Age- and Sex-Related Differences in Music Requests on a Radio Show for Children

Abstract

In research on the development of musical preferences, children are often asked either to evaluate musical pieces that have been previously selected by adult experimenters, or to give self-reports on preferences for different musical styles. This study involved analyzing a sample of 1,412 freely and publicly expressed music requests from children aged between 4 and 11 years taken from a German radio program, with regard to age- and sex-specific differences. The music was categorized into genres, as listed on Spotify, and examined using methods of music information retrieval provided by the Spotify Developer application programming interface. Results showed that, at younger ages, the requests were generally more evenly distributed across different genres. Regarding single genres, we observed small positive relationships between age and the likelihood of requesting the genres pop or electro and small negative relationships between age and the likelihood of requesting A capella, German songwriter, indie, theme songs, or children's music. Furthermore, we found that boys requested significantly more rock or hip-hop, whereas girls had a higher tendency to ask for pop. Finally, age- (but no sex-) related differences regarding the Spotify features of valence and liveness of the requested music were found, which were related to the preference for children's music at younger ages. The results thus suggest that previous findings regarding differences between boys and girls and an increasing formation of distinct genre preferences during infancy also apply to single song requests made on a radio show.

Introduction

The questions of what music is popular at what age and how musical preferences and tastes change with age have long been the subject of research in music psychology. Musical preference usually refers to choosing one musical piece over another in a given situation, while the overall liking of certain musical genres and style characteristics is termed musical taste (Hargreaves et al., 2016). In many studies on children's musical preferences, participants rated music samples that had been selected and sometimes created by adult experimenters (Busch et al., 2009; Dobrota & Sarajčev, 2021; Finnäs, 1989; Gembris & Schellberg, 2003; Gregory et al., 1996; Hargreaves et al., 1995; Hemming, 2013; Kopiez & Lehmann, 2008; LeBlanc et al., 1996; Louven, 2016; Nieminen et al., 2012; Peery & Peery, 1986). The selection of a usually small set of musical genres and pieces, however, might severely bias studies on children's musical preferences. For example, Lehmann and Kopiez (2011) argued that, as children grow older, preference ratings might be affected by an increasingly critical attitude toward content presented by adults.

Another methodological approach is based on self-reports on the preferences for about 10–15 (Bakagiannis & Tarrant, 2006; Colley, 2008; Hargreaves et al., 1995; North et al., 2000; Rentfrow & Gosling, 2003; Tanner et al., 2008) or, occasionally, more (Christenson & Peterson, 1988; Gardikiotis & Baltzis, 2012; Lorenzo-Quiles et al., 2020; North, 2010) musical genres. These reports depend, of course, on the objectivity and honesty of young people's self-assessment, and the ultimately arbitrary number of genres that they have to rate.

A third type of study relies on the experience sampling method (ESM) (Greb et al., 2019; Lamont, 2008; North et al., 2004; Randall & Rickard, 2017), which has become much more feasible through the availability of specific smartphone apps (Randall & Rickard, 2013), but still requires participants’ willingness to include the repeated answering of questionnaires in their daily life over a period of several days.

This study utilizes another methodological approach by collecting and analyzing music that was freely requested on a children's radio show. There are fundamental differences between these song requests and the data retrieved from the aforementioned methodological approaches.

Advantages of Assessing Musical Preference Through Single Song Requests

When children choose the requested songs themselves, there are no issues with selecting appropriate music for a comprehensive analysis of participants’ musical preferences. As a consequence, the song requests exclusively consist of music that the children already know and like. In studies with previously selected stimuli, some or all of the musical pieces might be unfamiliar to some or all participants, respectively, and how much children and adolescents like the given musical pieces is usually one of the primary objects of measurement (Busch et al., 2009; Dobrota & Sarajčev, 2021; Finnäs, 1989; Gembris & Schellberg, 2003; Gregory et al., 1996; Hargreaves et al., 1995; Hemming, 2013; Holbrook & Schindler, 1989; Kopiez & Lehmann, 2008; LeBlanc et al., 1996; Louven, 2016; Nieminen et al., 2012; Peery & Peery, 1986).

This study's objects of measurement are quasi-self-reported song preferences, which are more specific than, for example, genre preferences reported in surveys conducted in a classroom or online. As the requests concern a particular song—contrary to studies asking for participants’ (stable) preferences for musical genres—the song's selection as a request is affected by several situational factors, such as the current mood, the presence of parents, siblings, or friends, or a momentary high preference for a recently discovered song. Therefore, the preferences captured by the song requests are probably more time-variant and context-dependent. However, although making a call to a radio station to request a song is perhaps unusual for the children, it is a deliberate, most likely self-initiated act of their real musical engagement, and thus an interesting alternative assessment of preference, compared with the outcome of rating tasks created by adult experimenters.

Taken together, this study involved testing the extent to which established concepts of children's and adolescents’ musical tastes and preferences apply to single song requests made on a radio show. The following section gives a brief overview of theories on age- and sex-related differences in both musical preferences and musical taste of children and adolescents, which serve as a basis for the research questions addressed in this article.

Age-Related Changes in Musical Preferences From Childhood to Adolescence

Previous research has observed vast changes in musical preference during childhood and adolescence that are linked to three mechanisms: the development of children's aesthetic experiences (Nieminen et al., 2012), the decrease of so-called open-earedness at the transition from childhood to adolescence (Gembris & Schellberg, 2003), and adolescents’ formation and expression of social identity through distinct musical preferences (Hargreaves et al., 2016; North & Hargreaves, 1999).

Concerning the aesthetic experiences of music, it has been shown that in preschool age (4–6 years), it seems to be easier for children to recognize happiness in music as opposed to sadness or anger (Cunningham & Sterling, 1988). According to the theory of mind, these developments in musical emotion recognition might be linked to cognitive abilities in this age range, as preschoolers would be able to recognize false beliefs, but would not understand another person's perspective until the age of 6–8 years (Wellman, 2002).

