Music represents a key element of modern campaign ads. In fact, many candidates use either upbeat or melancholic music in some of their campaign ads because they believe that music—like images—can shape individuals’ perceptions regarding, for instance, their competence or compassion (Brader, Reference Brader2005; Holtz-Bacha & Just, Reference Holtz-Bacha and Just2017). However, is this belief empirically valid? That is, does music affect trait perceptions? Answering this question is key because it would further our knowledge of where trait perceptions—which represent important determinants of the vote (Campbell et al. Reference Campbell, Converse, Miller and Stokes1960; Funk & Carolyn, Reference Funk1999; Hayes, Reference Hayes2009)—come from. Moreover, considering that music may be used to manipulate public opinion and push individuals to vote for a candidate who does not reflect their best interests, it is crucial that we understand whether music is effective, or not, at shaping trait perceptions.
This article investigates the relationship between music and trait perceptions. It does so using data from an online survey experiment conducted between October 30 and November 12, 2020. In this experiment, individuals were exposed to a random sequence of five campaign ads, either with their original music or with no music. Following each campaign ad, individuals were asked a battery of questions tapping their perceptions of the five candidates on the four most relevant trait dimensions (competence, honesty, compassion, and leadership) (Geer, Reference Geer2006; Kinder, Reference Kinder, Lau and Sears1986).Footnote 1 , Footnote 2
In line with prior work (Thorson et al., Reference Thorson, Christ and Caywood1991), the results suggest that music marginally affects perceptions of competency, honesty, and leadership, implying that individuals are not much more likely to perceive a candidate as competent, honest, or a strong leader when they are exposed to campaign ads with their original music than when they are exposed to the same ads with no music. Moreover, music exerts no significant effect on the degree to which individuals perceive a candidate as compassionate. Importantly, these results hold when we look at the effect of music on trait perceptions with data from each campaign ad: in only 3 out of 20 cases, the relationship between music and trait perceptions reaches statistical significance.
With these findings, this study advances knowledge on the relationship between music and trait perceptions. More specifically, it offers evidence that, unlike images, music may not represent a source of trait perceptions. Therefore, this study suggests that music may not serve to manipulate public opinion and, in this sense, push individuals to vote for a candidate who does not reflect their best interests.
Does music affect trait perceptions?
Within the psychology and political science fields, there is indirect evidence that music can affect trait perceptions. In fact, prior work in those fields showed that music can affect emotions and, in turn, emotions can influence trait perceptions. Beginning with the first causal link (between music and emotions), research showed that Mahler’s and Mozart’s music can generate a strong emotional reaction in its listeners (Niedenthal & Setterlund, Reference Niedenthal and Setterlund1994) (see also the work of Baumgartner et al., Reference Baumgartner, Esslen and Jäncke2006; Suguna & Deepika, Reference Suguna and Deepika2017; Västfjäll, Reference Västfjäll2002). Research revealed as well that brain regions normally associated with emotional activity—the amygdala and the hypothalamus—are both activated with exposure to music (Sachs et al., Reference Sachs, Damasio and Habibi2015), a reaction that emerges in only a few seconds (Bachorik et al., Reference Bachorik, Bangert, Loui, Larke, Berger, Rowe and Schlaug2009). Finally, comparing different types of emotion stimuli, Krumhansl (Reference Krumhansl1997) found that music not only induces emotions but does so more strongly than images.
With respect to the second causal link (between emotions and trait perceptions), prior work provided evidence that emotions affect the extent to which individuals seek new information. More precisely, when individuals feel enthusiastic in a given context, they are likely to rely on habitual responses. By contrast, when they feel fear, they are likely to “stop ongoing action, shift attention to new stimuli (and) uncouple reliance on habit” (Marcus et al., Reference Marcus, Neuman and MacKuen2000, p. 54). In addition to triggering a process of information seeking, research showed, emotions also affect how individuals evaluate a candidate. More specifically, Isbell and Wyer (Reference Isbell and Wyer1999) and Yamasaki et al. (Reference Yamasaki, Yamada and Laukka2015) showed that exposure to a happy stimulus results in more favorable evaluations of a candidate than exposure to a sad stimulus or no stimulus. Following this body of work—specifically, the evidence on the connection between music and emotions, and between emotions and traits perceptions—we derive the following hypothesis: music can affect individuals’ trait perceptions.
