 45% on a higher professional level. These numbers seem to indicate that the distribution on these demographics is non-normal, which is confirmed by the Shapiro-Wilk Test of Normality being significant (p=.00 for all three variables). The same is the case for another interesting descriptive variable: mood at the time of participation (p=.00). The data is skewed right for age and gender and skewed left for education and mood. These non-normal distributions implicate that the mean isn’t located in the middle. Parametric tests such as an Independent samples T-test assume that it does, which could produce biased results. However, following the central limit theorem the sample size is large enough to perform parametric tests (Rice 1995), making the T-test fitting to check for significant differences in demographics and mood at the time of participation.
As depicted in table 4, the Independent samples T-test reveals that there are no significant differences between the two conditions. First, age for the humorous condition (M=34.57, SD=12.27) doesn’t differ significantly from age for the irritation condition (M=31.81, SD=11.93), t(95)=1.12, p=.27. Second, the male/female ratio for the
humorous condition (M=.41, SD=.50) is nearly equal to the irritation condition (M=.40, SD=.49), t(95)=.12, p=.90. Third, the humorous condition is equally well educated (M=6.00, SD=.94) compared to the irritation condition (M=5.96, SD=1.10), t(95)=.20, p=.84. Lastly, the mood at the time of participation in the humorous condition (M=4.91, SD=1.03) was almost equal to the mood in the irritation condition (M=5.10, SD=1.02), t(95)=-.89, p=.38. As such, the two groups are not significantly different from each other and thus suitable for further analysis. An additional Mann-Whitney U test (the T-test’s non-parametric counterpart) was carried out to be sure the non-normal distribution didn’t play a role. The test statistics also proved to be not significant for all variables: age (U=966, p=.13), gender (U=1161.5, p=.90), education (U=1160.5, p=.91) and mood at the time of participation (U=1053, p=.37).
Mean SD T-test Significance
Age 0 1 34.57 31.81 12.27 11.93 1.12 .27 Gender 0 1 .41 .40 .50 .49 .12 .90 Education 0 1 6.00 5.96 .94 1.10 .20 .84 Mood 0 1 4.91 5.10 1.03 1.02 .89 .38
19 4.4. MAIN DESCRIPTIVE STATISTICS
As was explained in the
methodology chapter, the
experiment included an open question measuring brand name
and product recall. This question isn’t used for hypotheses testing. It does however provide interesting statistics, as can be seen in table 3. It reveals that 47.4% of the respondents recalls Geico as a brand, but only 19.6% can recall what the product was that was being advertised by this brand. Taken together, 52.6% of respondents could recall either brand, product or both. This shows that for example a person that remembers the brand does not necessarily remember which product was being advertised. Moreover, taking the humorous and irritation conditions apart shows that both brand and product recall is higher after exposure to the irritating Geico advertisement than after exposure to the humorous advertisement. Brand recall is 50% for the respondents who were exposed to the irritating commercial as opposed to 44.9% for those exposed to the humorous commercial. For product recall these figures are respectively 20.8% and 18.4%. For either brand recall, product recall or both the difference is the most sizable, namely 58.3% and 46.4%.
Again an Independent samples T-test is employed to test whether the difference between the two groups on their brand and product recall is of statistical significance. The Shapiro-Wilk test of Normality being significant for
all variables (p=.00) means that the data are distributed non-normal (all skewed right), but with respect to the central limit theorem the T-test should be able to handle this. The T-test as summarized in table 6 shows that brand recall for the irritation condition isn’t significantly higher (M=.50, SD=.51) than brand recall for the humorous condition (M=.45, SD=.50), t(95)=-.50, p=.62. The same goes for product recall, which isn’t statistically higher for the irritation condition (M=.21, SD=.41) than for the humorous condition (M=.18, SD=.39), t(95)=-.30, p=.76. Also brand and product recall combined isn’t higher for the irritation condition (M=.58, SD=.50) compared to the humorous condition (M=.47, SD=.50), t(95)=-1.12, p=.27. An additional non-parametric test in the form of the Mann-Whitney U test provides almost identical results and thus reveals the non-normal distribution
Table 5. Descriptive statistics.