By the age of 6 years, children in Western societies associate music in the major mode with happiness and music in the minor mode with sadness (Dalla Bella et al., 2001; Gregory et al., 1996). Nieminen et al. (2012) observed that children between 6 and 9 years consistently preferred music written in the major mode or with free tonality over music written in the minor mode, concluding that children prefer happy over sad music. Their explanation for these preferences is that children are not yet capable of dissociating the aesthetic quality of sad music from the negative emotions it expresses. The assumption that music whose affective qualities can be better understood would be more stimulating also supports a preference for music expressing happiness among children.

As, in adolescence, music becomes an important tool for mood regulation (Londsdale & North, 2011), these preference patterns partly change.

Regardless of their preference for positively valenced music, children have been found to give higher preference ratings in general (LeBlanc et al., 1996) and to be more receptive to different musical genres at preschool age and in the first years of elementary school. Here, Hargreaves (1982) coined the term open-earedness to describe an openness toward music that adults would consider to be unconventional. This open-earedness would rapidly decline when children are about 9–11 years old (Gembris & Schellberg, 2003; Hargreaves et al., 2016). Although several researchers found evidence of the existence of open-earedness (Dobrota & Sarajčev, 2021; Hargreaves et al., 2016), Louven (2016) indicated some conceptual issues and proposed tolerance for (disliked) music and openness to unknown music as the underlying principles of open-earedness. Hargreaves and Bonneville-Roussy (2018) partially adopted this approach and linked openness to unknown styles to an omnivorous musical taste and introduced the flexibility of personal genre constructs as an additional aspect of open-earedness.

Because, in this study, we analyze data of (mostly) one request of familiar music per child, the liking of different, potentially unknown, musical styles by an individual child is beyond the scope of this research; thus, open-earedness will not be further addressed in the remainder of this article.

Instead, in our study, requests were aggregated to infer the popularity of musical styles within a group of children at a certain age. In this context, previous studies found that children develop an increasing sensitivity to style as they approach adolescence (Hargreaves, 1982); this consolidates preferences for distinct genres, especially for pop music (Gembris & Schellberg, 2003; de Vries, 2010)—whether open-earedness decreases or not (Kopiez & Lehmann, 2008). Adolescents use musical taste as a badge to express their own identity, judge the personality of others, and organize their social environments in peer groups characterized by shared musical preferences (Bakagiannis & Tarrant, 2006; North & Hargreaves, 1999). At least in terms of musical genre, Mulder et al. (2010) found that preferences acquired in this process remain relatively stable in a sample of adolescents and adults from 12 to 29 years old.

A crucial factor in forming an identity through musical taste is the shift from parent-orientation to peer-orientation, and musical preferences have been found to be mostly peer- and self-driven by the age of 9 years (Thompson, 1985; Troué & Bruhn, 2000). In a later study, ter Bogt et al. (2011) showed that the impact of parents’ musical tastes on those of their children persists during adolescence, but is accompanied by peer influences, from which new musical preferences emerge. They concluded from their results that, for example, girls’ preferences for rock were primarily passed on from parents, whereas preferences for rock among boys were predominantly peer-oriented.

Sex Differences in Musical Preferences

Since music makes a decisive contribution to the formation of identity, it is not surprising that there are also sex differences in preferences for specific genres (Dobrota & Sarajčev, 2021; Gembris & Schellberg, 2007; Habe et al., 2018; Hargreaves et al., 1995). Despite the varying number and selection of genres tested in different studies, girls tend to prefer more calm and reflective musical styles, while boys prefer harder, more rebellious, or aggressive music and tend to avoid mainstream music (Colley, 2008; North, 2010; P. A. Russell, 1997). North et al. (2000) found that the expression of identity through music is, overall, more important to boys; that is, boys favor music that expresses rebellion and individuality, whereas girls rather use music for mood regulation, for which emotional and romantic themes of chart pop might be more admissible (Colley, 2008). As rock often includes more negative and angry attitudes than pop music, these sex differences can also be attributed to general preferences for higher-valenced stimuli among girls that have been observed for both music (Hunter et al., 2011) and pictures (Sharp et al., 2006).

Furthermore, it has been observed that girls seem to give overall more positive ratings to music of different genres, such as jazz, classical music, and opera, than boys (Gembris & Schellberg, 2007; Hargreaves et al., 1995). Hargreaves et al. (1995) observed about 15%–20% more dislike ratings for these genres among boys, which they explained as reflecting a higher degree of musical training in the girls in their test sample, leading to more openness to styles with higher musical complexity.

Research Questions and Hypotheses

Frequencies of requests for certain genres were analyzed as indicators of the genres’ popularity at different ages and between boys and girls. Therefore, three hypotheses were formulated to test whether established concepts on age- and sex-related differences in musical preferences can also be found in aggregated, freely expressed music requests.

First, expecting different developmental stages regarding abilities to recognize emotion and dissociate negative emotions from sad music, it was hypothesized that younger children in our sample would request more happy songs, characterized by a higher valence (H1).

Given the narrowing in style sensitivity in early adolescence, the second hypothesis was that music requests of younger children would be more diversified over different genres and be concentrated on pop music with increasing age (H2).

Third, it was hypothesized that there would be a significant sex differences in the genres of the requested songs, with girls requesting more pop music than boys (H3).

Because the data provided actual songs as quasi-self-reported musical preferences (and not just favored genres), further exploratory analyses were included to investigate whether there were age- and sex-related preferences for musical style characteristics beyond the songs’ genre. These analyses were based on additional audio features provided by the Spotify Developer application programming interface (API), namely, valence, energy, acousticness, danceability, instrumentalness, liveness, and speechiness (Spotify, 2023), and metadata of the songs taken from the radio station's internal music database, such as the release year, the singer's sex, and how often a requested song was used on the radio show.