To the best of our knowledge, one study tested the connection between music and trait perceptions empirically: Thorson, Christ, and Caywood (Reference Thorson, Christ and Caywood1991). These scholars examined the impact of music on trait perceptions by means of a lab experiment with 161 undergraduate students from the communications department of a large midwestern university. In this experiment, those in the treatment condition watched four senatorial ads with an upbeat, voiceless, patriotic song.Footnote 3 By contrast, those in the control condition watched the same ads with no song. The authors observed no statistically significant difference in how candidates were perceived by the treatment and the control groups, which led them to conclude that music bears no influence on trait perceptions.
While we acknowledge the important contribution of Thorson, Christ, and Caywood (Reference Thorson, Christ and Caywood1991) to the understanding of whether and the degree to which music can affect trait perceptions, there is good reason to be skeptical about their null findings and to believe that more research is needed before any firm conclusion is made on the topic. More precisely, one of the key problems in their study is that the authors did not use the original music from the ads as the treatment. As a result, individuals may not have perceived the treatment as realistic because of the incongruence between the music and the images, affecting the overall validity of the experiment.
The experiment
This study examines the relationship between music and trait perceptions by means of an online survey experiment conducted between October 30 and November 12, 2020, with either the original music from campaign ads (the treatment condition) or no music (the control condition) (see Appendix A in the supplementary material online for an overview of the experimental design and Appendix B for the experimental protocol).Footnote 4 Following each campaign ad, individuals in both treatment and control conditions were asked to evaluate the candidates regarding their competency, honesty, leadership, and compassion.Footnote 5 Importantly, this experiment was conducted in Canada, where individuals feel less attached to a political party and are more likely to switch their vote between elections (Bischoff, Reference Bischoff2013). While examining the relationship between music and trait perceptions in highly partisan and volatile contexts seems relevant, it adds a complexity to the study of that relationship that is beyond the scope of this study.
The experiment was conducted with 362 individuals (yielding 1,810 observations) who were recruited by means of Prolific, based on four criteria: (1) absence of a hearing or a visual impairment, (2) good internet connection and a functional audio and visual interface, (3) no affiliation with a political science department (either as a student or as a professor), and (4) fluency in French.Footnote 6 , Footnote 7 , Footnote 8 The first two criteria (absence of a hearing or a visual impairment, and good internet connection and a functional audio and visual interface) were established because being able to see the candidates and hear the music in the campaign ads is imperative in the context of this study. By contrast, the third and the fourth criteria (no affiliation with a political science department, and fluency in French) were established to reduce the risk of prior knowledge of the candidates. The 362 individuals who satisfied those four criteria were paid $5.90 Canadian dollars each, which corresponds to almost half (47 percent) of the minimum wage in the Canadian province of Quebec.Footnote 9,Footnote 10
To avoid potential biases in the treatment estimates, the experiment included ads featuring only non-Canadian politicians (who should be unknown to participants) and containing no party references. Moreover, the ads were played in a random order to reduce the risk of between-treatment contamination. Finally, in line with Kroodsma (Reference Kroodsma1990) and Machlis et al. (Reference Machlis, Dodd and Fentress1985), the ads contained different music styles as well as different times of exposure to the candidates and to the music. More precisely, Ben Carson’s ad included an original piano-based, voiceless, melancholic song, and Carson appeared in 9 seconds of a 59-second video; Rick Santorum’s ad included an original guitar-based, voiceless, upbeat song, and Santorum appeared in 51 seconds of a 1-minute, 4-second video; Pete Buttigieg’s ad included an original electronic, voiceless, upbeat song, and Buttigieg appeared in 14 seconds of a 29-second video; Bill English’s included an original folk-inspired, voiceless, upbeat song, and English appeared in 27 seconds of a 1-minute, 59-second video; lastly, Jeremy Corbyn’s ad included Lilly Allen’s version of “Somewhere Only We Know,” and Corbyn appeared in 26 seconds of a 1-minute, 45-second video. Therefore, the results are unlikely to be contaminated with pseudo-replication bias.
