• Keine Ergebnisse gefunden

Popular music lyrics and musicians’ gender over time (S10)

2 Scientific Publications 9

2.3 Real world applications

2.3.4 Popular music lyrics and musicians’ gender over time (S10)

in the UK over time (see Appendix J for the full text). Data on the singles sales charts from 1960 to 2015 was analysed as a proxy of music preferences. Note that singles sales charts is determined by weekly sales, downloads, and streaming of music. With this data, the study focused on how the gender distribution of the United Kingdom’s most popular artists has changed over time and the extent to which these changes might relate to popular music lyrics.

Using data mining and machine learning techniques, all songs that reached the UK weekly top 5 sales charts from 1960 to 2015 were analysed (4,222 songs). A computational analysis of the lyrics was conducted to measure a total of 36 lyrical variables per song. Results showed a significant inequality in gender representation on the charts. However, the presence of female musicians increased significantly over the period covered in the study.

The most critical inflection points leading to changes in the prevalence of female musicians were in 1968, 1976, and 1984. Linear mixed-effects models showed that the total number of words and the use of self-reference in popular music lyrics changed significantly as a function of musicians’ gender distribution over time, and particularly around the three

24 2.3 Real world applications

critical inflection points identified. One of the most interesting trends found in the study is shown in Figure 2.6: Regardless of gender, there was a drastic increase in the total number of words over time, whereas the diversity of vocabulary (i.e., the number of different words in a song divided by the total words) decreased significantly over time, suggesting that UK popular music lyrics have become more repetitive over time. The use of data mining and machine learning techniques (e.g., classification tree models and random forest) offered several advantages in comparison to the statistical tools used in earlier studies.

Fig. 2.6 Mean words per song (left) and diversity of vocabulary (right) in UK popular music lyrics from 1960 to 2015.

600

Total number of words

0.5

Diversity of vocabulary

400

0.4

0.3

Artist Gender Both Female Male

200

1960 1980 Year

2000

0.2

1960 1980 Year

2000

There was a drastic increase in the total number of words used in popular songs from 1960 to 2015 (left plot).

In contrast, when looking at a measure of the diversity of vocabulary (i.e., the number of different words divided by the total words in a song), the results showed that popular songs became less varied and more repetitive over time. This finding occurred regardless of the gender of the artists or band. Note that Both

indicate bands or artists with both female and male members.

Mean words per song Diversity of vocabulary

Chapter 3 Discussion

This section starts by discussing the main theoretical and practical contributions of this thesis, with a focus on what have we learned from applying behavioural economics in the context of music. The section ends with discussing directions for new insights and valuable future research and, finally, concludes.

3.1 Theoretical contributions

The main theoretical contribution of this thesis is the conception of the BEM, an interdisciplinary but unified framework with which we can increase our understanding of musical behaviour (see Figure 1.1). Two literature reviews and eight empirical investigations (see Table 2.1 for a list of publications; see Appendix A-J for the full texts) demonstrate the value and potential of this novel approach.

The BEM contributes most significantly to the existing bodies of music research in both standard economics and psychology. Economists interested in music will benefit by moving away from the more rigid assumptions of standard economics and consider the psychological underpinnings known to be involved in musical behaviour. For example, in several studies conducted within this thesis, we learned that listeners are not utility maximisers who use all information and time available to make optimal musical choices.

Instead, there are several psychological constraints that limit their ability to evaluate and choose music, such as memory and the contextual information often presented with music.

Incorporating these insights will help building a more realistic and comprehensive account of music-related decision making.

26 3.3 Future directions

In comparison to other stimuli, music is experiential, multisensory, aesthetic, social, and highly emotional. Such intrinsic properties may prove particularly useful to test economic theories and enhance their generalizability and scope. For example, music is highly effective in evoking strong emotions in their listeners, such as chill experiences – i.e., the phenomenon of chills or goosebumps caused by intense emotion that come from listening to a specific piece of music (Goldstein, 1980). Thus, music can be an efficient and inexpensive stimulus to study the role of emotion in decision making. For instance, S5 (see section 2.2.3 - Appendix E) found an interaction between the affect heuristic and the emotional content of the music, suggesting that the impact of this heuristic on decision making may differ when using music stimuli in comparison with other stimuli. Similarly, music is social and largely influenced by culture. By investigating properties of popular music (e.g., lyrics) and characteristics of the artists (e.g., gender), S10 (see section 2.3.4 - Appendix J) showed how popular music can be used as a cultural product to study how the preferences and values of a society are shaped by political and socioeconomic changes.