Sample Humorous (0) Irritating (1)
Brand recall 47.4% 44.9% 50%
Product recall 19.6% 18.4% 20.8%
Brand and product recall 52.6% 46.6% 58.3%
Mean SD T-test Significance Brand recall 0 1 .45 .50 .50 .51 -.50 .62 Product recall 0 1 .18 .21 .39 .41 -.30 .76 Brand and product recall 0 1 .47 .58 .50 .50 -1.12 .27
20 isn’t of influence. The test statistics for neither brand recall (U=1116, p=.62), product recall (U=1147, p=.76) or brand and product recall combined (U=1042, p=.26) are significant.
4.5. MAIN HYPOTHESIS The main hypothesis (H1) of this study states that irritating advertisements have a higher advertising effectiveness than humorous ones, where advertising
effectiveness is measured via claim recall. 47.4% didn’t remember any claims, 29.9% remembered one and 22.7% remembered two claims. For H1 to be true, claim recall has to be significantly higher for the irritation condition than the humorous condition. An Independent samples T-test is appropriate to test the two groups on this relationship, since the test results provide a clear view on the difference between groups and give insight in the size of the difference. This is of value since in the current research it is interesting to know which type of advertising has higher effectiveness, and how much the difference is.
First, a significant Shapiro-Wilks of Normality test (p=.00) reveals that the data of claim recall have a non-normal distribution (skewed right). However, following the central limit theorem once more, this shouldn’t be of influence. The T-test reveals that claim recall is significantly higher for the irritating commercial (M=.96, SD=.85) over the humorous commercial (M=.55, SD=.71), t(95)=-2.57, p=.01. The standard deviations however are relatively high, which suggests that the various data points do notably deviate from the mean. This suggests a large variation in the observed sample, logically explainable as some respondents remember zero claims and others two. To control for a possible influence of the non-normal distribution an additional Mann-Whitney U test is performed. The test statistic is significant, and thus discloses that this is not the case: U=867, p=.02. Based on this analysis the assumption can safely be made that claim recall for irritating advertising is significantly higher than for humorous advertising. As such, H1 can be accepted.
4.6. INFORMATION-PROCESSING DEPTH The second hypothesis of this study
assumes that irritating advertising
Table 7. Independent samples T-test type of ad and claim recall.
Mean SD T-test Significance Claim recall 0 1 .55 .96 .71 .85 -2.57 .01
Mean SD T-test Significance Information-processing depth 0 1 2.83 2.62 1.41 1.35 .76 .45
21 results in more extensive information-processing depth than humorous advertising. Similarly to the dependent variable, the data of information-processing depth are not normally distributed (skewed right), as revealed by a significant Shapiro-Wilks test (p=.00). Again an Independent samples T-test is carried out for the same reasons as with the first hypothesis, the results of which are depicted in table 8. The T-test proves what the means already indicate: the means for the humorous condition (M=2.83, SD=1.41) do not significantly differ from the means of the irritation condition (M=2.62, SD=1.35), t(95)=.76, p=.45. To make sure the non-normal distribution isn’t of influence, an additional Mann-Whitney U test was again carried out. The test statistic reveals that there is no effect (U=1072, p=.45). Therefore, it can be assumed that there is no significant relation between type of advertising and information-processing depth. Interestingly enough, if significant, the higher mean for the humorous situation would even mean that increasing irritation would decrease information-processing depth. Therefore, H2 cannot be accepted.
The third hypothesis predicts that the relation between information-processing depth and claim recall is a positive one. Since the independent variable in this relation is a continuous variable (information-processing depth), a simple linear regression model is appropriate. The results of the basic regression analysis are given in table 9. The model is overall significant, as the accompanying ANOVA reveals (p=.02). Furthermore, in this model the R square is very low, suggesting that only approximately 6% of the variance in claim recall is explained by information-processing depth and conversely approximately 94% is attributable to other factors. The regression analysis reveals that indeed information-processing depth has a positive influence on
claim recall (p=.02). The unstandardized beta indicates that an increase by 1 in information-processing depth would result in an increase by .14 in claim
recall. The regression analysis shows that H3 can be accepted.