With these elements from the experimental design in mind, the results from balance tests suggest a successful randomization of participants in either the treatment or the control group. As shown in Table 1, there is no significant difference between the two groups regarding key attitudinal and sociodemographic factors. Specifically, individuals in the treatment group and those in the control group are, on average, equivalent when it comes to the extent to which they are interested in politics, their ideological position, their level of education, their age, and their gender.Footnote 11 Therefore, the following estimations of the impact of music on trait perceptions are based on conventional two-tailed t-tests, which are conducted first with the pooled experimental data and then with ad-specific data. Importantly, the analyses exclude those who knew the candidate before the experimentFootnote 12 or did not pay sufficient attention to the campaign ads.Footnote 13
Balance tests of the experimental sample.
Notes: Results from two-tailed t-tests with participants as the unit of analysis. Differences are calculated by subtracting the mean value in the treatment group from the mean value in the control group. * p < .05; ** p < .01; *** p < .001. Standard errors in parentheses.
The results
Table 2 presents the results of the online survey experiment with the pooled data. They reveal a statistically significant difference between the treatment and the control groups in three of the four trait perceptions that are measured in the context of this study. Specifically, individuals who were exposed to campaign ads with their original music are more likely to perceive candidates as competent, honest, and a strong leader than individuals who were exposed to the same campaign ads but with no music. Yet these differences are substantially small: they vary between .02 and .03 points in a 0–1 scale.
The impact of music on trait perceptions.
Notes: Results from two-tailed t-tests. Differences are calculated by subtracting the mean value in the treatment group from the mean value in the control group. * p < .05; ** p < .01; *** p < .001. Standard error in parentheses. For the impact of the music from each campaign ad, see Appendix E in the supplementary material.
Importantly, the results are not different when we analyze differences with respect to each specific ad: in fact, in only three of the 20 ad-trait combinations, the effect of music reaches statistical significance. More specifically, those who were exposed to the music from Buttigieg’s campaign ad are more likely to evaluate that candidate positively when it comes to the competency and the honesty dimensions than those who were not exposed. By the same token, those who listened to the music from English’s campaign ad are more likely to see that candidate as compassionate than those who did not (see Appendix E). In most cases (17 of the 20 ad-trait combinations), however, music bears no influence on how individuals evaluate the candidates.
Additional tests suggest that music does not affect the candidates’ likability or participants’ willingness to vote for the candidates in a hypothetical election.Footnote 14 As shown in Appendix F, the difference between those who are exposed to music and those who are not fails to reach statistical significance when it comes to both likability and willingness to vote. Furthermore, in one of the 10 ad-trait combinations, the treatment estimate stands out as significant: as expected, those who watched Buttigieg’s campaign ad with music are more likely to vote for him in a hypothetical election than those who watched the same ad without music. Given that a candidate’s likeability and individuals’ willingness to vote for a candidate are correlates of trait perceptions, these results provide additional support to the view that music may exert a marginal (or no) influence on trait perceptions.
Discussion and conclusion
This study explored a common assumption in politics: that music can shape the way in which candidates are perceived by the public. To this end, it relied on an online survey experiment exposing participants to a random sequence of five campaign ads, either with their original music or with no music. This study showed that music marginally affects individuals’ perceptions of candidates on three trait dimensions: competence, honesty, and leadership. It also showed that music exerts no significant effect on perceptions of compassion.
