Using Facial Electromyography to Test the Peak–End Rule in Tourism Advertising

Emotion in Tourism Advertising

Bagozzi, Gopinath, and Nyer (1999) define emotion as a psychological state resulting from cognitive appraisals of events or environments, which is accompanied by physiological responses and expressive behavior. Advertisement-elicited emotion is regarded as having the same level of importance as the conscious and rational thoughts consumers may have when viewing advertisements (Micu & Plummer, 2010). Marketers use various strategies to evoke consumers’ positive emotional responses to engage viewers with the advertisements (Teixeira & Wedel, 2012).

Two prominent approaches to describing individuals’ emotional responses are frequently used in advertising: the basic emotion and dimensional approaches. The basic emotion approach includes several subtheories (most based on cross-cultural studies) that identify a finite set of discrete emotions (Lzard, 1992; Plutchik, 1982). For instance, Plutchik (1980) identifies eight basic emotions which includes surprise, expectancy, disgust, acceptance, sadness, joy, anger, and fear. He believes that more complicated emotions are mixtures of these basic or primary emotions.

A second approach—the dimensional approach—identifies a set of common dimensions of affect that differentiate specific emotions from one another. For example, pleasure and arousal are the two main dimensions applied in describing consumers’ emotional responses to advertising (Holbrook & Batra, 1987; Olney, Holbrook, & Batra, 1991). The dimensional approach is considered more parsimonious than the basic emotion approach, because it classifies emotions along two dimensions: pleasure and arousal (Mauss & Robinson, 2009). Furthermore, emotion in advertising is often short lived and rarely seen in its pure form, thus the basic emotion approach focusing on discrete emotions is problematic (Huang, 2001). Therefore, the dimensional approach is adopted in this study to describe consumers’ emotions in response to tourism advertising.

Past advertising research demonstrates that ad-evoked emotion is positively related to various post-viewing global evaluative responses, such as attitudes toward the advertisement and brand (Edell & Burke, 1987), brand familiarity (Mano, 1996), brand choice (Stayman & Batra, 1991), brand attribute evaluations (Burke & Edell, 1989), attention to the advertisement (Olney et al., 1991), and recall (Pieters & de Klerk-Warmerdam, 1996). Recently, ad-evoked emotion has taken a central role in research on individual attitudes toward safe driving. Hamelin et al. (2017) used advertisements featuring messages about the importance of safe driving to demonstrate that consumers who watched the highly emotive versions presented greater safe-driving attitude scores compared with those who viewed the less emotive versions.

In tourism research, the majority of research to date has focused on identifying the most effective visual stimuli in an advertisement (e.g., Chou & Lien, 2012; Luoh & Lo, 2012; Wang, Hsieh, & Chen, 2002). For instance, Luoh and Lo (2012) investigated the influence of gender when exploring the effectiveness of chef endorsements in restaurant advertisements. Their experiment reveals that a middle-aged male chef is more likely to convince consumers than a middle-aged/young female chef. However, this research does not explain the psychological mechanism through which consumers processed the advertising messages (Li et al., 2016). Walters and Sparks (2012) found that the emotions evoked by tourism advertisements are able to predict consumers’ purchase interest and intentions. However, ad-evoked emotions were measured by self-report scales, which are not able to capture the dynamic process of tourism consumers’ moment-to-moment emotions. Li, Walters, Packer, and Scott (2017) further investigated the role of emotion in predicting tourism advertising effectiveness and found that pleasure, rather than arousal level, was positively related to consumers’ attitudes and behavioral intentions.

In summary, existing studies have shown the pivotal role ad-evoked emotion plays when explaining consumers’ post-viewing judgments. While recent studies have demonstrated the importance of ad-evoked emotion in influencing advertising evaluation, extant research does not examine consumers’ moment-to-moment affective viewing experience, particularly the impact of the key moments of their affective watching experience on retrospective evaluations.