Method

Sample

A total of 1,412 music requests were collected from a German daily radio show for children called Die Sendung mit der Maus zum Hören (The Radio Show With the Mouse), hosted by the German public broadcaster Westdeutscher Rundfunk. The show is based on a popular TV show for children that has been running on German public television for more than 50 years. The episodes are also published as podcasts on various streaming services and are accessible without subscription at the broadcaster's homepage (https://www.wdrmaus.de/hoeren/).

The radio show invites children to call and ask questions that will be answered on the show, greet friends and family, or request music. An automated answering system can be called at any time and free of charge. Callers are asked, but not required, to provide their names and ages.

The music requests analyzed here were received between November 17, 2020, and September 9, 2021. Of the callers, 651 were male (46.1%) and 761 were female (53.9%). The ages of the children ranged from 4 to 11 years (M = 7.28, SD = 1.73), with negligible age differences between boys (M = 7.10 years, SD = 1.71) and girls (M = 7.42 years, SD = 1.74). A total of 210 children did not provide age information, so all analyses of age effects are based on a subsample of 1,202 requests.

Originally, 1,442 requests were received in this time period, but five requests for which neither age nor sex information was available were removed from the data. Also, requests from 1-year-olds (where the parents made the call), and from children aged 3, 12, or 13 years were removed from the dataset, because there were only 10 or fewer requests for each of these ages.

Informed Consent

Since the described analyses were performed on an existing body of received calls, it was not possible to obtain any form of consent from the children or their parents. However, as the broadcaster did not provide us with the actual voice messages, but only the requested titles and, if available, the sexes and ages of the children, their anonymity was guaranteed and the study was ethically approved.

Song Analysis

The ages and sexes of the children were obtained from the broadcaster's internal database. As no other information on the children was available to the authors, it was not possible to identify several calls from the same child in the analysis. However, the production staff reported that this happened rather irregularly, with few children making more than one call, and very few children calling several times.

Several songs were requested repeatedly by different children, and the 1,412 music requests comprised 511different songs. ¹ Because the genre information stored in the broadcaster's internal music database was incomplete, the Spotify Developer API was used to obtain the requests’ genres and also to calculate several audio features. These features were: valence, which is described on the API's website as “musical positiveness”; energy, including “dynamic range, perceived loudness, timbre, onset rate, and general entropy”; acousticness, as a descriptor of “whether a track is acoustic” (with no further explanation); danceability, based on “musical elements, including tempo, rhythm stability, beat strength, and overall regularity”; instrumentalness, estimating “whether a track contains no vocals”; liveness, which indicates “the presence of an audience in the recording”; and speechiness, which “detects presence of spoken words in a track” (Spotify, 2023).

Seven songs (requested 15 times in total) were not available on Spotify. Therefore, they had to be excluded from the audio feature analysis and the genre information was retrieved from the broadcaster's internal music database instead.

Furthermore, Spotify treats children's music as a distinct musical genre. However, different musical styles within the domain of children's music were of interest to this research. Thus, for songs that Spotify categorized as children's music, the genre information was obtained from the broadcaster's music database. In some cases, this database provided no genre information either, and one of the authors manually annotated genre information, for example, based on other songs of the same artist with similar style characteristics, but for adult audiences. Additionally, a binary identifier for children's music was taken from the broadcaster's database. After these adjustments, our data contained a musical genre and a separate identifier for children's music for each requested song (see Table 1 for examples).

Table 1.

The 10 most frequently requested songs. Except for the third and fifth songs, all have German lyrics (in the eighth song, only a few lines and single words are in French).

	Song	Artist	Genre	Children's music	Requests
1	Ich frag die Maus	Mark Forster	Pop	Yes	55
2	194 Länder	Mark Forster	Pop	No	45
3	Dance Monkey	Tones And I	Pop	No	41
4	Roboter	Eule und Lerche	Pop	Yes	33
5	Wellerman (Sea Shanty)	Nathan Evans	Pop	No	32
6	99 Luftballons	Nena	Pop	No	28
7	Mädchen gegen Jungs	Lina Larissa Strahl	Pop	Yes	23
8	Je ne parle pas français	Namika	Pop	No	20
9	Bist du okay?	Mark Forster	Pop	No	17
10	Hausaufgaben	Deine Freunde	Electro	Yes	16

The requests comprised 19 genres (Table 2). Additional information on the songs’ release dates and usage on the radio show was taken from the broadcaster's music database.

Table 2.

Requests for different genres at each age. The last row displays the number of requests classified as children's music at each age. The last two columns show, respectively, the number and percentage of each genre for all requests.

Genre	Age									Total
	4	5	6	7	8	9	10	11	NA	N	%
Pop	18	63	108	119	218	101	63	19	122	741	52.5
Hip-hop	8	11	22	23	29	21	7	2	16	139	9.8
Rock	4	17	16	14	29	9	8	1	13	111	7.9
German songwriter	11	19	20	24	12	10	4		7	107	7.6
Electro	1	5	12	11	13	20	6	3	18	89	6.3
Indie	3	13	10	10	9	5	4		10	64	4.5
A capella	5	4	6	3	2	3	1		5	29	2.1
Theme song	3	6	5	1	2	4			4	25	1.8
Singer-songwriter	1	5	2	2	2			2	4	18	1.3
Traditional	2	2	3	2		1		1	5	16	1.1
Reggae		1	1	2	4	2	2		2	14	1.0
World		1	4	2	4	2			1	14	1.0
Oldies		1	2	4	1	3	1			12	0.9
Classical	1		1	2		2	1	2		9	0.6
Funk	1	1	5			1				8	0.6
German Schlager			1	2	1	1			1	6	0.4
Eurodance		2	1		1	1				5	0.4
Worship							2		2	4	0.3
Country					1					1	0.1
Sample	58	151	219	221	238	186	99	30	210	1,412
Children's music (%)	81.0	61.6	49.8	50.2	40.8	35.5	35.4	30	39.5	46.0