4.7. MEDIATION
The research question of this study proposes a mediator effect, which is put forward in H4. Increased information-processing depth is the process that is suggested to mediate the relationship between type of advertisement and claim recall. For this statement to hold, four conditions have to be
Claim recall
R Square .06
Adjusted R Square -.05 Model significance .02 Unstandardized Beta Standardized Beta Significance
(Constant) .36 .04
Information-processing depth
.14 .25 .02
22 satisfied (Hayes 2009): (1) the independent variable must be related to the dependent variable, (2) the independent variable must be related to the mediator, (3) the mediator has to be significant in the model and (4) when the mediator is included, the direct effect of the independent variable should be non-significant (full mediation) or at least have a much lower effect (partial mediation). The mediation is represented as follows:
(1) Υ = β0+ β₁Χ + 𝑒 (2) 𝑀 = β0+ β₁Χ + 𝑒 (3) Υ = β0+ β₁𝑀 + 𝑒 (4) Υ = β0+ β₁Χ + β2𝑀 + 𝑒 where β0 = intercept
β₁Χ = linear effect of type of advertising
β₁𝑀 = linear effect of information-processing depth
There are multiple tests designed to assess mediation. The most common is the Baron and Kenny (1986) test. This method however is said to have low statistical power (MacKinnon et al., 2002; Hayes, 2009). The Baron and Kenny method tests if the independent variable is related to the mediator and the mediator related to the dependent variable separately. Bootstrap methods test this relationship directly and are said to be more reliable (Shrout and Bolger, 2002). This study uses both methods to provide results as accurately as possible.
Testing mediation via the Baron and Kenny (1986) approach can be done with the PROCESS macro for SPSS by Hayes (2013). This macro is based on a simple linear regression
model. Confirming the ANOVA as described in paragraph 4.4, the regression analysis proves that there is a relation between the independent variable type of ad and the dependent variable claim recall, thus fulfilling
condition one. However as specified before, H2 cannot be accepted, which means that there is no relation between the independent variable and the mediator and condition two isn’t met. Condition three states that the mediator has to be significant within the model, which is the case based on the p-value as depicted in table 10. Condition four means that the original relation between the
Table 10. Simple mediation model type of ad, information-processing depth and claim recall.
Mediation
R Square .14
Model significance .01 Unstandardized Beta Significance
(Constant) .11 .56
Type of ad .44 .01
Information-processing depth
23 independent and dependent variable now has to be non-significant or at least of much lower effect. Table 11 shows that this isn’t the case. The effect even increases slightly. Since two conditions aren’t met, the Baron and Kenny approach reveals that there is no mediation going on. This is underlined by the bootstrap method, which directly tests the indirect effect of type of advertising of claim recall. The bootstrapping method asks whether it is possible that the indirect effect is zero, or in other words, that there is no indirect effect. Within a 95% confidence interval, zero lies within the lower and upper level (LL=-.16, UL=.05). H4 therefore cannot be accepted for failing to meet multiple conditions over two tests of mediation.
Even though mediation can’t be proved, an interesting observation can be made. The full model in its entirety and the mediator within the model are significant, even though condition two isn’t met. At the same time, the intercept of the full model is highly insignificant, indicating that it is not significantly different from zero when the predictor variables are zero. This simply implies that advertising effectiveness can be zero. However, this observation could also be caused by multicollinearity, meaning that there is a linear relationship between type of advertising and information-processing depth. The VIF values (1.01) however do not exceed 10, and suggest that this is not the case. A bivariate correlation matrix confirms this with a non-significant Pearson Correlation coefficient (r=.08, p=.45).
4.8. CONTROL VARIABLES To control for the possible influence of other variables on the relation between type of advertising and claim recall, demographics and mood at the time of participation in the experiment were added to the mediation model as independent variables. Table 11 shows that the model overall stays significant
when including these control variables. Also, type of advertisement and information-processing depth remain significant. It furthermore reveals that neither mood (t=-.55, p=.59), gender (t=-1.10, p=.28), age (t=-.27, p=.79) or education (t=.80, p=.42) has a significant effect on claim recall. Therefore the tested variables do not need to be included as covariates in the full mediation model.
Unstandardized Beta Significance
(Constant) .06 .93 Type of ad .44 .01 Information-processing depth .16 .01 Mood -.04 .59 Gender -.18 .28 Age -.00 .79 Education .07 .42
24
5. DISCUSSION
This chapter provides a thorough discussion on the results as presented in the previous chapter, thereby connecting it to existent research and practice. Furthermore, the limitations of this study are acknowledged and possible future research paths are suggested.