The Peak–End Rule

The peak–end rule posits that individuals base their retrospective evaluation of an experience or event on the peak moment (the most affectively intense moment of an experience) and the final moment (Kahneman et al., 1993), rather than on total amount of pain or pleasure experienced. Due to memory instability, only a limited information was available for individuals to evaluate their entire experience at any given recall time, it is possible that this limited information was dominated by the peak and end moments (Kemp, Burt, & Furneaux, 2008). Previous research has confirmed the peak–end rule across various situations with negative stimuli such as aversive sounds (Schreiber & Kahneman, 2000) and pain (Kahneman et al., 1993). Kahneman et al. (1993) have also demonstrated the robustness of peak–end rule with pleasant stimuli (i.e., film clips of playful penguins; Fredrickson & Kahneman, 1993). Research has also shown that neither the total duration of an event or experience (Kemp et al., 2008; Varey & Kahneman, 1992) nor the number of the peaks (Schreiber & Kahneman, 2000) is important when an individual is constructing an overall judgment of an experience.

In the field of marketing, Baumgartner, Sujan, and Padgett (1997) used a computerized “feeling monitor” to track viewers’ real-time “liking” of an advertisement and found that the peak and end moments were related strongly with the overall evaluation that occurred after the consumer had viewed the advertisement. Thomas et al. (2018) found additional support for the peak–end rule in the context of financial decision making and charity advertisements. The peak–end rule was also found to be applicable to the understanding of the mechanisms that form consumers’ reference prices (De Maeyer & Estelami, 2013).

However, some studies have reported deviations from the peak–end rule. In attempting to locate the “key selling second” of an advertisement, Polsfuss and Hess (1991) found that the overall liking of the advertisement was more closely associated with the average rating scores than with peak or end moments. Hui et al. (2013) found that the role of specific patterns, such as peak moments in driving the overall judgment, is limited in the context of TV shows. Rather, the final segment was found to have a greater effect on the overall evaluation. Asking participants to report their feelings from the previous day, Miron-Shatz (2009) found that the average, rather than the peak or end segment, was the best indicator of retrospective evaluations of the day. Similarly, the power of peak moments in predicting the overall evaluative outcome in the context of video games was not confirmed in the work of Mukherjee and Lau-Gesk (2016). Rather, they found that end moments were correlated with the evaluation, especially when the end sections were meaningful.

These conflicting evidences could be partially attributable to the measurements used to capture emotions. The majority of studies to date have measured moment-to-moment responses via a self-report method, although the specific measure may vary (Geng, Chen, Lam, & Zheng, 2013; Kemp & Chen, 2012; Kemp et al., 2008). For instance, Mukherjee and Lau-Gesk (2016) asked game players to report their feelings from the beginning to the end of the arcade gaming session by drawing a continuous line. This line could be further used to showcase participants’ emotional experience. In contrast, Kemp et al. (2008) used Likert-type scale (1 = extremely unhappy to 9 = extremely happy) and asked participants to recall and rate their emotions at different times (e.g., current, best, or worst moments). In essence, most of the research in this area has relied on participants’ reconstruction of their completed experience, despite the fact that participants may not have been able to remember the details of their experience accurately. This measurement issue needs to be addressed before further testing the peak–end rule in various contexts.

The traditional self-report method, in which participants have to rely on their memory to recall their affective experience or feelings, has been widely criticized (Brodien Hapairai, Walters, & Li, 2018; Kim & Fesenmaier, 2015; Li et al., 2016). Inevitably, the recollection of the emotional experience may involve a series of cognitive biases, which could distort an individual’s original emotion (Ravaja, 2004). As noted by Kemp et al. (2008), “Ordinary people may recall experiences quite inaccurately or in a way that has been greatly distorted” (p. 133).

Psychophysiological measures provide an alternative measurement technique that involves attaching sensors to participants and recording their emotional responses in real time. Psychophysiological measures of emotions are based on individuals’ automatic body responses that are largely beyond their conscious control, and thus they are less likely to be influenced by cognitive bias (Kim & Fesenmaier, 2015). Possible measurements include facial electromyography (EMG), skin conductance, and heart rate measures.

Compared with traditional questionnaire methods, one of the advantages of psychophysiological measures is continuality. Emotions are continuous and dynamic, and the psychophysiological measures enable consumers’ moment-to-moment emotions to be tracked, which in turn can be used to examine the emotional flow pattern (Li et al., 2016). Practically, these real-time emotional curves can be used to fine-tune advertisements (Hui et al., 2013).