Why does music exert a marginal (or no) effect on trait perceptions? While providing an answer to this question is out of the scope of this study, we still conducted additional tests with a reduced version of the Positive and Negative Effect Schedule (PANAS) tapping four emotions (anger, anxiety, enthusiasm, and sadness).Footnote 15 As Appendix G suggests, it is possible that music is not able to change how individuals feel when it is an element of campaign ads. In fact, individuals assigned to the treatment group are not, on average, statistically different from those assigned to the control group when it comes to how much they feel angry, anxious, enthusiastic, or sad (for the impact of emotions on trait perceptions, see also Appendix G). Note, however, that this exploratory analysis was carried out beyond the purview of the preregistered pre-analysis plan. Furthermore, we are cautious about drawing any firm conclusion about the connection between music and emotions as a better question may ask individuals how they feel with respect to each campaign ad in the experiment, not how they feel in general.
Having said that, it is quite interesting that music marginally affects trait perceptions. While we took many precautions to best assess the impact of music on trait perceptions, our survey experiment is still limited in some respects. Most importantly, it is possible that individuals assigned to the treatment group did not listen to the music in the campaign ads, and therefore did not receive the treatment. To ensure this would not happen, participants in the treatment group were asked to confirm (before the experiment began) that they were able to watch the campaigns ads and that they could listen to the music in those ads. In addition, participants were asked a series of attention checks, including “Is the candidate in the ad young or old?,” “Is the weather in the ad sunny or rainy?,” or “Is the candidate in the ad a man or a woman?” While we expect noncompliers to be the exception in our experiment, we cannot rule out this possibility.
Despite this potential limitation, this study makes a contribution. It furthers our understanding of whether a common non-verbal element in campaign ads (music) can influence the way in which citizens perceive candidates. In other words, if music represents a source of trait perceptions. Given the link between music and emotions (Bachorik et al., Reference Bachorik, Bangert, Loui, Larke, Berger, Rowe and Schlaug2009; Baumgartner et al., Reference Baumgartner, Esslen and Jäncke2006; Krumhansl, Reference Krumhansl1997; Niedenthal & Setterlund, Reference Niedenthal and Setterlund1994; Suguna & Deepika, Reference Suguna and Deepika2017) and between emotions and trait perceptions (Isbell & Wyer, Reference Isbell and Wyer1999), it seems obvious that music can shape trait perceptions. Yet our research suggests that music may in fact not influence individuals’ trait perceptions.
From the perceptive of electoral democracy, the weak (or null) relationship between music and trait perceptions is good news. The reason is that while some candidates may use music in their campaign ads to more easily inform the citizenry about their actual traits, other candidates may use it to mischaracterize some of their traits or even to create traits that are not real. Furthermore, while we do not reject the usefulness of non-verbal cues in helping individuals to have an opinion about a candidate and to make a decision about who they should vote for, there is a risk that individuals are influenced unconsciously by the music from the campaign ads and, consequently, that they end up voting for candidates who are not in line with their political preferences.
In addition to making those contributions, this study opens many avenues of research. More specifically, this study shows that the music from campaign ads whose goal is to foster positive evaluations of a candidate does not affect (or marginally affects) trait perceptions. Yet it is possible that music is more influential on trait perceptions in the context of campaign ads whose goal is to foster negative evaluations of a candidate. Future research may build on this study and examine this possibility, providing evidence on whether and the degree to which our findings generalize to other types of campaign ad (including those that are sponsored by female candidates).
Future research might also explore whether and the degree to which music yields physiological responses on the part of individuals or activates certain parts of their brain. While we do not expect a different conclusion regarding the relationship between music and trait perceptions, this research would further our knowledge of why that relationship proves itself to be empirically weak (or inexistant).
To conclude, the key message from this study is that, despite the use of music by some candidates as a way of influencing how citizens perceive them, music seems to play a limited (or no) role in that respect. From this perspective, music may not represent a tool with which campaign ads may shift public opinion. It may capture the attention of individuals or facilitate the task of remembering a candidate, but it seems unable to shape the way in which individuals perceive candidates regarding their competency, honesty, leadership, and compassion, important determinants of the vote.
Supplementary Materials
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/pls.2021.22.
Competing Interest
None.