5.1. DISCUSSION OF RESULTS
The main goal of this study is to further explore the power of irritating elements in advertising, especially in relation to the popular form of humorous advertising. It is hypothesized that irritating advertising is more effective, mainly via increased information-processing depth caused by the negativity bias. Through experimentation, claim recall of both humorous and irritating advertising was measured, as well as the depth of information-processing that went with it. The hypotheses that have been accepted or rejected are summarized in table 12. While two hypotheses are rejected, this doesn’t explicitly imply that these relations are non-existent. Rather, these relations can be very much actual concepts but this study hasn’t been able to prove that this is indeed the case.
H1 Irritating advertisements have a higher advertising effectiveness than humorous
advertisements. Accepted
H2 Irritating advertisements result in greater information-processing depth than humorous
advertisements. Rejected
H3 Increased information-processing depth has a positive effect on advertising
effectiveness. Accepted
H4 The relationship between type of advertising and advertising effectiveness is mediated
by information-processing depth. Rejected
First of all, the hypothesized stronger effectiveness of irritating advertising over humorous advertising is found to be true. The experiment shows that claim recall is significantly higher for the irritating advertisement than for its humorous counterpart. This confirms findings by Fritz (1979), who also experimented with TV commercials. The results however contradict a substantial amount of research which assumes a J-shaped relationship, attributing humorous advertising a stronger effect (e.g. Edell & Burke 1987; Silk and Vavra 1974). A possible explanation can be found in the fact that
25 this literature mostly originated minimally two or three decades ago, in an advertising landscape very different from the one that is currently existent. In current times, commercial allocated time on mass media like radio and television has skyrocketed. At the same time, people are subject to advertisements on the internet on nearly every website, from Google to social media like Facebook and Twitter, as well as in the applications on their smartphones. People are likely to be more acclimated to, or more forthright, even numbed by the enormous amount of advertising they encounter on a daily basis. Therefore it is more and more challenging for advertisements to stand out from the clutter of various advertisements to attract the attention of the consumer (Pieters, Warlop and Wedel 2002). It is reasonable to assume therefore that standing out through irritating elements actually makes an important difference, especially since humorous advertising is barely unique.
26 the irritation condition respondents were more inclined to check which brand brought them the unpleasant experience, while in the humorous condition they just enjoyed the storyline and moved on, a reasoning which does match James and Kover’s (1992) findings.
So while no relation between type of advertisement and information-processing depth can be determined, this study does confirm earlier findings that information-processing depth and claim recall are positively related. In the experiment there is indeed a significant positive relationship between information-processing depth and claim recall, which means that respondents who indicated that they awarded either the humorous or irritating commercial more attention also score high on claim recall. This is a probable result, since it makes sense that people who were more concentrated on the commercials would indicate a higher involvement, as well as remember more details. The experiment however also reveals that information-processing depth only explains a small fraction of the variance in advertising effectiveness, which means that other factors may exist that have a more sizeable role.
27 are cognitively occupied are relying less on their knowledge of the persuasive nature of advertising compared to people who aren’t. This would imply that distraction here is the factor that mediates the relationship between type of ad and claim recall.
5.2. MANAGERIAL IMPLICATIONS
The results of this study confirm that irritating advertising indeed has its value and is effective in attracting attention. Moreover, the outcomes prove that irritating advertising can provide even higher advertising effectiveness than humorous advertising.
First off, for firms the effectiveness of irritation implies that implementing irritating elements in advertising has a positive effect on advertising effectiveness (as an antecedent for purchase behavior). Advertising effectiveness was measured by claim recall, which doesn’t take into account attitudes the perceiver may have toward the ad, unlike other antecedents of actual purchase behavior, like brand attitudes and purchase intentions. It is however accepted that negative associations resulting from irritation can deflect upon the brand and therefore be harmful for brand attitudes and purchase intentions, although scholars do not seem to fully agree on this thesis. For example Chakrabarty and Yelkur (2005) find ad irritation to be no predictor of brand attitude. At the same time, following Moore and Hutchinson’s (1983) reasoning, unfavorable attitudes may deflect over time and result in brand consideration. It is however advised to use irritating elements with caution.