Psychophysiological Measures of Emotions in Tourism

Given the criticism of self-report measures of emotion, recent research in the field of tourism investigated tourism consumers’ moment-to-moment responses via psychophysiological measurements. For example, Kim and Fesenmaier (2015) measured travelers’ emotions in natural settings based on skin conductance and discussed the potential of using psychophysiological measures in designing and managing tourist experiences. Kim et al. (2014) focused on tourism consumers’ skin conductance and heart rate responses to two types of advertising communications (video vs. high-imagery audio advertisements) and found that the arousal level evoked by video was higher than that elicited by audio advertisements. Li and his colleagues tested the capabilities of facial EMG and skin conductance in tracking consumers’ real-time emotional responses to tourism advertisements (Li et al., 2016), and further explored the relationship between ad-evoked emotional responses and tourism advertising effectiveness using both self-reporting and psychophysiological measures (Li et al., 2017).

While psychophysiological measurements can measure moment-to-moment emotional responses, previous research adopting such measures in tourism and marketing has typically used an average score as a proxy measure to indicate consumers’ emotional states and examine the effects of emotions on other advertising measures such as attitude toward advertisement (Kim et al., 2014; Li et al., 2017; Ravaja & Somervuori, 2013). This is problematic, because both a flat emotion pattern and emotion curves with positive or negative slopes could generate the same mean score (Hughes, 1992), even though consumers may not evaluate them equally.

This study used facial EMG to track consumers’ moment-to-moment positive emotional responses to tourism advertisements, with an emphasis on the effects of key emotion moments on retrospective global evaluations. Facial EMG is the modern measurement of individuals’ subtle facial muscle contractions (Bolls, Lang, & Potter, 2001). Facial EMG measures the electrical signal produced by the occurrence of action potentials across different facial muscle groups. In particular, the experience of positive emotions is closely related to the activation of the zygomatic muscle group (around the cheek), and negative emotions are indicated by contractions of the corrugator muscle group (Lang, Greenwald, Bradley, & Hamm, 1993). Li et al. (2016) have demonstrated the capabilities of facial EMG in tracking tourism consumers’ positive emotional responses to tourism advertisements. Tourism produces hedonic consumption experiences and involves fantasy, feelings, and fun (Williams, 2006); hence, it is appropriate that the current study focuses on positive emotional responses.

Research Model and Hypothesis Development

Based on the peak–end rule, we investigated the effect of ad-evoked moment-to-moment emotions on consumers’ post-viewing global evaluations in the context of tourism advertising. Consistent with previous research (e.g., Chang, Wall, & Tsai, 2005; van der Veen & Haiyan, 2014; Luoh & Lo, 2012), this study measured post-viewing global evaluations using attitude toward the advertisement (A_ad), postexposure brand attitude (A_bp), and purchasing intention. The majority of existing research adopting self-report measures of emotion has confirmed the direct effect of ad-evoked emotions on A_ad (Pieters & de Klerk-Warmerdam, 1996), A_bp (Wells, 1989), and purchasing intention (Shahin Sharifi, 2014). The indirect effect of ad-evoked emotions on purchase intention via A_ad and A_bp has also been established (Batra & Ray, 1986; Pelsmacker & Geuens, 1998). In the context of tourism, postexposure brand attitude is represented as postexposure destination attitude (A_dp) whereas purchase intention is often represented as visit intention (VI). The model to be evaluated in this study is illustrated in Figure 1.

OPEN IN VIEWER

Figure 1

The Research Model

As stated above, when consumers make global evaluations of an emotional tourism advertisement, their judgments are affected by their emotions in response to the advertisement. The peak–end rule posits that the peak and end moments are key emotional moments and are more predictive of consumers’ post-viewing evaluation than a temporal integration or sum of the affective values associated with a tourism advertisement. Given that tourism advertisements are more likely to incite positive emotions, the following hypotheses are centered on pleasant, rather than aversive, advertisements:

Participants

We conducted this research in a research laboratory located in an Australian university. One hundred and one university students and staff volunteered to be the participants in this study in exchange for an AUD10 gift voucher. A small sample is common in studies involving psychophysiological measures (Ravaja & Kallinen, 2005; Ravaja & Somervuori, 2013) and did not present any validity issues concerning the necessary analysis techniques. Demographic characteristics of the samples are shown in Table 1.

