Alison Johnston Thesis

Mixed Emotions: Emotional Juxtaposition in Online Advertising Alison Johnston, Clifford Nass Communication between Humans and Interactive Media Lab Communication Department, Stanford University Stanford, CA 94305

ABSTRACT

However, while advertisers look at the type of content next to the ad, little is done in terms of evaluating the emotion of the content and the effect that this might have on the actual advertisement and perception of the product.

In The Emotion Commotion [3], Aaron Marcus discusses the importance of emotional factors on user interface and product design. While the effects of emotions on consumer behavior have been studied for decades, little is known on the effects of emotional juxtaposition in advertising. In order to test this effect, we gave participants four news stories combined with banner advertisements of matching and mismatching emotions. Our results found that while the emotional juxtaposition of content and advertisements did not affect perception of the ads, it did affect perceptions of the content. In particular, sad advertisements significantly improve the reception of happy content and increased the likelihood that the participant would recommend it to a friend.

Additionally, very few content providers consider the effects that ads have on their content—beyond suggesting certain morals (e.g., the negative connotations of a cigarette or liquor advertisement) and visual irritation (e.g., flashing banner ads). To better understand the affects of emotions on consumer behavior and interaction, one can look at the research discussed by Byron Reeves and Cliff Nass in The Media Equation [4]. After running a series of studies assessing the effects of emotions on the actions and memory of participants, the authors state, “Negativity not only causes people to remember whatever is happening at the moment it is experienced. It also changes the way information is processed after the negative experience is over.”

KEYWORDS

Emotions, Emotional Juxtaposition, Ads, Advertising, Online Advertising Placement

They also find that “painful and negative experiences get attention…People reallocate thinking, store information differently and experience feelings that change everything from how we feel to how our bodies function.”

INTRODUCTION

For decades, advertisers have been concerned with ad placement. Beer advertisements adorned sports stadiums while make-up and designer clothing advertisements filled women’s magazines. With the advent of the Internet, advertisers were suddenly able to pair ads with keywords and content leading to more specific targeting. Banner advertisements on a Facebook page have information regarding your age, gender, club affiliations, interests and romantic status and can target the user accordingly. Similarly, ads on Google search are able to know what a person is looking for at that very moment.

This suggests that the emotions of ads and the content surrounding them would change the way that the ads and content are perceived as readers adjust their focus and trigger their memories in different manners to accommodate for happy or sad feelings. As a result, one could hypothesize that sad content will improve the performance of an ad and a sad ad will improve the perceived quality of content, because subjects will pay more attention and remember more

1

about the content or ad after the fact. However, past research has also shown that people are more likely to remember mood-congruent information than mood-incongruent information. In a study conducted in 1981, Bower, Gilligan and Monteir [1] [2] put subjects in a happy or sad mood before reading stories about several characters. The following day, the subjects remembered more about the characters whose moods were the same as the subjects’ than about the characters with opposing moods. This leads us to assume that while sad content and ads will positively affect impressions and recollections, the juxtaposition of similar emotions will also improve recall, meaning that people will be most likely to remember the content and ads from the sad/sad combination, but will also be able to recall happy/happy combinations nearly as well. METHOD

Design To test this hypothesis, we designed a study that paired happy and sad content with happy and sad banner advertisements in matching and contrasting situations. In order to maximize emotional reactions to the content, we pretested a series of recent news stories, taking care to select stories that couldn’t easily be pinned to a certain date. We selected the two stories with the happiest and the saddest ratings based on the pretest. Our two happy stories covered the rising role of women in politics and the current positive state of the country of Mauritius. Our sad stories covered Mexican drug killings and AIDS in Rwanda. These would serve as HappyContent1 (HC1), HappyContent2 (HC2), SadContent1 (SC1), and SadContent2 (SC2) in our study. To portray the same products and services in both a happy and sad light to prevent product bias, we created banner advertisements from scratch. After a series of pretests, we were left with ads for a dating site, a job site, a leukemia research organization and a brand of wine. Each had two versions of the ad with the same image and fonts—one with text focusing on something happy and the other with text focusing on something sad. For example, the happy Leukemia