On the other hand, phycologists will gain from considering behavioural economics as a toolkit by which to address key music problems that. In particular, the BEM approach allows psychologists to rethink the study of musical behaviour using a new (and empirically supported) set of concepts and theories, such as bounded rationality, dual-process theory, and behavioural game theory. Whilst these insights have been highly influential in the study of human behaviour and decision making, they have rarely been applied to examine musical behaviour. For example, to date, the notion of heuristic processing has been mostly overlooked in the music psychology literature. Nevertheless, four studies conducted in this thesis (see section 2.2) indicated that heuristics play a central role in music listening and choice behaviour.

The studies conducted in this thesis are important in demonstrating the value of applying insights from behavioural economics to study music decision making. However, they mostly focused on one BEM area (i.e., cognitive biases and heuristics) and one aspect of music decision making (i.e., music preferences and listening behaviour). To address this issue, S2 (see section 2.1.2 - Appendix B) provided an up-to-date account of all studies that utilised behavioural economics for research on music-related decision making. This study contributes significantly to the literature by showing which areas within behavioural economics can generate new and valuable insights into the study of music decision making, both in terms of research methods and theory. The systematic review identified 33 studies organised in four distinctive BEM that readily apply to music decision making: cognitive biases and heuristics, social decision making, behavioural time preferences, and dual-process theory. Each of these BEM areas adds value to the existing bodies of music research in both psychology and economics. For instance, social decision making is an area within behavioural economics that examines how decisions are influenced by social information and preferences. Although at odds with neoclassical economic theory, social preferences (i.e., altruism, reciprocity, and fairness concern) can explain why consumers choose to pay

27 3.3 Future directions

voluntarily for music, a phenomenon that has puzzled researchers for a long time.

Similarly, behavioural time preferences can enable a deeper understanding of how music is valued and consumed over time. Notably, individuals exhibit present-biased time preferences, i.e., they have a strong preference for immediate gratification (O’ Donoghue &

Rabin, 1999). Since music is a hedonic good (i.e., multisensory based on experiential consumption), individuals may place an even higher weight on outcomes that occur in the present rather than the future. This has implications for how consumers select music, particularly with the emergence of music streaming platforms providing music instantaneously. A further area of behavioural economics, dual-process theory, explores the interaction between emotional and cognitive processes in the brain. Dual-process theory can be used to study decision making in the context of music composition and performance. For example, investigating the interaction between these two systems can help better understand conscious states while musicians perform and how these may impact on the quality of their performances.

Another theoretical implication of this thesis is the focus on understanding the role of context in music evaluation and decision making. Research within music psychology has identified three main interconnected factors that influence people when listening to and evaluating music: the music, the listener, and the listening context (see Hargreaves, North, & Tarrant, 2006; LeBlanc, 1982, for theoretical models considering the three factors; see Greasley &

Lamont, 2016; North & Hargreaves, 2008, for research reviews). Traditionally, the vast majority of studies have focused on the music and the listener. Comparatively, less attention has been paid to the listening context. In this thesis, six empirical investigations manipulated contextual factors presented with the music stimuli to investigate its effects on musical behaviour, including artists names, song titles, information about the artists, post-event information about the music piece, and the source of the music. These studies consistently show that music decision making does not happen in a vacuum, but is significantly influenced by the context. More specifically, contextual information can lead listeners to perceive different musical performances when in fact they are identical (S3; see section 2.2.1 - Appendix C); generate false musical memories of a past music event (S4; see section 2.2.2 - Appendix D); influence music judgments and decision even when the contextual manipulation is minimal, such as only changing linguistic aspects of titles presented with music (S5 and S6; see section 2.2.3 and 2.2.4 - Appendix E and F); and cause potentially negative biases amongst ad professionals when choosing music for advertising (S7; see section 2.3.1 - Appendix G).

Furthermore, the studies conducted in this thesis contribute towards a better understanding of the role of music expertise in music evaluation and decision making. Previous research consistently shows that highly trained musicians outperform non-musicians in several musical tasks, such as short-term and working memory tasks with music stimuli (see Talamini, Altoe, Carretti, & Grassi, 2017, for a review). Thus, it seems plausible that since musicians’ cognitive abilities to perceive and process music are higher than non-musicians,

28 3.3 Future directions

they should perhaps be less influenced by contextual factors and cognitive heuristics.