OPEN IN VIEWER

Table 1

Demographic Characteristics of the Participant Samples

Item	Total number	%
Gender
Male	46	45.5
Female	55	54.5
Age (years)
18-24	45	44.6
25-34	32	31.7
35-44	9	8.9
45-54	11	10.8
>55	4	4
Educational level
High school	36	35.6
Bachelor’s degree	39	38.7
Master’s degree	19	18.8
Doctoral degree	6	5.9
Others	1	1
Prior experience with the TV commercials
Have seen before	4	4
Have not seen before	97	96
Prior experience with the destination
Have visited before	26	26
Have not visited before	75	74

Stimuli

The stimuli consisted of three 60- to 90-second destination advertisements from www.myswitzerland.com, the website of a Switzerland-based destination marketing organization. These advertisements use the same actors, but each of the advertisements present Switzerland as offering a distinct tourism experience ranging from a city escape, to perfect winter holiday destination, to a place to spend one’s summer break. The hypotheses in this study rely on the evocation of positive emotions, and the Switzerland tourism promotion advertisements present a significant element of humor and were therefore deemed appropriate for the purposes of this study.

Emotion Measures

Viewers’ positive emotional responses to the advertisements were collected and recorded using a facial EMG monitor that required the attachment of electrodes to participants’ zygomatic major muscles. A trained researcher, proficient in the use of facial EMG, carefully located the recording sites on each participant’s face in accordance with published facial EMG guidelines (e.g., Fridlund & Cacioppo, 1986). Consistent with the work of Baumgartner et al. (1997), we used the last second of each advertisement as the indicator of final moment. The peak moment was represented by the maximum of the facial EMG trace. The mean facial EMG score was used as the indicator of average affective intensity in this study (Boxtel, 2001).

Global Retrospective Evaluation

The scales measuring A_ad, A_dp, and VI were adapted from prior research, as shown in Table 2. The authors considered the theoretical foundations of one question compared with multiple questions to capture a construct (Dolnicar, 2013; Rossiter, 2011) and concluded that using a single question was suitable for the research aims and additionally reduced respondent fatigue.

OPEN IN VIEWER

Table 2

Measures of Global Retrospective Evaluations and Preexposure Measures of Covariates

Measures	Items	Reference
The attitude toward the ad	How would you describe your overall attitude toward the ad you have just watched? (1 = I don’t like this ad at all to 7 = I like this ad very much)	Derbaix (1995), Morris et al. (2002)
The attitude toward the destination	Based on this ad, how positive is your impression of Switzerland as a tourism destination? (1 = not at all positive to 7 = very much)	Kim and Stepchenkova (2015)
Visit intention	Does this ad make you want to visit Switzerland? (1 = not at all to 7 = very much)	Kim and Stepchenkova (2015)
Preexposure destination attitude	How positive is your impression of Switzerland as a tourism destination? (1 = not at all positive to 7 = very much)	Kim and Stepchenkova (2015)
Mood states	“Currently, I am in good mood”; “As I answer these questions, I feel cheerful”; “At this moment, I feel edgy or irritable”; “For some reason, I am not very comfortable right now” (1 = strongly disagree to 7 = strongly agree)	Peterson (1983)

Covariates

Participants’ moods and preexposure attitude toward the destination (Switzerland) were included to control for their effect on dependent variables.

Procedure

Participants were tested one at a time. Before the commencement of the experiment, each participant indicated their current mood and existing attitude toward Switzerland via a questionnaire. A practice video (an unrelated tourism advertisement) was first shown to the participants to allow them to become familiar with the procedure, while simultaneously allowing the investigator to check the quality of the facial EMG data. To reduce the risk of participant fatigue, each participant was randomly assigned to one of the three Switzerland tourism promotion advertisements. After watching the tourism advertisement, each participant completed a questionnaire regarding their A_ad, A_dp, and VI. Demographic information such as gender, age, and prior experience with the advertisement or destination was also collected.

Data Analysis

AcqKnowledge^® software was used to analyze facial EMG data. To evaluate the structural model, the partial least squares (PLS) path modeling method (Hair et al., 2012) was used with SmartPLS 3.0 software (Ringle, Sven, & Becker, 2015). The PLS path modeling method has the advantage of being able to handle small sample sizes and process data that are not normally distributed. The PLS was thus considered more suitable than the commonly used covariance-based structural equation modeling technique for this study. To compare the differences between two correlation coefficients that involve a common variable, Steiger’s z test was used with an online calculator (Hoerger, 2013; Steiger, 1980).