research ad boasted about a young boy who had beat leukemia, while the sad ad discussed the number of children who had died from leukemia. The banner ads were then embedded on the side of the story text exactly as in traditional news websites. These story/ad combinations were combined to create eight questionnaire variations, each with four different story/ad combinations: happy content/happy ad, sad content/sad ad, sad content/happy ad and happy content/sad ad. We also ensured that each questionnaire had all four stories and all four products and services advertised. Diagram 1 shows the structure of the eight versions of the study. The four happy ads are represented by HA1, HA2, HA3 and HA4. The four sad ads are represented by SA1, SA2, SA3 and SA4. Of course, a given participant would not receive both a happy ad and a sad on the same topic. Q1 HC1/HA1

SC1/SA2

SC2/HA3

HC2/SA4

Q2 HC1/SA1

SC1/HA2

SC2/SA3

HC2/HA4

Q3 SC1/HA1

HC1/SA2

HC2/HA3

SC2/HA4

Q4 SC1/SA1

SC1/HA2

HC2/SA3

SC2/HA4

Q5 HC1/HA3

SC1/SA4

SC2/HA1

HC2/SA2

Q6 HC1/SA3

SC1/HA4

SC2/SA1

HC2/HA1

Q7 SC1/HA3

HC1/SA4

HC2/HA1

SC2/SA2

Q8 SC1/SA3

HC1/HA4

HC2/SA1

SC2/HA2

Figure 1. Experimental Design Participants We enlisted 64 self-selected undergraduate and graduate-level students from classes, assigning eight participants to each questionnaire. Procedure We asked participants to take our study online at their own conveniences within a two-week time frame. When each participant began the study, he/she was asked to close all other windows on the computer to prevent distractions. The participant was then shown one story/ad combination at a time, followed by a

series of questions pertaining to emotions and liking, first about the story and then the advertisement next to it.

A second variable for which we found significant results was a person’s likelihood to recommend the content to others. Content Ad Recommend Recommend F p F p Content 13.21 .001 0.00 .99 Ad 3.91 .05 1.53 .22 Content*Ad 3.95 .05 0.18 .68 HH Mean 3.49 2.63 HS Mean 5.94 2.43 SH Mean 4.97 2.73 SS Mean 4.94 2.32 Table 2: Recommend the Content to Others

Two days after completing the initial questionnaire, each participant was emailed the link to a follow-up questionnaire with questions testing memory of the stories and advertisements. They were asked to select the topics they had read about, the types of products and services advertised, and the brand names of the products and services from a multiple choice list. Analysis After gathering data from all 64 participants, we grouped responses based on emotion combination, not actual stories or ads. For example, all happy content/happy ad combinations were grouped with the other happy content/happy ad combinations, regardless of topic.

As displayed in Table 2, the content is affected by the emotion of the content and the emotion of the ad, as well as an interaction between the two.

A factor analysis of the questionnaire suggested a single factor. Good content was an index of enjoyable, interesting, educational, and interested in the topic.

The two main effects are an artifact of the interaction. The key result is that happy content is much more likely to be recommended to others when it is juxtaposed with a sad ad as compared to a happy ad, consistent with the previous result.

RESULTS

After running a multivariate linear regression on our data, we found two main domains that were influenced by the emotional juxtaposition. First, there were main effects for good content. Content Good F p Content 49.48 .001 Ad 5.80 .02 Content*Ad 1.97 .17 HH Mean 24.58 HS Mean 26.67 SH Mean 21.85 SS Mean 22.34 Table 1. Good Content

There were no predictors for recommendations of the advertisements. As seen by both variables, sad ads improved response to the content, particularly in the case of happy content. Table 3 below shows the significant effects caused by happy and sad ads on happy content.