Several of the studies conducted in this thesis addressed this issue by collecting data on participants’ musical background, including both musical training and active engagement to music. These studies, which collected data on more than 500 participants, showed that music expertise does not have a protective effect against contextual factors. Besides, highly trained musicians are not any more or any less susceptible to cognitive biases and heuristics than non-musicians. Thus, contextual factors and heuristics seem to influence listeners regardless of their previous experience to music. Although these results might seem counterintuitive at first, they are consistent with the behavioural economics literature on the "expert problem"

(e.g., Hall, Ariss, and Todorov, 2007; Reyna, Chick, Corbin, and Hsia, 2014; Taleb, 2007), showing that in certain conditions and domains, more knowledge and expertise does not necessarily lead to more accurate and less biased judgments and decisions.

3.2 Practical contributions

A main practical contribution of this thesis is the wide variety of methods and paradigms used to investigate different aspects of music decision making. For example, S3 (see section 2.2.1 - Appendix C) proposed the repeated recording illusion, a novel paradigm that is useful to investigate non-musical factors in music evaluation because it allows for the study of their effects while the music remains the same. S4 (see section 2.2.2 - Appendix D) applied, for the first time, the misinformation paradigm using music instead of visual materials, showing that listeners generate false memories in a music context. Both S5 (see section 2.2.3 - Appendix E) and S6 (see section 2.2.4 - Appendix F) adapted successfully existing paradigms in the behavioural economics literature to study the effects of cognitive heuristics in music evaluation and decision making. S5 adapted a well-known experiment from behavioural economics (Shah & Oppenheimer, 2007) to examine linguistic fluency, whereas S6 adapted a common paradigm to investigate the recognition heuristic in preferential choice tasks (Oeusoonthornwattana & Shanks, 2010) to study musical choices when listeners search for music in playlists. Overall, these studies emphasize the potential of applying methods and paradigms from behavioural economics to study similar phenomena in music.

This thesis also explored other methods beyond those commonly used in controlled and artificial studies. This is important because controlled studies conducted in laboratories and other artificial environments are susceptible, among other things, to two major problems (Carpenter, Harrison, & List, 2005; Reis & Judd, 2000): a lack of external validity—the extent to which the results are generalizable beyond the research setting and participant pool—and a lack of ecological validity—the degree to which the results apply to the real world situation under study. Note that issues related to poor ecological validity and generalizability are taken particularly seriously by economists and behavioural scientists (Harrison & List, 2004; Levitt & List, 2007). As argued by Levitt and List (2007), “Perhaps

29 3.3 Future directions

the most fundamental question in experimental economics is whether findings from the lab are likely to provide reliable inferences outside of the laboratory” (p. 179). Thus, it was important to consider further ways to examine behavioural responses to music in natural environments, once sufficient scientific grounding has been obtained based on laboratory-generated data.

This was one of the motivations for S9 (see section 2.3.3 - Appendix I), which used a field research approach to investigate responses to musical performances in a naturalistic busking environment. The charitable behaviour of passersby’ (i.e., amount of money donated) was recorded while a professional busker performed in the London Underground over the course of 24 days. Two factors of the performance, commonly investigated in lab studies, were manipulated: familiarity of the music and body movements. Contrary to common findings in lab studies looking at the same factors, the results indicated that neither music familiarity nor the performer’s body movements had a significant impact on the amount of money donated. These discrepancies might be due to differences in the ecological validity between laboratory and field studies. For instance, in the laboratory, participants are always aware of their participation in a scientific study and their only goal is to listen carefully to the music while evaluating it in a highly controlled and quiet environment. Therefore, the advantages of field studies over lab studies include high ecological validity and avoiding problems associated with self-reported assessments. On the other hand, S10 (see section 2.3.4 - Appendix J) used a big data approach that offers high ecological validity but also allows to analyse large datasets. In particular, the study analysed naturalistic data from the singles sales charts to examine how preferences for popular music have changed over time. This approach allowed for the study of responses to music in the real world that are neither obtained nor affected by the actions of researchers. Moreover, it made it relatively easy to collect and analyse a large dataset (4,222 songs and 2,287 artists) covering 55 years (1960-2015). Overall, S9 and S10 are useful in exploring alternative methods to study musical behaviour that do not suffer from poor external and ecological validity.