Scale Reliability and Validity

Mood is the only construct that has multiple items in this study. The value of the average variance extracted was 0.726, which was well above the minimum threshold of 0.50 (Henseler, Ringle, & Sinkovics, 2009). Thus, the convergent validity of the mood scale was established. Indicator reliability was evaluated by checking indicator loadings. Four items measuring mood (0.979, 0.897, 0.788, and 0.721) showed significant standardized loadings above 0.70, p < .01. The high Cronbach’s alpha and composite reliability (ρ) values (all >0.80) also demonstrate good internal consistency. The value of the square root of average variance extracted for the mood scale was 0.852, which is greater than the correlation values between mood and other constructs. Therefore, the discriminant validity was also well established.

Evaluation of the Structural Model (Hypotheses Testing)

This section reports the results separately for the structural model with average emotion intensity, peak emotional experience value, and final emotional experience value as the independent variables. To test the hypotheses proposed earlier, a nonparametric bootstrapping analysis (5,000 subsamples and 101 cases) was used to calculate the t values and significance level. The outcome of the structural model test using different emotion indexes is shown in Table 3.

OPEN IN VIEWER

Table 3

Results of Hypothesis Testing With Different Emotion Indexes

Hypothesis	Emotion Index	Path Coefficient	t a	Supported
Hypothesis 1: Emotions → Aad	Average	0.252	2.447*	Yes
	Peak	0.266	3.505**	Yes
	End	0.225	3.047**	Yes
Hypothesis 2: Emotions → Adp	Average	ns	0.441	No
	Peak	ns	0.086	No
	End	ns	1.263	No
Hypothesis 3: Emotions → VI	Average	ns	1.410	No
	Peak	ns	0.009	No
	End	ns	0.841	No
Hypothesis 4: Emotions → Aad → Adp → VI	Average	0.185	2.086*	Yes
	Peak	0.204	2.767**	Yes
	End	0.2	3.1**	Yes

Note: A_ad = attitude toward the advertisement; A_dp = postexposure attitude toward the destination; VI = visit intention; ns = not significant.

a.

We applied a nonparametric bootstrapping route to test the significance of the partial least squares path modeling results.

*

p < .05. **p < .01 (two-sided test; number of observations = 101).

Average Emotion Intensity

Figure 2A displays the outcome of the structural model when positive emotions were represented by average emotion intensity. The PLS path model estimation provided an R² value of .083 for A_ad, .661 for A_dp, and .715 for VI, respectively. In support of Hypothesis 1a_a, we found that the average emotion intensity had a significant positive impact on A_ad. However, the direct effects of emotions on A_dp (Hypothesis 2a_a) and VI (Hypothesis 3a_a) are not supported. To test Hypothesis 4_a, we conducted a mediation analysis using a nonparametric bootstrapping procedure. In support of Hypothesis 4_a, we found that A_ad and A_dp fully mediated the effect of emotions on VI.

OPEN IN VIEWER

Figure 2

Partial Least Square Result of the Structural Models With Different Emotion Indexes

Peak Emotional Experience

Figure 2B displays the outcome of the structural model test when positive emotions were indexed by peak emotional experience. The PLS path model estimation provided an R² value of .09 for A_ad, .66 for A_dp, and .737 for VI. Hypothesis 1a_a received strong support when emotional responses were represented by the peak moment value. However, the direct effects of emotions on A_dp (Hypothesis 2a_a) and VI (Hypothesis 3a_a) are not supported. Ad-evoked emotion had an indirect impact on VI via A_ad and A_dp, which supports Hypothesis 4_a.

Final Emotional Experience

Figure 2C displays the outcome of the structural model test when positive emotions were represented by final emotional experience. The PLS path model estimation provided an R² value of.07 for A_ad, .662 for A_dp, and .739 for VI. We found a significant impact of emotion on A_ad, which supports Hypothesis 1a_a. Furthermore, Hypothesis 2a_a and Hypothesis 3a_a positing a direct effect of emotion on A_dp and VI, respectively, are not supported. The results support Hypothesis 4_a that A_ad and A_dp mediate the impact of emotion on VI.

The Peak–End Rule

According to the Steiger’s z test, we compared the correlation coefficients between peak moment/average and A_ad and yielded a z value of −0.246, p = .805, which indicates the correlation coefficient of peak moment and A_ad is not significantly higher than that of average score and A_ad. Similarly, we compared the correlation coefficients between end moment/average and A_ad and obtained a z value of 0.405 with p = .68. These results reveal that the peak and end moments are no better than the average emotion intensity in predicting the attitude toward the advertisement when more objective emotion measurements were employed. Therefore, Hypothesis 1b_a, Hypothesis 2b_a, and Hypothesis 3b_a are rejected.