To measure the preference of certain genres over others at different ages, an index for genre preferences (IGP) was calculated for each age, as

$$\mathrm{IGP}={\displaystyle \frac{\left(\sqrt{{\sum }_{\mathrm{g}}{\left({\displaystyle \frac{r_{\mathrm{g}}}{N_{\mathrm{r}}}}\right)}^2}-\sqrt{{\displaystyle \frac{1}{G}}}\right)}{1-\sqrt{{\displaystyle \frac{1}{G}}}}}$$

where $r_{\mathrm{g}}$ denotes the number of requests per genre, $N_{\mathrm{r}}$ the total number of requests, and G the total number of genres at a specific age. The IGP describes the length of a vector in a G-dimensional space, which is scaled to range from 0 to 1. Because this vector is based on the relative number of requests for each genre, the sum of its elements is always 1. The more the requests concentrate on a few genres, the more the vector points toward the respective axes in the G-dimensional space, increasing its length. The vector has a minimum length of $\sqrt{1/G}$ , if there are equal numbers of requests for all genres, and a maximum value of 1, if all requests are for the same genre. Therefore, the IGP corrects the vector length to range from 0 to 1, allowing for comparison of IGP values calculated for different ages with different total numbers of genres.

Note, that the IGP treats all genres as orthogonal; because our data only contain one request per child, we cannot test for latent dimensions within genre preferences (Colley, 2008; Gardikiotis & Baltzis, 2012; Rentfrow & Gosling, 2003; ter Bogt et al., 2011). However, as the IGP is used to observe trends over a wide genre space and to analyze the popularity of distinct genres in a different manner, we believe that the assumption of genre orthogonality does not substantially affect the results.

Statistical Analysis of Effect Size Measurement

As the examined data include different scale levels (genre information, age of children, song valence scores, etc.), different statistical models with different effect size measures had to be calculated; those used most frequently will be briefly described here.

Linear regression models were used to compare the requesting children's ages with the requests’ audio features. Cohen's f² was obtained from the coefficient of determination, R², representing the amount of explained variance, and the benchmarks for f² proposed by Cohen (1988) were applied as a preliminary assessment of the effect's practical significance.

To test the association between age and whether a request would belong to a certain genre, logistic regression models were calculated, and the adjustment of McFadden's pseudo-R² proposed by Horowitz (1982), R²_MFH, was used to assess the model fit. Hemmert et al. (2018) found this pseudo-R² to be most stable across different sample sizes and asymmetric distributions of the binary outcome variable, which is a desirable feature in this analysis, where the requests for a certain genre are often just a small percentage of the total requests (Table 2). They also propose benchmark values, depending on sample size and the number of successes in the dependent variable. For sample sizes greater than 200, they identify values of R²_MFH between .09 and .17 as indicators of good fit, if the number of successes is below 38% or above 62%, and values of R²_MFH between .11 and .28, if the number of successes is above 38% and below 62% (Hemmert et al., 2018).

The likelihood for boys and girls to request certain genres was estimated using 2 × 2 contingency tables of the requesting child's sex and whether the request was of the genre in question. A χ²-test for group independency was conducted and odds ratios (OR), including 95% confidence intervals, as well as the better-interpretable relative risk (RR), were calculated as effect size measures. Chen et al. (2010) proposed OR equivalents to Cohen's $d$ -benchmarks for epidemiological research. Benchmark values are given for outcome rates of interest in the control group (corresponding to, for example, the number of girls requesting rock songs when, in general, boys request more rock songs) between 1% and 10%, but it was concluded that, generally, OR < 1.5 is comparable to Cohen's d < .20. Cramér's V was calculated as an additional effect size measure, because the number of request for a genre in the control group was not always as small as 10%. All statistical analyses were performed using R (R Core Team, 2022).

Results

The majority of requests was for songs that were categorized as pop music (52.5%; 741 requests for 197 different songs by 116 artists). Table 2 shows the relative number of requests for each genre at each age, as well as the total relative number of requests for each genre, and the portion of requests categorized as children's music at each age. At all ages, pop was the most requested genre. The next most requested genres were hip-hop (9.8%; 139 requests for 61 different songs by 25 artists) and rock (7.9%; 111 requests for 64 different songs by 38 artists).

Table 1 lists the 10 songs that were requested most often. The maximum was 55 requests for the same song (3.9% of all requests). Except for the last song in Table 1, all songs are pop music and four of them are additionally categorized as children's music. Only two songs (“Dance Monkey” and “Wellerman (Sea Shanty)”) have English lyrics, whereas all others contain German lyrics (in the eighth song, only the title and a few lines and single words are French).

A total of 650 requests (46.0%) were additionally categorized as children's music. Table 3 shows how these requests were distributed in terms of the musical genre (third column) and how many requests of each genre were marked as children's music (second column). The three most-represented genres in children's music were pop, with 39.5%, hip-hop, with 18.0%, and German songwriter (Liedermacher in German), with 14.5%. Within the a capella, traditional, and theme song genres, all or nearly all requests were categorized as children's music; within German songwriter, hip-hop, and reggae the proportion was greater than 80%.

Table 3.

Requests for each genre that were classified as children's music, relative to the total number of requests for the genre and the total number of requests for children's music, respectively. Only genres whose requests included children's music are shown.

Genre	Proportion classified as children's music (%)	Proportion of all children's music (%)
A capella	100.0	4.4
Electro	23.6	3.2
Funk	25.0	0.3
German Schlager	50.0	0.5
German songwriter	87.9	14.5
Hip-hop	84.2	18.0
Indie	60.9	6.0
Pop	34.7	39.5
Reggae	85.7	1.9
Rock	19.8	3.4
Singer-songwriter	77.8	2.2
Theme song	96.0	3.7
Traditional	100.0	2.5

Age and Preferred Audio Features

Figure 1 depicts the Spotify valence and energy features of the 511 requested songs, mapped onto the circumplex model of affect by J. A. Russell (1980). Energy is assumed to be an appropriate proxy of arousal, used in the original model (Panda et al., 2021); following the model's scaling, the audio features were transformed to a range from −1 to +1. The figure illustrates that the majority of requests lies within the first quadrant of the model, corresponding to both positive energy and valence, and the second most requests are in the second quadrant of negative valence and positive energy. Overall, there are only a few requested songs with negative energy.