Another implication that can be put forward has to do with the fact that the study proves that information-processing depth has a positive influence on claim recall. Therefore inducing people to process an advertisement will help in increasing claim recall, and in turn brand familiarity and consideration. No prove is provided whether irritating advertising has a role in this relationship, other than that attracting attention is an important prerequisite for people to do so. This means that for example an irritating element added to an otherwise informative type of advertisement may very well increase its effectiveness.
28 important as well. Managers should aim to use irritation to their advantage, and focus it on catching attention. This means for instance balancing a TV commercial out with an irritating aspect and sufficient pleasant and/or humorous aspects to refrain from irritating people even more when their attention has been incited. Irritation elements can also provide a significant role in marketing campaigns for new brands, creating brand awareness in the early stages and shift to for example more informative advertising once this goal has been reached.
5.3. LIMITATIONS AND FUTURE RESEARCH
As with all experimental studies, the present research is confined to certain limitations. First of all there are the demographics of the sample. Considering age, the average of the sample is fairly representative since the average age of the Dutch population is 39. But, as stated before, it consists of mainly highly educated people. With 33% of respondents educated on a scientific level and 45% educated on a higher professional level, a total of 78% has followed higher education. The average of people educated on a higher level in The Netherlands lies around 28% (CBS 2013). As a result, the findings done in this study can’t be generalized for the entire Dutch population, but only reflect people with a higher educational background. Moreover, the sample size of 97 respondents who participated in the experiment is relatively small, which gives the study limited statistical accuracy. Future research would be advised to improve statistical accuracy through using larger samples and including a wider variety in educational backgrounds.
Another limitation results from the fact that TV commercials were used in the experiment. The commercials originally aired in the USA, thus were English spoken. While The Netherlands has the second highest English proficiency of non-native speaking countries in the world according to the EF English Proficiency Index (EF Education First 2014), it still isn’t a native tongue. It therefore is reasonable to assume that for example claim recall results might be influenced by an individual’s English proficiency. However, using foreign commercials did allow to control for the possible influence of brand familiarity on the results. The influence of product category involvement is a possible factor as well. For example, there is more existing product involvement with car insurance for people who own cars than for people who have no affinity with cars whatsoever. This could be of influence to the results. The experiment lacks a variable to control for this possible factor.
29 attention. After a short filler task they were subsequently asked to answer questions about the commercials in order to measure advertising effectiveness. In a real-world situation it is likely that people pay less attention to commercials than when participating in an experiment. Furthermore, a buying situation is likely to occur much later after being exposed to the commercial than in the experiment. The delayed effect caused by the filler task might as such be insufficient to reflect a real-world situation. More attention combined with a short delayed effect need to be kept in mind when trying to generalize this study’s results on advertising effectiveness. The two Geico commercials that reprised the two levels of the independent variable however were identical in length, structure, advertising claims and accompanying narration. As a result the difference in claim recall and information-processing depth may be contributed in a large portion to the difference in the depicted humorous or irritating situations. With that reasoning the situation can be considered appropriate for research purposes but interpreting the results in a real-life setting should be done with caution.
Caution should also be taken considering other advertising forms and product categories. First, the results of this study possibly aren’t generalizable over multiple advertising forms. It isn’t clear whether irritating elements that are exclusively visual like in for example print advertising or direct mail will have the same effect as TV commercials which are both visual and audial. Other results could be found with other types of advertising. Similarly, the results possibly aren’t generalizable over product categories either, since the commercials used in the experiment focus around one product category only: car insurance. This product category might be better suitable for irritating than humorous advertising. It is reasonable to assume that certain types of advertising like irritation and humorous match better with specific product categories than others. Future research employing other product categories and advertising forms is needed.
31
BIBLIOGRAPHY
Aaker, David A. and Donald E. Bruzzone (1985), “Causes of Irritation in Advertising,” Journal
of Marketing, 49(2), 47-57.
Aaker, David A. and Douglas M. Stayman (1990), “Measuring Audience Perceptions of Commercials and Relating them to Ad Impact,” Journal of Advertising Research, 30(4), 7-18.
Ad Meter (2015), “2015 Ad Meter results: All 61 Super Bowl XLIX ads, from best to worst”, [Available at: http://admeter.usatoday.com], (Accessed on March 18th, 2015).
Alden, Dana L., Wayne D. Hoyer and Chol Lee (1993), “Identifying global and culture-specific dimensions of humor in advertising: A multinational Analysis,” Journal of Marketing, 57(2), 64-75.