Practical Implications

Our findings have direct implications for tourism destination marketers seeking to make effective advertisements and maximize viewers’ overall evaluations. The results of this study show that the peak and end emotional experience of an advertisement are as important as the overall emotional experience. Thus, producers may optimize resources at critical points, such as the end of an advertisement, or create a peak experience (e.g., exciting or funny scene or plots) for the viewers. Facial EMG and other psychophysiological measures provide an appropriate measurement tool to test if a peak experience can be evoked among potential tourists.

While creating a peak emotional experience is important, advertisements should bolster personal meaning to coincide with these key moments in order to achieve the most favorable evaluations (Mukherjee & Lau-Gesk, 2016). In tourism marketing, the peak moment should not only be humorous but also reflect what the target audience really desire when seeking a destination for their holidays. For instance, Kim, Guo, and Agrusa (2005) found that mainland Chinese tourists considered “beautiful scenery” and “safety” as most important factors when choosing an overseas destination. Therefore, destination marketers could integrate these elements into their consideration when designing advertisements targeting Chinese tourists.

Our findings also indicate that the peak and end emotional experiences are no better than an average emotion intensity when evaluating “overall” attitude toward an advertisement. Average emotions reflect participants’ overall affective impression of the stimuli and can be used as a proxy to predict their retrospective evaluation (Miron-Shatz, 2009). Thus, the destination advertisement maker may also need to improve the quality of the entire advertisement instead of only specific moments. In addition to evoking a memorable peak experience among potential viewers, the designer should create an innovative advertisements story line to keep viewers stay focused while watching or design the best scene change in a video sequence through pretesting.

Limitations and Recommendations for Future Research

This study has certain limitations. First, we focused on tourism consumers’ moment-to-moment positive emotions, and thus only included humorous advertisements as the stimuli. As noted by Kemp et al. (2008), it is still not clear whether the peak and end rule will apply in the mixed experience that involve both positive and negative emotions. Thus, it may be insightful for future research to test the applicability of the peak–end rule in other types of tourism advertisements (e.g., those advertisements adopting adventure, family, and romance appeals) that are more likely to produce mixed emotions.

Second, the small sample in this study may, to some extent, limit the decisive influence of the peak and end moment in driving tourism consumers’ post-viewing attitude. Future studies could enlarge the sample size and test the influence of the key moments in shaping tourism consumers’ retrospective judgments.

Third, this study adopted the dimensional approach to describing tourism consumers’ emotional responses to the advertisements. While consumers’ pleasure level was measured by facial EMG in this study, their arousal level was not measured. This is due to the consideration of initial results derived from the pretesting stage: All of three humorous advertisements were demonstrated to be capable of evoking viewer’s peak emotional experience (indicated by a significant increase in their facial EMG trace), which provided an ideal condition for testing peak and end rule in the context of tourism advertising. However, we found that it was hard to identify viewer’s peak moment with their arousal data measured by skin conductance device. It is possible that these advertisements were designed to elicit viewer’s pleasantness rather than excitement. Therefore, future studies could select tourism advertisements with adventure or excitement appeals and examine the applicability of the peak–end rule in testing the effect of moment-to-moment arousal on consumers’ post-viewing evaluations. Eye tracking measures may also be combined with emotion measurements to identify the specific stimuli viewers focus on when their emotion responses are elicited.

Fourth, individuals only have a limited amount of information when asked to recall or evaluate their overall experience, and this limited information may include peak and end experiences, which was the rationale of the peak and end rule proposed by Fredrickson and Kahneman (1993). However, this limited information may also include the beginning experience or the most unusual experience (Kemp et al., 2008). Future research could further explore which particular aspect of consumers’ moment-to-moment emotional responses could best explain their post-viewing global evaluation.

Finally, the participants in this study were instructed to focus on the stimuli from start to finish. However, consumers in real life may turn their attention to other tasks (e.g., check their phones or talk to their friends or families) while watching the advertisements. Future research may replicate our study in a more natural environment and further explore the impact of selective exposure and attention on the correlation between ad-evoked moment-to-moment emotions and evaluative responses to advertising.