Figure 1.

Spotify valence and energy values of requested songs, mapped onto the circumplex model of affect.

Bivariate correlations between the children's ages and the Spotify features revealed significant negative correlations between age and valence, r(1,185) = −.191, p < .001, and age and liveliness r(1,185) = −.101, p = .009.

The first hypothesis (H1) was tested by computing a linear regression model to predict the songs’ valence by the age of the children requesting the song. This model suggests a significant negative effect of age on the requested songs’ valence, confirming H1, F(1, 1,185) = 45.09, p < .001, R² = .038, Cohen's f² = .038. Figure 2(a) shows the valence of the songs requested at each age as box plots. From ages 4 to 8, the mean valence decreases; afterwards, it appears to remain at a rather steady level.

Figure 2.

(a) Valences of songs requested at each age. (b) Relative number of requests for children's music at each age.

A similar decrease with age was observed for the relative amount of songs categorized as children's music (Figure 2(b)). Here, a linear regression model confirmed a negative, but weak, association of age with the likelihood to request children's music (p < .001, R²_MFH= .034). As a Welch t-test showed that the songs’ valence was also significantly higher in children's music, t(460.3) = –5.463, p < .001, Cohen's d = .488, it was tested whether the relation between age and valence was just an artifact of the decrease in requests for children's music. However, recalculation of the regression model that included only the requests that were not categorized as children's music showed that the negative association of age and valence was still significant (p = .016), but only with a very small effect size, F(1,622) = 6.48, p < .001, R² = .010, Cohen's f² = .010.

Age and Genre Preferences

Figure 3 displays the relative number of requests for different genres as bars. For better visibility, genres that accounted for less than 5% of the requests at any age are aggregated as “other.” As can be seen in Table 2 and Figure 3, the relative frequency of requests for pop songs and electro increases rather steadily with age, while the number of requests categorized as a capella, German songwriter, and theme song mostly decreases. The numbers of requests categorized as hip-hop and rock, respectively, appear to be rather steady across different ages.

Figure 3.

Relative number of requests for different genres at each age.

Logistic regression models were fitted to test whether the likelihood of requesting each genre would increase or decrease with age. The country and worship genres were excluded because they were only requested at one age each. The regression models suggest significant positive effects of age on the likelihood of requesting electro (p = .007, R²_MFH= .020) or pop music (p < .001, R²_MFH= .019), and significant negative effects on the likelihood of requesting a capella (p = .004, R²_MFH= .040), German songwriter (p < .001, R²_MFH= .031), indie (p = .021, R²_MFH= .014), or theme song (p = .007, R²_MFH= .040).

All of these models suggest a very minor effect size, according to Hemmert et al. (2018); even though Figure 3 shows that the genre distribution within the request changes with age, there seem to be only minor effects of age on the likelihood to request a certain genre. It should also be noted that, in the cases of a capella, German songwriter, and theme song, either most or all of the requests were categorized as children's music.

Furthermore, the IGP was calculated as an overall representation of the genre distribution at different ages. Figure 4(a) displays the number of genres and the relative frequency of requests for pop music at each age, and Figure 4(b) shows the IGP, which closely follows the relative number of requests for pop music. At every age, most requests belong to this genre (Table 2 and Figure 3); therefore, the genre vector described by the IGP is always skewed toward pop music. The rather strong decrease at age 11 illustrates that, despite its normalization, the IGP depends on the total number of genres, which also considerably drops at this age.

Figure 4.

(a) Relative amount of pop music requests (black line) and total number of requested genres (gray line) at each age. (b) Index for genre preferences (IGP) at each age.

To test the second hypothesis (H2), the IGP was correlated with age, revealing a substantial significant correlation of r(6) = .881 (p = .004), confirming that musical preferences for a few genres (in this case, pop music in particular) consolidate at the transition from childhood to adolescence.

Sex and Genre Preferences

The third hypothesis (H3) was tested by performing chi-square tests on contingency tables, showing how the genre requests were distributed between boys and girls (Table 4). This analysis included the full sample of 1,412 requests (including the 210 requests where no age was given).

Table 4.

Boys’ and girls’ requests for each genre and for children's music.

Genre	Requests by boys		Requests by girls
	Absolute	Relative (%)	Absolute	Relative (%)
A capella	18	62.1	11	37.9
Classical	4	44.4	5	55.6
Country	1	100.0	0	0.0
Electro	49	55.1	40	44.9
Eurodance	2	40.0	3	60.0
Funk	5	62.5	3	37.5
German Schlager	2	33.3	4	66.7
German songwriter	57	53.3	50	46.7
Hip-hop	85	61.2	54	38.8
Indie	25	39.1	39	60.9
Oldies	10	83.3	2	16.7
Pop	293	38.9	462	61.1
Reggae	5	35.7	9	64.3
Rock	67	60.4	44	39.6
Singer-songwriter	6	33.3	12	66.7
Theme song	12	48.0	13	52.0
Traditional	8	50.0	8	50.0
World	5	35.7	9	64.3
Worship	2	50.0	2	50.0
Children's music	272	44.6	338	55.4
Total	661	46.0	777	54.0

Table 5 shows the results of analyzing 2 × 2 contingency tables comparing the requests for each genre with those of other genres for boys and girls, respectively. The table shows the χ²-test statistic and values of p, OR with 95% confidence intervals, RR, and Cramér's V. For better readability, dummy coding of the children's sex was carried out so as to lead to positive values of OR and RR by always assigning the value 1 to the sex with more requests for the genre of interest. The second column in Table 5 shows which sex requested more songs of each genre.