Aronson, Elliot, Timothy D. Wilson and Marilynn B. Brewer (1998), “Experimentation in Social Psychology,” in The Handbook of Social Psychology, Daniel T. Gilbert, Susan T. Fiske, and Gardner Lindzey, eds., Boston: McGraw-Hill, 99-142.
Baron, Reuben M., & David A. Kenny (1986), “The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic and Statistical Considerations,” Journal of
Personality and Social Psychology, 51(6), 1173-1182.
Bartos, Rena (1981), “Ads That Irritate May Erode Trust in Advertised Brands,” Harvard
Business Review, 59(4), 138-140.
Batra, Rajeev and Michael L. Ray (1986), “Affective Responses Mediating Acceptance of Advertising,” Journal of Consumer Research, 13(2), 234-249.
Baumeister, Roy F., Ellen Bratslavsky, Catrin Finkenauer, and Kathleen D. Vohs (2001), “Bad is Stronger than Good,” Review of General Psychology, 5(4), 323-370.
CBS (2013), “Onderwijsniveau Bevolking Gestegen,” [Available at: http://www.cbs.nl], (Accessed on June 4th, 2015).
Campbell, Margaret C. and Amna Kirmani (2000), “Consumers’ Use of Persuasion Knowledge: The Effects of Accessability and Cognitive Capacity on Perceptions of an Influence Agent,” Journal of
Consumer Research, 27(1), 69-83.
Chakrabarty, Subhra and Rama Yelkur (2005), “The Effects of Ad Irritation on Brand Attitudes,” Journal of Promotion Management, 11(2-3), 37-48.
32 De Pelsmacker, Patrick and Joeri van den Bergh (1998), “Advertising Content and Irritation: A Study of 226 TV Commercials,” Journal of International Consumer Marketing, 10(4), 5-27.
Derbaix, Christian M. (1995), “The Impact of Affective Reactions on Attitudes toward the Advertisement and the Brand: A Step toward Ecological Validity,” Journal of Marketing Research, 32(4), 470-479.
Duncan, Calvin P. and James E. Nelson (1985), “Effects of Humor in a Radio Advertising Experiment,” Journal of Advertising, 14(2), 33-40.
EF Education First (2014), “EF English Proficiency Index,” [Available at: http://www.ef.nl], (Accessed on June 4th, 2015).
Eisend, Martin (2009), “A Meta-Analysis of Humor in Advertising,” Journal of the Academy of
Marketing Science, 37(2), 191-203.
Fang, Xiang, Surendra Singh, and Rohini Ahluwalia (2007), “An Examination of Different Explanations for the Mere Exposure Effect,” Journal of Consumer Research, 34(1), 97-103.
Fennis, Bob M. and Arnold B. Bakker (2001), “Stay Tuned-We Will be Back Right after These Messages: Need to Evaluate Moderates the Transfer of Irritation in Advertising,” Journal of
Advertising, 30(3), 15-25.
Finkenauer, Catrin and Bernard Rimé (1998), “Socially Shared Emotional Experiences vs. Emotional Experiences Kept Secret: Differential Characteristics and Consequences,” Journal of Social
and Clinical Psychology, 17(3), 295-318.
Fritz, Nancy K. (1979), “Claim Recall and Irritation in Television Commercials: An Advertising Effectiveness Study,” Journal of the Academy of Marketing Science, 7(1), 1-13.
Greyser, Stephen A. (1973), “Irritation in Advertising,” Journal of Advertising Research, 13(1), 3-10.
Hayes, Andrew F. (2009), “Beyond Baron and Kenny: Statistical Mediation Analysis in the New Millenium,” Communication Monographs, 76(4), 408-420.
Hayes, Andrew F. (2013), Introduction to Mediation, Moderation, and Conditional Process
Analysis: A Regression-Based Approach. New York: Guilford.
Hicks, Joshua A. and Laura A. King (2011), “Subliminal Mere Exposure and Explicit and Implicit Positive Affective Responses,” Cognition & Emotion, 25(4), 726-729.
James, William L. and Arthur J. Kover (1992), “Observations: Do Overall Attitudes toward Advertising Affect Involvement with Specific Advertisements?” Journal of Advertising Research, 32(5), 78-83.