Ad Good F p 0.83 .37 0.94 .34 0.94 .34 12.01 12.85 11.88 11.88

Variable 1 HHCgood HHCrecommend

Variable 2 HSCgood HSCrecommend

Effect Ad Ad

F 6.75 8.06

p .01 .01

Table 3. Happy Content Comparisons Interestingly, we did not find any significant effects of emotional juxtaposition on the recall of news stories, ads and products two days later

Both the emotion of the content and the ad significantly affected the perceived quality of the content. Happy content was perceived as better than sad content, but this could be an artifact of the actual stories. The more striking result is that participants liked the content better when it was juxtaposed with sad ads rather than happy ads. There was no interaction. The quality of the ad was not predicted by either variable nor by the interaction.

DISCUSSION

While we found that the emotional juxtaposition created by the advertisement affected the subjects’ perceptions of the news story, our hypothesis was incorrect that the juxtaposition would affect participants’ impressions and recollection of the advertisements as well.

3

These results are certainly interesting. They suggest that user perception of web content is affected by the advertisements that surround it. In an industry that is dominated by advertising revenue, web designers should consider these effects when placing advertisements, especially because this emotional juxtaposition changes the probability that a person recommends web content. It should be noted that sad advertisements helped the content to perform far better than happy advertisements did. This suggests that it is in the best interest of a website to post sad ads next to content. However, websites displaying sad content do not need to be as concerned with the emotion of nearby advertisements as those with happy content. While containing what we defined as Good content may or may not be necessary to a website’s survival, our second effect, Recommend is hugely important— and is much more affected by the emotional juxtaposition of advertisements.

emotions when subject were asked questions about the news stories and advertisements two days later. In order to determine whether there is really no effect of emotional juxtaposition in online advertising on memory, one would have to run the same study using varying time lengths for the follow-up. Further studies should also be conducted to evaluate the effects of aspects other than advertisements on web content. For example, if a website has happy content, but a sad color scheme, does this color scheme affect the content in the same way a sad ad does? It would also be important to test the effects of happy and sad ads on emotionally neutral content. In terms of advertisements, it would be interesting to repeat this study with more complex advertisements, forcing the participant to spend the same duration of time focusing on the ad as on the content. CONCLUSION

This is also something that is important for the web user to understand. As technology improves, it can better pinpoint the emotions caused by web content. For example, it is not difficult to imagine a program that could evaluate whether an email bears good or bad news and target advertisements accordingly. As these surrounding advertisements have the ability to affect the user’s perception of an email, it is also important that users understand these effects in order to react appropriately.

In conclusion, this study firmly demonstrates an important emotional link between original content and the advertisements that surround it. Sadnessprovoking advertisements have the distinct ability to significantly improve perceptions of content and the likeliness that it is shared. In an Internet generation when many sites live or die on the virality of their content, this is something to which domain owners need to begin to pay attention.

Another interesting finding that this study reveals is the small amount of emotional manipulation needed in order to alter perceptions of content. Past emotion studies have used high levels of emotion manipulation (e.g., showing subjects video clips in the case of Nass and Reeves).

REFERENCES

Though we can assume that the participants in our study spent the vast majority of their time reading the news stories and only a few seconds glancing at the ads, those small ads on the side of each page were able to significantly change perceptions of the content. Finally, we must note that our hypotheses on content and ad recollection proved to be incorrect. Inconsistent with other studies, the matching of emotions did not perform any better than mismatching

1.

Bower, G. H., Gilligan, S. G., & Monteiro, K. P. (1981). Selectivity of learning caused by affective states. Journal of Experimental Psychology: General, 110, 451-473.

Brave, S. & Nass, C. (2002). Emotion in humancomputer interaction. In J. Jacko & A. Sears (Eds.), Handbook of human-computer interaction (pp. 251-271). Hillsdale, NJ: Lawrence Erlbaum Associates. 3. Marcus, A. "The Emotion Commotion." Interactions (2003): 28-34. 4. Nass, C. & Reeves B. Media equation how people treat computers, television, and new media like real people and places. New York: Cambridge University Press. 2.

5