Table 5.

Sex differences in the likelihood of requesting a genre.

Genre	Direction	χ²	p	OR	OR-CI	RR	Cramér's V
A capella	m > f	2.416	.120	1.939	0.909–4.135	1.913	.046
Classical	m < f	<0.001	1.000	1.070	0.286–4.000	1.069	.003
Electro	m > f	2.691	.101	1.467	0.953–2.259	1.432	.047
Eurodance	m < f	<0.001	1.000	1.284	0.214–7.710	1.283	.007
Funk	m > f	.333	.563	1.956	0.517–7.352	1.948	.025
German Schlager	m < f	.048	.827	1.715	0.313–9.392	1.711	.017
German songwriter	m > f	.733	.392	1.365	0.919–2.026	1.333	.041
Hip-hop	m > f	13.383	< .001**	1.966	1.374–2.814	1.840	.100
Indie	m < f	1.058	.304	1.353	0.809–2.260	1.335	.031
Oldies	m > f	5.324	.021*	5.920	1.293–27.119	5.845	.069
Pop	m < f	32.269	< .001	1.854	1.500–2.291	1.346	.153
Reggae	m < f	0.265	.607	1.546	0.516–4.637	1.540	.021
Rock	m > f	9.240	.002**	1.870	1.259–2.777	1.780	.084
Singer-songwriter	m < f	.733	.392	1.722	0.643–4.615	1.711	.029
Theme song	m > f	<0.001	1.000	1.081	0.490–2.385	1.079	.005
Traditional	m > f	.004	.950	1.171	0.437–3.138	1.169	.008
World	m < f	.265	.607	1.546	0.544–4.362	1.540	.021
Worship	m > f	<0.001	1.000	1.169	0.164–8.326	1.169	.004

The table gives: χ² statistic and p; OR with 95% confidence interval (OR-CI); relative risk (RR); and Cramér's V. Significant Holm–Bonferroni corrected correlations are indicated.

* p < .05; **p < .01).

Four genres were requested significantly more often by either boys or girls. As the RR shows, boys were 84% more likely to request hip-hop, nearly six times (484,5%) more likely to request oldies, and 78% more likely to request rock than girls. According to Chen et al. (2010), the OR of 1.966 for hip-hop (with 7% requests in the control group of girls) corresponds to Cohen's d = .300, and the OR of 1.870 for rock (with 6% requests for rock in the control group) corresponds to Cohen's d = .290, indicating small effects. For oldies (with less than 1% of oldies requests in the control group), the OR of 5.920 would be comparable to Cohen's d = .750, indicating a medium effect size. On the contrary, Cramér's V is exactly at or below the threshold for small effects of .10 proposed by Cohen (1988). However, this measure is based on the χ² statistic, which is susceptible to low cell frequencies (Kateri, 2014) and therefore less reliable, as the success cells for the 2 × 2 contingency table only contain less than 10% of all requests (which are further split into boys in girls) in the case of all three genres (Table 2). Yet it should be kept in mind that the association, in particular between the children's sex and the likelihood to request oldies, is based on a very small sample size, and the lower bounds of the confidence intervals are below the OR benchmark of 1.5, implying Cohen's $d<.20$ in all three cases, and making the observed effects somewhat uncertain.

Furthermore, girls are about 35% more likely to request pop. Here, the cell frequencies were more balanced, and indeed, Cramér's V confirms a small effect. On the contrary, the interpretation of the OR is not as straightforward, because Chen et al. (2010) define their benchmarks for outcome-of-interest-rates in the control group of up to 10%, whereas in these data, 45% of boys’ requests (control group) were categorized as pop (outcome of interest).

Finally, there was no significant difference in the number of requests for children's music between boys and girls: χ² = .396, p = .529, OR = 1.077, OR-CI = .871–1.331, RR = .958, Cramér's V = .018. Also, Welch t-tests showed no significant effects of the children's sex regarding the Spotify audio features: valence, t(1,366.2) = .836, p = .403; energy, t(1,353.7) = .347, p = .715; acousticness, t(1,359.3) = .350, p = .727; danceability, t(1,306) = –1.444, p = .149; instrumentalness, t(1,385.5) = 0.146, p = .884; liveness, t(1,285.3) = 1.791, p = .073; speechiness, t(1,353) = 1.597, p = .110.

Exploratory Results

In further exploratory analyses, children's age and sex were compared with other, nonmusical properties of the songs, such as the year of release, and the number of times a song was played on the radio show. A statistically significant, but practically negligible, effect of children's age on the release dates of the requested song was observed, F(1, 1,163) = 5.44, p = .020, R² = .005, Cohen's f² = .005.

Furthermore, a linear regression model indicates an association between increasing age and the number of times that a requested song was used on the radio show, F(1, 1,200) = 11.59, p < .001, R² = .010, Cohen's f² = .010, but, again, the effect size implies that this observation is of little practical significance. Also, a Welch t-test suggested that girls request songs more frequently used in the show. t(1,407.5) = –4.597, p < .001, Cohen's d = .242. However, a subsequent analysis of variance (ANOVA), testing both children's sex and genre as effects on the number of times that a song was played, revealed that genre had a higher effect, F(18, 1,392) = 6.776, p < .001, η_p²= .075, rather than sex F(1, 1,392) = 13.469, p < .001, η_p²= .010, whereas within the genres there was no clear pattern of either boys or girls requesting more or less frequently played songs.

There were also significant differences in the singer's sex between boys and girls (p < .001), with Cramér's V = .267. Table 6 lists the numbers of requests made by boys and girls, respectively, in relation to the vocalists’ sexes. For both boys and girls, most requests were for songs with male vocalists, but, for boys, the difference between the number of requests for songs with male and female vocalists, respectively, was considerably larger. Girls, however, made more requests for songs with mixed vocalists.