Another contribution of this thesis includes the wide variety of analysis techniques used across the eight empirical studies to examine human responses to music. Firstly, linear mixed-effect models proved to be very efficient to test main hypotheses when using repeated- measured designs, as they allowed the modelling of important sources of random noise, such as participants and songs’ variability. Secondly, several studies applied machine learning and data mining techniques that proved to be well-suited to examine certain music problems. For example, to analyse a large set of individual differences, S3 (see section 2.2.1 - Appendix C) and S4 (see section 2.2.2 - Appendix D) showed that random forests can be a very useful technique, as they can handle a large number of variables even when they are correlated between themselves. Similarly, classification tree models were successfully applied in S3 (see section 2.2.1 - Appendix C) and S10 (see section 2.3.4 - Appendix J) to model higher-order interactions between several predictor variables and the dependent variable of the study. In S10, for instance, this approach allowed for the identification of three main inflection points in which the prevalence of female artists in UK’s popular music

30 3.3 Future directions

changed considerably over time, i.e., 1968, 1976, 1984. Interestingly, these inflection points coincide with some significant moments in UK’s culture, such as the surge in popularity of the women’s rights movements (1968), the rise of punk (1976), and the peak in popularity of Margaret Thatcher’s prime ministership (1984).

Finally, the last part of this thesis focused on the application of the BEM to improve music- related decision making in the real world. Two studies focused on music decision making in advertising and marketing, an area where music choices are particularly relevant. Musical choices can have profound effects on brand communications, consumer behaviour, and can be costly for brands. Despite this, the process of choosing and evaluating music for advertising is poorly understood. Thus, S7 and S8 applied insights from behavioural economics to better inform the decision-making process of selecting music for ads. S7 (see section 2.3.1 - Appendix G) found that source bias has a significant impact on how ad professionals evaluate music for advertising purposes, whereas source cues have no effect on consumers. The differences between these two groups can be costly for brands, as professionals may recommend that their clients pay a premium for music coming from specific sources (e.g., performing artists), but brands may see little or no added benefit if ultimately, the source of the music does not matter to the consumer. Potential solutions to mitigate this bias are discussed, including increasing awareness among professionals and measuring the impact of advertising music and source on target consumers. In addition, S8 (see section 2.3.2 - Appendix H) provided a first estimation of the effectiveness of using music as a recognition cue to influence consumer choice by means of the recognition heuristic. The results showed that music can only be successfully used as a recognition cue when it is liked by the target consumers, whereas recognition-based heuristics are not influential when the music is disliked. This finding is valuable to brands in terms of the importance of measuring the value of their investment when working with music.

31 3.3 Future directions

3.3 Future directions

The diversity of the ten studies presented in Chapter 2 begin to illustrate the breadth and potential of the role that behavioural economics can play in music research. Moreover, a network visualisation map of behavioural economics demonstrates the rich array of concepts and theories yet to be applied within the domain of music research (see Figure 3.12). The map was generated using 68,509 publications from the field of behavioural economics and shows the 200 most frequently used concepts in these publications (the key terms provided by the author (s) to describe their work). Thus, the map highlights the potential of behavioural economics for future research on music decision making. Based on this map and further accumulated insights from the ten empirical investigations summarised in Chapter 2, this section proposes ideas and directions for future research within the BEM (RQ 5). The future directions are organised around different areas within music research that can significantly benefit from considering the behavioural economics toolkit.

Fig. 3.1 Network visualization map of behavioural economics.

The map shows the 200 most influential keywords in the behavioural economics literature (based on 68,509 publications) and how often they co-occur with others, identifying main research areas and concepts in the field. On the map, each concept is represented by a circle, its size determining how frequently the concept was used in the retrieved literature. The lines between concepts indicate how connected the concepts are with each other. The stronger the connection,

the wider the line. Highly connected concepts are grouped into clusters, with different colours representing different clusters.

323.3 Future directions

33 3.3 Future directions

Music reward value

Music reward value is at the core of musical behaviour. Yet, if musical sounds have not

Music reward value is at the core of musical behaviour. Yet, if musical sounds have not