Table 6.

Vocalists’ sexes in requests made by boys and girls.

	No vocalists	Mixed vocalists	Female vocalists	Male vocalists
Requests by boys	4	26	115	506
Requests by girls	7	42	313	399

Finally, we used multiple logistic regression to test the association between Spotify audio features and the likelihood of a song in our sample being categorized as children's music. The model revealed significant positive effects of valence (b = 2.202, p < .001) and speechiness (b = 4.380, p < .001), and a significant negative effect of energy (b = −2.459, p < .001). Liveness (b = .058, p = .927), danceability (b = 1.046, p = .137), and acousticness (b = .434, p = .356) had nonsignificant positive effects, while instrumentalness (p = .075) had a nonsignificant negative effect. For the model, R²_MFH was 147.

Discussion

In this study, we analyzed 1,412 freely expressed music requests for a children's radio show, comparing children's age and sex with the requested songs' audio features and genres. Significant associations with age and the songs’ valence and genres were observed, as well as sex differences regarding the frequencies of certain genres in the requests.

Effects of Age on Songs’ Valence

The first hypothesis, that younger children would request more happy (i.e., positive-valence) music, was confirmed. However, this decrease in valence appeared to be connected to fewer requests for children's music with increasing age; children’s music also typically exhibited higher valence than the requested adult music. Even though the songs tagged as children's music spread over 13 musical genres, the audio features of energy and speechiness were also indicative of a song's categorization as children's music. This implies that there is some kind of common denominator beyond genre and high valence, such as an increased presence of singing—since the lyrics establish the child-directedness of the song. However, the audio features tested here explained only a small portion of this common denominator.

For children aged 4, children's music was the most requested type of music. This is in agreement with the findings of an ESM study by Lamont (2008), who found that the same was true for the music that 3-year-olds chose to listen to in everyday situations. Interestingly, the two following years of age (4–6) show the steepest decline of requests for children's music (i.e., before and within the first year of school). This decline might be confounded with a parallel increase in sample size (see the discussion on limitations to the study) but, according to Dalla Bella et al. (2001), in this period, children learn to assign affective values to structural musical qualities, such as mode or tempo, and their ability to recognize different emotions in music improves (Cunningham & Sterling, 1988). In this process, they might acquire abilities to appreciate music that is not explicitly directed at them, based on an increasing sensitivity to musical expression of emotion. Here, positive emotions would still be more popular, as children are not yet capable of dissociating negative emotions of sad music from its aesthetic qualities (Nieminen et al., 2012).

To conclude, even though H1 was confirmed, the song requests were still overall at high valence levels and the observed slight decrease in valence was mostly a result of the children requesting less children's music. The reason for this might be that children start to understand the aesthetic qualities of music even if it does not address them as explicitly as children's music does; this is possibly a precursor of the enjoyment of serious or sad music later on in adolescence and adulthood (Londsdale & North, 2011; Vuoskoski & Eerola, 2017). In the course of this development, children might also start to become aware of the suitability of a song for the radio show. A requirement for this awareness is understanding other peoples’ perspectives—an ability that, according to the theory of mind, is acquired between 6 and 8 (Wellman, 2002). In fact, there is anecdotal evidence of a child calling again and explaining that a previously made request was probably not appropriate for the radio show. This awareness could contribute to the persistence of overall higher valence in the requests of older children. Also, the children's parents or other caregivers might bias the song selection in different ways. First, they might usually introduce their children to positively valenced, supposedly child-appropriate, music and thus bias the children's musical repertoire in general. Second, they could actively encourage their children to select such music for the request, or practice other forms of control, or even censorship.

Finally, assuming that children are excited about making a call for a song request (as their call might be played on air), mood congruency effects (Bower, 1991), which have been observed in children as young as 3 years old (Christodoulou & Burke, 2016), could bias the song selection toward happy music.

Genre Distribution at Different Ages

As the younger children in the sample develop aesthetic judgments of music, their requests appear to be more evenly distributed across different genres. Toward the end of elementary school (which is at age 10 in Germany), pop music became the predominantly requested genre, even though the statistical effect sizes for associations with age and the likelihood to request pop were fairly small and clearly below the threshold of Hemmert et al. (2018) for a good model fit.

When children enter puberty, they become more accultured to normative standards; this manifests in a stronger tendency to request certain musical styles, especially popular contemporary music. This kind of music also received higher preference ratings in studies using previously selected musical excerpts (Finnäs, 1989; Hargreaves et al., 2016; Kopiez & Lehmann, 2008; LeBlanc et al., 1996). It is furthermore consistent with the findings of Busch et al. (2009) that pop music has already become the most popular genre a few years before adolescence and includes more than half of all requests among 7-year-olds.

The development of musical style sensitivity is accompanied by a shift from parent-oriented toward peer-oriented musical preferences (Troué & Bruhn, 2000). As during adolescence, musical preferences concentrate on a few popular genres, the parents’ musical preferences can be expected to be more widely spread than those of adolescent peers. Perhaps this also affects the music children request on a radio show, and the higher genre diversity at younger ages might thus primarily be a result of younger children requesting more songs they have been introduced to by their parents.

In this particular dataset, there is a third possible source of influence on the repertoire of songs a child might request, which is the radio show itself. It has already been mentioned that some children were apparently aware of different musical styles’ suitability. At the same time, other music played on the show might affect the general musical preferences of its listeners and serve as inspiration for song requests. The minor effect of age on the number of times a request was played on the show implies that the role of the radio show for the selection of a request does not change much with the children's age.

Analysis of Sex Differences

Overall, sex differences found in previous studies on self-reported genre preferences (Christenson & Peterson, 1988; Colley, 2008; Hargreaves et al., 1995; P. A. Russell, 1997), also seem to apply to single song requests, as boys tended to request rock music, while girls rather tended to request pop music. From a restrictive perspective that considers the OR's confidence interval, boys’ observed preference for rock requests is somewhat uncertain. One reason could be the already mentioned awareness, for some children, of the requests’ suitability, which might have reduced the number of requests for rock songs in general. Furthermore, boys’ preference for rock music has been explained by the masculinity associated with this genre (Colley, 2008; Hargreaves et al., 1995), and preference for rock music has been associated with expressing individual identity (Gardikiotis & Baltzis, 2012) which was found to be more important to boys (North et al., 2000). Given the young age of our sample, compared with the referenced studies conducted among adolescents or young adults of 11 years and older (Christenson & Peterson, 1988; Colley, 2008; Hargreaves et al., 1995), it seems plausible that the stereotyping of rock is still in its development.

Boys were also more likely to request hip-hop; however, 83.7% of the hip-hop requests were children's music, so this can hardly be regarded as evidence for a generally higher preference for hip-hop among boys. Also, the (statistically significant) sex differences in the requests for oldies should be interpreted with caution, as there were only 12 requests.

Furthermore, boys requested more songs with male vocalists. Again, there are similar findings regarding preferences for preselected stimuli by Busch et al. (2009), who found excerpts of different singing styles to be more appreciated by boys when sung by male singers, and did not observe such a preference for same-sex singers among girls; this also corroborates the results of this study. According to North et al. (2000), boys are more invested in communicating their identity through music and might favor music performed by the same sex more clearly, whereas girls rather use music for mood regulation, which is expectably less affected by the performers’ sex. Again, these structures of musical preference are supposedly in the early stages of development in this sample, but older boys might at least understand song requests on the radio more as a display of musical identity than girls, and conclusively give more importance to selecting male music for their request.

Another explanation is based on evidence of similar differences between boys and girls regarding the sex of role models (Ivaldi & O'Neill, 2008). Bromnick and Swallow (1999) explained this finding with a higher acceptance among adolescents for girls having male role models and idols than vice versa, which has also been observed in a broader context of social cognitive theory (Bussey & Bandura, 1999). Possibly, similar principles apply to the musicians that children and adolescents listen to, and, perhaps even more, to songs they publicly request on a radio show.

We found no differences between boys and girls regarding the songs’ valences, contrary to findings by Hunter et al. (2011) and Sharp et al. (2006). This might be attributable to the overall high valence in the requests, as discussed previously.

In nearly all statistical models, the effect sizes were in the lower spectrum of the benchmarks recommended in the literature. However, especially for these data, it does not seem appropriate to understand small effect sizes as indicators of no or little evidence of the impact of the children's age and sex. As these are only two of many personal and situational factors that affect the selection of a song request, one would hardly expect them to have any more than a small influence on the request's genre or other audio features.

Limitations to the Study

The examination of individual song requests on a radio show poses some noteworthy limitations to this investigation. For instance, in the requests, children utter musical preference by choosing one song out of a pool of other songs they know. As these other songs remain unknown, the requests do not provide any information on which music these children like or dislike in general, and can only be aggregated to infer the popularity of musical style characteristics among the radio show's audience.

There is also no information on the children's motivation to make a request. Musical preferences, as such, are subject to different psychological functions (Schäfer et al., 2013), and making a request on a radio show introduces a particular situation in which, for instance, social functions could be particular important. Conclusively, there might have been reasons to select a certain song other than personal liking. As mentioned already, there was at least some awareness of the suitability of a song to be broadcast on the radio; possibly, the observed genre spectrum might be a somewhat incomplete representation of the music that children at different ages would request if there were no such constraints.

Another core limitation is the age distribution in our sample. The oldest children were at the beginning of adolescence, where the formation of identity through musical preferences is probably just starting; it remains an open question if, for example, preferences for single genres, such as pop or rock, would further consolidate in request patterns of older adolescents. As children become adolescents, they probably stop listening to this children's radio program and prefer other forms of musical engagement. Thus, toward the upper limit of the age range, our sample might also be biased toward children who have not yet entered these developmental stages.

Additionally, the number of requests at each age varies considerably (since the older children drift away). Some studies on children's musical preferences analyze shifts in preference patterns between distinct ages or grades at school (Kopiez & Lehmann, 2008; Nieminen et al., 2012). In this study, only trends over the entire age spectrum can be observed, as comparisons of single age groups would be confounded by their respective sample size.

In the analysis, it was furthermore assumed that there is only one request per child but, according to the production staff, there were a few exceptions of children calling several times. From a statistical point of view, this introduces repeated measurements that would require the use of linear mixed-effects models. However, as this affected only a few children, modeling random effects would not have been feasible, even if several requests from the same child could have been reliably identified.

Another uncertainty is introduced by the scarce information that Spotify provides on how exactly the genre information and audio features are computed, apart from the verbal descriptions given in the methods section. Panda et al. (2021) found the valence, energy, and acousticness features to be valid descriptors of perceived musical emotion, but the accuracy of the other features and the used genre information could not be validated before this research.

Finally, this study was conducted with an exclusively German-speaking sample. Especially in light of the importance of children's music, which is predominantly German, the findings of this study are thus not only limited to German-speaking children, but also to the musical culture in German-speaking countries.

Outlook and Conclusion

This study was an analysis of the effects of children's age and sex on song requests made on a radio show. Such requests are a fairly special case of uttered musical preference, which is influenced by many factors beyond age and sex that could not be controlled here. Yet the association between genre frequencies and children's age and sex revealed popularity patterns that agree quite well with findings on children's and adolescents’ individual musical preferences for different genres. In other words, several known effects of age and sex on musical preferences appear to generalize to the music requests that children made on this radio show.

The results also emphasize the importance of children's music at younger ages, which has been rarely addressed in previous research. In this context, it would be an interesting subject for future research to investigate these kinds of song request in comparison with both the children's self-reported reasons to make the request, and the parents’ musical taste, allowing for a further validation of our findings.